VDI Calculator

Hello and welcome to the New VDI calculator.

To access the new Java calculator just click the green button below.

Click-Here-Button-psd50511

 If you want to know more about the VDI calculator options and features read the Manual at http://myvirtualcloud.net/?p=1927 or check out the VMware View Calculator Video Training at http://myvirtualcloud.net/?p=2551(the video is still based on the old version on the calculator).

Permanent link to this article: http://myvirtualcloud.net/?page_id=1076

253 comments

Skip to comment form

  1. Andre Leibovici

    Release 6.0
    - Horizon View and XenDesktop Infrastructure Calculation
    - Java Code Signing
    - Write Cache Sizing

    Release 5.0
    - Multi Desktop Type pool support

    Release 4.2
    - Added backend feature to support multi-pool and multi-vm calculations. (Coming Soon!)
    - Added backend functionality to support a feature where you will be able to ask for help from VDI vendors.
    - Couple User Interface improvements

    Release 4.1
    - Added auto version checking.
    - Added vendor logo display when sponsored calculation is selected
    - Fixed host high-availability calculation issue
    - Fixed “Desktop per Host” unit when using Print option
    - Few UI improvements

    Release 4.0
    - Support for hyper-converged infrastructure deployments
    - Little fixes to validations

    Release 3.0
    - Added support for 3rd party calculations

    Release 2.0
    - Added support for XenDesktop

    NEW Java Calculator Released – version 1.0

    Release 3.0
    - Added support for pools with 2,000 desktops
    – Added support for new processors with 12 and 16 cores
    – Added support for user-defined VM limit in vCenter
    – Added support for user-defined vSphere cluster size
    – Added support for 10,000 desktops with vCenter 5.1
    – Multiple backend changes to support Horizon View 5.2
    – Multiple backend changes to support vCenter 5.1
    – Enabled high availability features as Default options
    – Removed support for vCenter Server earlier than 5.0
    – Removed support for View earlier than 5.0
    – Removed cluster size dependency on storage protocol

    Release 2.9

    - Increased number of allowed VMs per CPU/Core to 20 (maximum supported is 16)
    - Added support for manual CBRC Base % IOP savings
    - Added support for vSphere 5.1
    - Added framework to support future VMware View features
    - Added additional validations according to VMware compatibility matrix
    - Fixed layout issues

    Release 2.8.1
    - Added support for 12 trough 24 core microprocessors
    - Small bug fixes

    Release 2.8
    - Added VMware View 5.1 support
    - Added support for VMware View CBRC System Disk (Content Based Read Cache)
    - Added support for NFS Linked Clone vSphere clusters with more than 8 hosts
    - Added support for scale up architectures with high density servers
    - Added support for 2560×1600 display resolution
    - Small UI changes

    Release 2.7.1
    - Fixed issue with Total Backend IOPS when Storage Read Cache was in use
    - Small UI changes

    Release 2.7
    - Added support for storage array caching calculation capabilities
    - Added support for storage block de-duplication
    - Added configuration validations based on storage protocol (Block and File)
    - Added support for VAAI (vStorage API for Array Integration) validations
    - Added higher degree of control over refresh cycle for floating pools
    - Small UI changes

    Release 2.6
    - Added additional field to demonstrate Total Frontend IOPS
    - Added support for RAID6 configurations
    - Few UI changes
    - Fixed issue with % Host Used Memory

    Releae 2.5.1
    - Added information about Frontend IOPS
    - Re-added Print to PDF feature

    Release 2.5

    - Added higher degree of control over read/write ratios
    - Added segmentation between number of sockets and cores per socket
    - Added storage overhead support for swap on host
    - Fixed video swap issue with earlier versions of vCenter and ESXi
    - Complete overhaul of the IO calculation
    – Calculation of IO per disk type (replica, clones and persistent)
    – Calculation of IO for of simultaneous boot VMs

    Release 2.4.1 (FIX)
    - Added support for 20 cores/host
    - Fixed issue with size of video swap file when 3D is not selected

    Release 2.4

    - Added support for VMware View 5.0
    - Added support for hosts with 24 and 128 cores
    - Added support for 1000 desktop per desktop pool
    - Added support for vCenter 5.0 and 10K validated architecture
    - Added support for 32 bit display resolutions
    - Added support for vSphere 5.0
    - Added support for 3D
    - Several improvements to the calculator structure
    - Small UI changes

    Known Issues
    - Calculation of View 5.0 3D vRAM using VMs with 3 or 4 vCPU not available

    Release 2.3
    - Added support for 1-to-Many relationship between ParentVM and Desktop Pools.
    - Added support for calculation of Full Clones.

    Release 2.1
    - Added support for VMware View 4.6 architecture
    - Added support for MHz per VM
    - Added host/core MHz calculation
    - Added support for CPU MHz overhead
    - Fixed bug when displaying Linked Clone LUN size with Local .vswp in use (Thanks to Andrew Lash for reporting the inconsistency)
    - Minor UI adjustments

    Release 2.0
    - Added support for VM Swap files in local storage
    - Added support for Connection Server High Availability (N+1)
    - Minor UI adjustments

    Release 1.9
    - Added support for Printing
    - Added IOPs calculation based on Workload IOPS, RAID Type and Read/Write ratio
    - Re-positioned fields that relate to each other
    - Minor UI adjustments

    Release 1.8
    - Added field to show number of VMs per host
    - Fix for storage consumption calculation when using Suspend Power Policy

    Release 1.7
    - Added dropdown menu to allow selection of number of vCPUs per VM
    - Added dropdown menu to allow selection of number of displays per VM
    - Added dropdown menu to allow selection of maximum display resolution per VM
    - Added automated calculation of virtual machine memory overhead

    Release 1.6
    - Added dropdown menu to allow selection of maximum number of VMs per vCenter. Only applies to vCenter 4.1.
    - Changed from Dropdown menu to free entry of Number of VMs per Datastore to allow more than 64 VMs.
    - Fixed bug when calculating delta sizes whilst Refresh OS on logoff set to 0%.
    - Fixed calculation of vCenter Servers when number of VMs was equal to vCenter Server limit.

    Release 1.5
    - Added support for Dedicated Datastore for Replicas
    - Added support for use of VM memory reservation
    - Added support for Cluster with HA (N+1)
    - Added information on number of clusters required
    - Fixed bug with overcommitment memory calculation

    Release 1.4
    - Added support for .vswp calculation with Linked Clones
    - Added configurable Hypervizor memory overhead
    - Added configurable VM memory overhead

    Release 1.3
    - Added support for VMware View 4.5 Disposable disks
    - Added Storage Breakdown summary for VMware View 4.5 Tiering

    Release 1.2
    - Added Host number calculation based on Desktop State when Not in Use (On, Suspend, Power Off)
    - Added Storage calculation based on Desktop State when Not in Use (On, Suspend, Power Off)
    - Added 4Gb overhead for vmKernel when calculating Host Memory with Overcommit
    - Fixed bug when calculating Hosts Memory overcommit with low number of VMs per core

    Release 1.1
    - Added ERR validation for Number of Desktop Pools (max. 512 Linked Clone VMs per Desktop Pool)
    - Added developer name and blog URL at bottom

    Release 1.0

  2. Daniel Whittaker

    Nice calculator Andre. One question: How are the “ESX Host memory – no overcommit” and “ESX Host memory – overcommit” calculating their numbers? For a medium deployment, 500 desktops, I’m getting a larger amount for the “overcommit” value than the “no overcommit” value.

    The factor that determines a change in those values seems to be the VMs/core value (associated with the cores/host value). At lower VMs/core values you end up with more memory required when “overcommited” then when not. Calculation error, or working as intended?

  3. Andre Leibovici

    @Daniel Whittaker
    The %Memory allocated is based on %Used Memory * %Shared Memory. To be precise (100-%Used Memory) * (100-%Shared Memory).
    I was not able to simulate what you tell me. Have you filled all fields?
    In normal conditions the %Used Memory = 80% and %Shared Memory = 40% or 50%.

    Would you be able to send me a screenshot of the scenario you describe. Please send to aleibovici@myvirtualcloud.net

  4. Andre Leibovici

    @Daniel Whittaker
    Thanks for the note. As you can see on Release Notes 1.2 I have fixed the issue.

  5. Graham Zulauf

    Would you be willing to publish the formulas behind the results column? I’d like to recreate the calculator in Excel.

    1. Andre Leibovici

      I won’t be publishing the formulas as the whole calculation is a bit complicated, but if you have any doubts I can publish an explanation.

      All calculations are based on published VMware reference architectures and guidelines for vSphere and VMware View.

  6. sam

    hi,

    when I change memory the “ESX Host Memory ” memory changes, but when I change the number ov VMs, it stays the same, thanks

    Sam

    1. Andre Leibovici

      Hi Sam and thanks for your comment.

      Host memory is a resulting equation between VM memory size and the number of VM’s per core and number of CPU sockets per host. The number of VMs for the total solution does not influence host memory itself. In some cases host memory is also affected by the number of concurrent VMs and the powering option, such as suspend VMs.

      Try changing the number or sockets and VMs per core.

  7. Miguel Angel Alonso

    A great calculator!!!. Only one thing: When calculator calculate the LUN numbers doesn,t use the best practices where The LUns must to have 500 GB maximum RAID 5 (4+1). If you can to modify this would be perfect.

    Thanks a lot of

    see you later

    1. Andre Leibovici

      Miguel Thanks for you comment.
      The size of the LUN is a resulting equation between the final size of the virtual desktops and the number of VMs per datastore. You may configure the number of VM’s per datastore using the drop-down menu to reduce the size of the LUN. VMware recommends a maximum of 64-128.
      BTW – Can you point me to the 500 GB best practice documentation?

  8. Miguel Angel Alonso

    Hi Andre. This is a DELL and EMC recommendations of the LUNs for Server and Dektops VMware. I imagine that is a IOPS theme and proofs with their disk array. Mike Laverik speak in his new book (VMware View 4,5 on LULU Web) about it.

    Tomorrow I,ll give you one or two links about it.

    Thanks

    Its a great work

    Congratulations

  9. Robert Kadish

    Andre,

    Great calculator! One question, I’m wondering where the input for IOPS is?

    1. Andre Leibovici

      Robert, I haven’t yet included an IOPS calculator, however you would have to know the numbers of IOPS utilised by your GuestOS.

  10. robert Kadish

    Andre what we are finding is that you can use tools like Veam or even native monitoring tools in the hosts to identify the maximum IOPS requeired per host and per VMDK. We have found that max IO per desktop is not really relevant.

  11. Suresh Thoppay

    When Desktop state when not in use is selected as suspend, I don’t see the VM storage size increasing.

    ESX suspend file This file, which has a .vmss extension, is created if you set the desktop pool
    logoff policy so that the virtual desktop is suspended when the end user logs
    off. The size of this file is equal to the size of guest RAM.

  12. Andre Leibovici

    @Suresh Thoppay
    Suresh, thanks for you comment.
    I have updated the calculator and published an article based on your comment. Read at http://myvirtualcloud.net/?p=1405

  13. Vasko

    Nice tool, but there is something wrong in the calcs – the ESX Memory results are dependant on the “Cores per Host”, not on the actual number of VMs running on the cluster. So, if I use 2 x Dual Socket Hosts with 6C Xeon CPUs running only 100VMs, do I still have to install 200+GB of RAM per host?

    1. Andre Leibovici

      @Vasko thanks for you comment.
      The actuall memory results are dependent upon a number of factors – CPUs per host, cores per CPU, VMs per Core, Memory per GuestOS, Memory Overhead per VM, Kernel Memory Overhead, % Transparent Page Sharing and the amount of memory you set as avaiable to be utilised, as an example 80%. VDI if normally memory bound so you will always run out of memory before you run out of CPU (unless you keep adding memory if supported by your server architecture).

      If you run 96 VMs per host in a Dual x 6 cores (Total 12 Cores) you would need 160GB RAM assuming the following:
      - Windows 7 Guest OS with 2GB RAM
      - 80% Maximum RAM utilisation
      - 10% TPS

      If you run the same configuration with Windows XP and TPS set TPS to 40% you would need 112GB RAM per host.
      I recommend you to run your calculations again as I believe you have missconfigured some settings. Send me the screenshot if you wish.

      regards

  14. Trc

    Nice tool!!!..grt work!!

    when ever my desktop count is increasing accordingly my connection server and Vcenter server count is not changing. it i sdisplaying an error.

    can you fix that.

    Regards
    Saty.

    1. Andre Leibovici

      @Trc
      The numbers or Connection Brokers and vCenters will only change when upper limits are reached.
      Try 2001 users and don’t forget to change the number of Concurrent Users to 2001 as well. You will see that 2 vCenters and 2 CS are required.
      The calculator will only use the Concurrent Users for the purpose of this calculation.
      Regards

  15. Pat

    So for 512 VM’s, all running at the same time with PCOIP enabled I just need 7 hosts?

    LOL!

  16. Pat

    Just noticed there isnt any difference between RDP and PCOIP at all according to the calculator… so where are all these CPU cycles magically dissapearing to?

    Also Linked Clones don’t scale after more than 64 VM’s (it’s VMWare’s own “best practice!”) so getting 96 VM’s on a host might be way above stretching it.

    VDI isnt simply put into a calculator.

  17. Andre Leibovici

    @Pat it will depend on the amount of cores and VMs per core you are running. If you plan to run around 12 VMs per core the number is 7. If you are running 16 VMs per core the number of hosts would be 5. Now, be aware that you may need to test your workload before assigning 16 VMs to a processor core.

  18. Andre Leibovici

    @Pat You must be doing somehting wrong. There are no CPU cycles change when changing between PCoIP or RDP.

    In regards to Linked Clones you have old information. Newest architectures from VMware have mention to up to 128 VMs per datastore. The 64 VMs was a safety number because of SCSI locks. Also, the limit is not per host as you mention, it’s per datastore. ESX 4.1 can host up to 320 VMs and you may do that if you wish and if you have enough memory for that. However, be carefull with downtimes when 320 VMs may go offline at the same time.

    In my view as long computing is mathematics you would still be able to calculate using a Calculator.

  19. Randy

    Andre, Great calculator.

    You’ve got all the major technical decisions for high level planning in there which frees up my brain to do more infrastructure planning and equipment ordering rather than always doing the calculations on paper (which always impresses people).

    Reading all the comments you certainly know your stuff!

  20. Andre Leibovici

    @Randy
    Thanks Randy, and I’m glad the calculator is helpfull to you.

  21. Alif

    HI,

    Just a quick question.

    In my environment, I am using Dell M1000e chassis with Poweredge M610 blade servers with the following details:
    Dell powerEdge M610
    CPU cores: 8 CPUs x 2.26 Ghz
    Processor Type: Intel Xeon CPU E5520 @ 2.27 GHz

    Currently, I am keeping a ratio of 20 VMs per host. But, I find that you are using around 90+ VMs in each host. Even when I make 30 VMs, and when I try to enable DRS/HA, gets an error, insufficient resources. I am running Windows 7 (normal) VMs.

    I would really appreciate, if you can help me to come up with the correct number of VMs in each host, without .

  22. Andre Leibovici

    @Alif
    The Insuficient resources error message most likely has nothing to do with the hosts itself, but the way you have configured your cluster and/or resource pools. Make sure you have enough capacity in your cluster to support a host failure, or that the resource pool is set to expandable (not recomendable in production environment), for tests purpose. Also, verify that you have enough RAM memory to cater for all virtual machines.

  23. Alif

    HI Andre,

    Thanks for your quick and prompt response.

    Please note that each host has 64-96 GB RAM. So, RAM is not an issue. I am concenred about the CPU cycles. And, it is just normal Windows 7 VMs (3GB RAM, 1vCPU). I have created a cluster of 7 ESX hosts, for a total of 150 VMs.

    Again, depending on this scenario, how many VMs can I host on each ESX host. Or let me rephrase it, Is 3 hosts more than enough for 150 user VMs (running Windows 7)?

    I have another implmentation of 500 users, soon. So, I would really appreciate if you can guide me on this. I would like not to over-provision, at the same time dont want bottlenecks. Please note that cost is not an issue., but the performance is.

  24. Andre Leibovici

    @Alif
    For Win7 with 2GB RAM and disconsidering any savings from TPS you may host aproximately 40 VMs (5 VMs per CPU/Core), if you hosts is a Dual-Quad. You will need 5 hosts (N+1) for your VDI solution, however I recommend you to assume some savings from TPS, something in between 10% and 20%. Something else to consider is making sure that you leave 10%-20% free RAM per host. I think it’s safe to assume that you hosts can host 48 VMs with 2GB RAM. Now, if you want performance and cost is not an issue, go for 40 VMs per Nehalen host. Also, CPU is not an issue for Nehalem processors even when running 128 VMs in a single host.

    Unfortunately the calculator will not compute 3GB RAM due the lack of CPU overhead numbers based on this amount of RAM. You may use either 2GB or 4GB.

    PS: Read my article on Multi-Core architectures for VDI. http://myvirtualcloud.net/?p=1519

  25. Alif

    Thanks very much

  26. Martin

    Great piece of work! Thank you for publishing it online.

    I would like to know can I use this calculator for VMware Lab Manager storage sizing? I read it uses also Linked Clones.

  27. Andre Leibovici

    @Martin
    I am not 100% familiar on how Lab Manager does it’s provisioning process. I know Linked Cloning is used, however I recommend you to refer to the installation and architecture guide for VMware Lab Manager.

  28. dlink7

    Hi

    Could work on the calcualtor. You have Windows 7 default to 10% for TPS, Is this what your are seeing out there?

    -Dwayne

  29. Andre Leibovici

    @dlink7
    Long story short – this is a long discussion and I am preparing an article with results around that. It is possible to reach 40% TPS for Windows 7 in some circumstances, however other times not possible. I’ll be addressing the question soon. In the meantime I recommend you to run your own workload tests.

  30. iwan 'e1' ang

    Excellent tool!
    For the IOPS, what’s the normal distribution among the 3 components (replica, persistent disk, disposable disk). For example, if the total IOPS is 3000, what would be the value of each component?

    Many thanks from Singapore!
    e1

  31. Andre Leibovici

    @iwan ‘e1? ang
    That’s the million dollar question, and the response is: you need to measure your environment. To start with I recomend you to read my article “Get hold of VDI IOPs, Read/Write Ratios and Storage Tiering” to understand the touch points and how the behave according to the RAID penalty, Read/Write ratio, LUN Size, IOPS etc.

  32. DAB

    Great job. Have you thought about accounting for the database requirement for vCenter/Composer/Event in your calculations? Specifically, in your calculations for compute, memory and storage requirements of the VM for the server. It would be good to see the requirement captured for a SQL or an Oracle installation. Thanks again.

  33. Andre Leibovici

    @DAB
    It’s a good idea and I might include database calculation in a future release. As of this moment I am working on Validated Profiles (Cisco, HP, Dell). Thanks for the feedback.

  34. Randy

    Andre, thanks for this tool. Any issues with linking to this (and some of your other writings) for an paper I am writing for school? I am interested in using some of your thoughts/opinions and research you have done for a paper I am writing for class, and while I will be citing, for linking to a tool like this I like to make an attempt to get permission first. Also I would be interested in a conversation about a scaled up approach we are working on as it comes to our vmware view host infrastructure. While I certainly agree with the under utilization of cores, there are several things I have run into that pushed us in this direction. Feel free to contact via the email I provided. Thanks Again!

  35. Oli

    Great Tool. Just to get it right: For my 120 VM calculation I got a result of 3000 IOPs total (RAID 5). Does this mean I need a total of 24 SAS-Disks (125 IOPs) or how do I have to understand this value?

  36. Andre Leibovici

    @Oli
    The number od disks in a RAID group is defined by the amount of IOPS required for workload (or vice-versa). The calculator does not tell you how many disks are required and I suggest you to talk to your storage vendor or administrator. However, the 3K IOPS you got from your calculation is only based on individual machines (full clones), RAID type and %Read/Write. There is more to it when using Linked Clones and I suggest you to read my article “Get hold of VDI IOPs, Read/Write Ratios and Storage Tiering” at http://myvirtualcloud.net/?p=1421.

  37. Ben S

    Andre,

    Nice work! I have a question on the Delta, when set to 100% of the Refresh on Logoff the Delta is == to the Master Image size, in reality do you think this would be closer to: Full Master Image Size – Thin Master Image Size = Delta?

    Not looking for a change, as I realize it’s correct, just an opinion.

  38. Andre Leibovici

    @Ben S
    The 100% refresh rate should be interpreted as “Never”. I have yet to rename the dropdown button. The delta size in this case will grow up to the size of the provivioned disk. Yes, Full Master Image Size – Thin Master Image Size = Delta, however this is only from a NTFS perspective. From a VMFS perspective, any writes will add to the size of the delta. I recommend you to read my article “Sizing Storage for VMware View Linked Clones” http://myvirtualcloud.net/?p=745. I am also running some tests around delta growth and will be able to soon publish an updated version.

  39. Dirk R

    Great work!

  40. Jordan

    I’m running into an issue, maybe my math is horrible.

    For my 10,000 concurrent Desktops I’m giving each Virtual Desktop 2 CPUs from the hosts. The host has (2) Quad-Core 2.93GHz CPUs. So in the calculation above I have average vCPU set to 2930Mhz (2.93 x 1000) and then 2 vCPUs.

    Running about 8 VMs per Core I get 64 VMs per box. The calculation is telling me that each host needs to have 56256Mhz or 56Ghz per box. On a Quad core box, each CPU would need to be 7.03GHz ([56256/1000]/8), which isn’t possible. Seems to be there should be a limit cap of whatever the max amount of GHz a box can handle at this point (3.60Ghz seems to be the max on Dell.com) where it would tell you, you need another host. Basically I would need double the amount of hosts recommended because 7.03Ghz per CPU isn’t possible.

  41. Andre Leibovici

    @Jordan
    Apparently you are using the physical CPU speed to calculate your VM vCPU. If you plan t run 8 VMs (1vcpu) per physical CPU/core you need to divide 2930Mhz by 8 = 366.25 MHz. If you plan to run 8 VMs (2vcpu) per physical CPU/core then you divide by 16 and you will have 183.12Mhz per vcpu. Most likely this will not be enough to support your multi-threaded applications if all VMs use all resources at same time. In this case you have two solutions:
    1 – Reduce the amount of VMs per physical CPU/core
    2 – Leverage ESX advanced CPU scheduling

    If your choice is number 2 you need to make sure you are not overcommitting your system beyond acceptable boundaries that will affect end user experience. I personally prefer to assign a single vCPU to VDI VMs and treat only the exceptions adding multi-SMP capabilities. I hope that helps.

  42. Andre Leibovici

    @Dirk R
    Thanks for the feedback.

  43. Jordan

    @Andre Leibovici
    Thanks for explanation, I knew I was calculating wrong. This is my first VDI deployment, so I appreciate the help.

  44. Andrew

    Andre,

    Are you able to update the calculator to include a field for the number of parent VMs required. Your calculations are assuming there is a parent VM per pool which is not necessarily the case as multiple pools can utilise the same parent VM.

    Andrew

  45. Andre Leibovici

    @Andrew
    Thanks for the feedback. I will try to include the feature request in the next release. I hope things all going allright in your project in NZ. Let me know if you ever visit my new shores.

  46. Andrew

    Thanks Andre. Send me an email and we can chat offline.

    Andrew

  47. Andre Leibovici

    @Andrew
    As per your request I have included the support for 1-to-Many relationship between ParentVM and Desktop Pools. Thanks for your sugestions.
    Andre

  48. Keith

    Well done and many thanks!

  49. rowdycholin

    Andre,

    What block size are you using to determine the IOPS? Also, what if any, write penalty are using for each RAID level

  50. Andre Leibovici

    @rowdycholin
    The calculator does not deal with block sizes at this stage. Block sizes will typically indicate the maximum sustained bandwidth based on X number of IOPS. Therefore, block size should be considered for sizing your backend array, however after you have the total number of IOPS. May I suggest you reading “Improve VDI performance with IO Length Trending” at http://myvirtualcloud.net/?p=988.

    Thanks

  51. wee kiong

    it will be able to print this out as similar to this update

  52. Ronny Steiner

    Hi Andre,

    your VDI calculator is really awesome!
    Our customers are now starting to purchase the IBM x3850 X5 servers with the new 10 core E7 CPU’s (Westmere-EX generation). Could you please add 40 Cores to the “Cores per Host” option?

    thank you very much!

  53. Andre Leibovici

    @Ronny Steiner
    Thanks for the feedback.
    I’ll include 40 cores support in the next iteration of the calculator.
    Remember that there will be a sweet spot on number of VMs per core. ESX4.1 support 320 VMs per hosts and ESXi5 supports 512.
    The maximmum number of vCPU per core supported by VMware View today is 16, however ESXi5 supports up to 25. However you migh run out of GHz beforehand.

  54. Wolfgang Motal

    @Andre Leibovici

    Maybe you could add the 10-Core processor stepping to your selection of “Cores per Host”.

    Best regards,
    Wolfgang

    Processor
    One or more of the following depending on model
    Ten-Core Processors
    Intel® Xeon® E7-4870 (2.40GHz/10-core/30MB/130W) Processor
    Intel® Xeon® E7-4860 (2.26GHz/10-core/24MB/130W) Processor
    Intel® Xeon® E7-4850 (2.00GHz/10-core/24MB/130W) Processor
    Intel® Xeon® E7-8867L (2.13GHz/10-core/30MB/105W) Processor
    Eight-Core Processors
    Intel® Xeon® E7-8837 (2.67GHz/8-core/24MB/130W) Processor
    Intel® Xeon® E7-4830 (2.13GHz/8-core/24MB/105W) Processor
    Intel® Xeon® E7-4820 (2.0GHz/8-core/18MB/105W) Processor
    Intel® Xeon® X7560 (2.26GHz/8-core/24MB/130W) Processor
    Intel® Xeon® X7550 (2.0GHz/8-core/18MB/130W) Processor
    Intel® Xeon® L7555 (1.86GHz/8-core/24MB/95W) Processor
    Six-Core Processors
    Intel® Xeon® E7-4807 (1.86GHz/6-core/18MB/95W) Processor
    Intel® Xeon® E7540 (2.0GHz/6-core/18MB/105W) Processor
    Intel® Xeon® E7530 (1.86GHz/6-core/12MB/105W) Processor
    Intel® Xeon® X7542 (2.66GHz/6-core/18MB/130W) Processor
    Quad-Core Processors
    Intel® Xeon® E7520 (1.86GHz/4-core/18MB/95W) Processor

  55. Wolfgang Motal

    Wolfgang Motal :@Andre Leibovici
    Just seen the repsonse about 10-Core systems.

    Maybe you could show most actual entries first.

    Wolfgang

  56. Andre Leibovici

    @Wolfgang Motal
    Thanks for the feedback. I have just uploaded a new version of the calculator that is mostly a fix for the video swap size. However, I have added calculation for 20 core/host.

  57. Derek

    A similar calculator for XenDesktop 5.5 would be most useful.

  58. Pablo

    Hi Andre,
    This calculator is a bless.
    It Would be nice to have a “Hosts Core MHz” limiter, as not to get 5000ghz or things like that, this could be an easy alternative to the test and error changes we make to the amount of VMs per core, just an opinion.

    Thank you so much.

  59. Dasa

    Hi, thanks a lot for a great tool, I am just starting to use it and not sure how to interpret the results. I would appreciate an aswer, if the host memory is per host or the overall memory needed, meaning:
    Example:
    Hosts: 8
    Host Memory – no overcommit: 112 GB
    Concurrent VMs: 720
    So the overall RAM of all hosts that I would need to accomodate given number of concurrent VMs (720) would be 112*8 GB? Thanks a lot in advance.

  60. Andre Leibovici

    @Dasa
    You need 8 hosts, each host with 112GB RAM.
    The overall RAM is not important since you need to sice each host iondividually.

  61. Dasa

    Andre Leibovici :
    @Dasa
    You need 8 hosts, each host with 112GB RAM.
    The overall RAM is not important since you need to sice each host iondividually.

    Thanks a lot, Andre, for quick reply. Everything is clear now. I am really thankful.

  62. John Dias

    Andre, great tool thanks for publishing. One question related to powered off when not in use. I entered 500 total desktops with 250 concurrent users. Selecting “powered off” when not in use doesn’t seem to change the host RAM requirements. Is that intentional? (i.e. are you assuming a worst case that all VMs would be powered on?)

  63. Andre Leibovici

    @John Dias
    Host RAM requirements are based on the number of VMs per Core, ammount of RAM per VM and TPS ratio. Powering VMs off does not change RAM requirements; however lower the numbers of hosts required to support the workload. For 500 total desktops with 250 active, the number of hosts will only support 250, however the amount of VMs per hosts is determined by VMs per Core.

    Andre

  64. Rodrigo Nuñez

    Hello Andre,

    First of all Thx a lot for this magnificent tool its awesome!! but I have a doubt with the 3D enable…

    I need to make a deployment for a University with 150 VM’s and only 100 will be concurrent, my question is…

    Are the students going to be able of executing apps like Autocad, Macromedia Studio, Corel Draw 11, Netbeans and not having performance issues?

    In your experience does VMware View 5 is already prepared to support this kind of applications that demand a lot of CPU, RAM and Network usage??

    Thx in advance and have a great Weekend!!!

    1. Andre Leibovici

      Rodrigo Nuñez,

      They will defenetely be able to execute such applications. It boils down to the amount of bandiwthd you will have to assign to each user.
      However, you are running the VMs in your local LAN that should not be much of an issue.
      In regards to RAM/CPU consumption, follow the vendors recomendations or run your own validation tests.

      Andre

  65. mikob

    Hello,

    Why I can’t calculate with 3 Gbyte Memory size? If I calculate 3 Gbyte Memory size than value of Host Memory – no overcommit and with overcommit = NaN Gb.

    1. Andre Leibovici

      mikob,

      VMware has not published VM memory overheads for VMs with 3GB RAM, thefore I am not able to calculate using them.\
      Maybe in the future is I have time to reverse engineer the implementation.

      Andre

  66. Marcelo

    Hi André,

    If you allow me for a suggestion, you could add the option to export and/or print the choices taken + results to a worksheet or pdf file to your tool.

    Cheers,

    Marcelo Vieira

  67. Celia

    Please include the option for 6 cores :D
    Excellent sizing tool! Thanks!

  68. Derek

    Please let us configure maximum physical RAM per server, so that the number of servers that you calculate reflects what our hardware can support. It’s great the calculator tells me I need 216GB per server, but what if our maximum is 144GB and we don’t want to overcommit?

  69. Frederik

    Would it be possible to let the Calculator work in reverse?

    Say; I’ve got theese hosts in my cluster, how many machines can i put on them?

  70. Andre Leibovici

    @Marcelo
    The calculator used to have an option to print to PDF, however I removed upon finding a bug. I will work to add the option back to the calculator.

  71. Andre Leibovici

    @Celia
    Thanks for the feedback. What type of server and CPU have you got that you require 6 cores per host?

  72. Andre Leibovici

    @Derek
    You may control the size of your hardware in few diferent ways that make more sense from a calculation standpoint.

    - Change VM memory size
    - Change Host Transparent Page Sharing
    - Change number of VMs per core

    It makes more sense because the calculator will not guess any of the items that Must be defined by the user.

  73. Andre Leibovici

    @Frederik
    Please see my response to @Derek

  74. Raj

    Andre,

    Thanks for all this, I have a quick question, I think the total IOPS output is very high. with your default 768 VMs, it’s coming as 42931.

    Almost 56 per VM, I think it’s way too high , please help me why do we need 56 IOPS per VM. I tried your calc 2 months ago, and the IOPS output was correct.

    Thanks,
    Raj

  75. Andre Leibovici

    @Raj
    You need to provide me with more details, perhaps some screenshots.
    The number of IOs is defined by you, not by the calculator. Send me the screenshot to andre.leibovici@yahoo.com.

    Andre

  76. Dan Turner

    Hey Andre,

    Could you explain briefly the difference between “frontend” and “backend” IOPS in the Linked Clones section of the report?

    Thanks.

  77. Andre Leibovici

    @Dan Turner
    Frontend represent the number of IOs transversing from the host to the storage array. Backend represent the number of IOs required from the spindles due to the write penaly caused by use of RAID protection. May I suggest you to read my article “Get hold of VDI IOPs, Read/Write Ratios and Storage Tiering” where I discuss write penalty.

    http://myvirtualcloud.net/?p=1421

    Thanks
    Andre

  78. umarzuki

    Hi Andre,

    Why do I need 2 clusters if I’m going to have 2000 VMs with only 1000 concurrent VM

  79. Andre Leibovici

    @umarzuki
    The number ofclusters are based on the number of hosts, not the number of VMs. Wyhen using Linked Clones the maximum host number ina cluster is 8. You probably have more than 8 clusters to run all your 1000 virtual desktops.

    Thansk,
    Andre

  80. Jim Nickel

    Hiya,

    Very nice!

    However, one problem I see is that most implementations will have both some linked-clones and some full-clones. There will always be some use cases that require a full-clone. I think you should have a section to allow for both.

    Jim

  81. Derek

    Does the calculator take into account the usage of CBRC (content based read cache) that’s available with ESXi 5.0? That factors into/reduces the number of physical reads required from the array.

  82. Andre Leibovici

    @Derek
    Despite available in ESXi 5 CBRC has not been officialy released nor integrate with VMware View. VMware tells me CBRC will be avaiable for use in View.next. Once the feature is released I will also release an update for the calculator.

    Andre

  83. Andre Leibovici

    @Jim Nickel
    As today you will have to run two calculations in paralel on the calculator.
    However, I will put the request on the log for future versions of the calculator.

    Thanks,
    Andre

  84. Biogazownie

    Nice Tool. TXH!

  85. Brett Dunn

    Hi There,

    When using the VMware View Calculator, there is mention that the calculator can be used for any VDI solution running on top of vSphere infrastructure.

    Considering my client wants to virtualize their infrastructure and desktops using vSphere, but use CITRIX Xendesktop as the desktop manager / controller server, which calculator would be more suited?

  86. Andre Leibovici

    @Brett Dunn
    For XenDesktop deployment I recommend you to use my XenDesktop Calculator (http://myvirtualcloud.net/?page_id=2303). The View Calculator has more features, and unless you fill comfortable using the View Calculator and converting the results to XenDesktop yourself may use the View Calculator for XenDesktop. The infrastructure calculations are very similar, but the XenDesktop calculator has different labels and lack of features like Dedicated Image Stores.

    Andre

  87. HoosierStorageGuy

    Andre,

    I’m seeing some discrepancy in the total calc of backend IOPS when I change the Read Cache Hit %.

    Assuming Read Cache Hit = 0%, I have total of 4770 IOPS on Replica and 4200 for Linked Clones. The Total backend IOPS shows 8970, as would be expected.

    If I change Read Cache Hit = 90% (assuming FAST Cache), then the Replica IOPS go down to 477, and the Linked Clones IOPS go down the 3660. However, the total backend IOPS calc shows 8430. It looks like instead of adding 477 and 3660, it added 4770 and 3660.

  88. Andre Leibovici

    @HoosierStorageGuy
    Thanks for reporting the bug. I have included a fix for the next release of the calculator.

    Regards,
    Andre

  89. Ivo

    That’s really a great calculator!

    Can you help me to interpret some results?

    INPUTS:
    Number of VMs: 100
    Concurrent VMs: 100
    Memory Size: 2048MB
    % Host Memory Used: 90%
    % TPS: 10%
    % VM Memory Reservation: 0%
    Connection Server HA: Yes

    RESULTS:
    Hosts: 3
    VMs per Host: 96
    Host Memory: 216GB
    Total Frontend IOPS: 3897
    Total Backend IOPS: 6897

    QUESTIONS:
    a) The reason that “Number of Hosts” is 3 depends on the “Number of VMs per Host” which is 96 (dependent on CPUs, number of Cores and number of VMs per Core). If I would calculate with just 96 concurrent VMs, then the number of Hosts would be 2. That’s everything clear! But my question is, why the “Host Memory” doesn’t change (in case of 100 versus 96 VMs / in case of 3 versus 2 hosts) – it is still 216GB which I also don’t need in case of one host failure (if one host fails, there would still be 432GB memory available for just 100VMs). I think it should be just the half of 216GB or what’s the reason therefor?

    b) I have measured some different Storage Arrays with IOMeter and different Access Specifications (%Read/Write, Block Sizes, Random/Sequentiell, ..) and the values are – of course – VERY different according to this Access Specifications. From your VDI-Calculator I now get some IOPS needs. Under which Access Specifications should I get the number of IOPS calculated by VDI-Calculator?

    For example some IOMeter-Results for an HP P4500 Storage Array:
    BlockSize / %Read / %Random => measured IOPS
    4KB / 100% / 0% => 28’000 IOPS
    4KB / 20% / 75% => 2’600 IOPS

    According to these IOMeter results, do I need just one of this Storage Array or two? or even more? Do I have to reach the calculated Frontend or Backend IOPS?

    c) Why does only the Backend IOPS change when I change the “Delta % Read IO” and not also the Frontend IOPS of the Linked Clones?

    Thank you very much for your support!

  90. FT

    Really nice calculator !!!

    Following things I want to understand which will really help me to understand VDI more

    (1) Disposable disk
    (2) Persistant disk
    (3) Full size of parent VM
    and
    (4) thin size of parent VM

    Appericiate your help in advance.

  91. Andre Leibovici

    @FT
    Please refer to my article “VMware View 4.5 Linked Cloning explained”
    http://myvirtualcloud.net/?p=1222

    Andre

  92. Kyle

    I visited this page in Chrome and was looking all over to try to find the calculator itself. I had to use IE to get it to work correctly. Just thought you might want to let people know this.

  93. Andre Leibovici

    @Kyle
    Thanks for the feedback.
    I am not planning to update for Chrome, but rather create an app for iPad.

    Andre

  94. Bozo Popovic

    Hi Andre,

    there is a bug in calculator when trying to allocate more than 2048 of RAM for vm. The sizing of the target server memory is displayed as NaN. I Have tried with other values but anything greater than 2GB is not possible. Is this left intentionally or you can fix it.

    Thank,
    Bozo

  95. Andre Leibovici

    @Bozo Popovic
    Thanks for your note. Yes, this is a known issue.
    The reason being is that VMware has not published memory overhead tests for virtual machines with more than 2GB RAM. I’ll look into this, but most likely will need to reverse engineer memory overhead. Please keep in tune as I’ll try to fix that as soon as possible.

    Andre

  96. elizabet

    Hi everybody I want bye license for 200 user desktop in VDI which type of and how many licenses I must prepare ?

  97. Andre Leibovici

    @elizabet
    I would recommend you to start looking at the user type you are catering your VDI solution for. However, in saying that VMware View is licensed per concurrent user and any type of deployment will be covered by the license. As I starting point I also recommend you to read VMware View 5 Desktop Virtualization Solutions. http://myvirtualcloud.net/?p=3435

    Andre

  98. Bonny

    @Andre Leibovici
    Hi Andre,

    I am trying the tool for linked clone model. How is the delta sized? It does not seem to change. I have tried 5, 50, 100 VM and Memory 2, 4 and 8. delta still remains the same. But I saw you formula of delta = VM Memoryx3.

    Bonny

  99. Bob

    Hi Andre,

    I’m confused with the terms Frontend IOPS and Backend IOPS used. Can you clarify the terms? Is it used in the same context as total IOPS and Write IOPS?

    When I do some basic calculations for example:

    10 – 15K RPM drives (175 IOPS)

    10×175 = 1750 IOPS Read

    1750 IOPS / Raid 6 Penatly = 292 Write IOPS

    Thanks

    Bob

  100. Andre Leibovici

    @Bonny
    The deltas are a combination of the total size of the desktop (thick provisioned) and how often you refresh that desktop. If your Parent VM is 40GB than if your refresh cycle is configured for 10% your Delta is 4096MB. If refresh cycle is 30% your Delta is 12288MB. If you have configured your desktop pools to refresh on logoff you will need to identify, on average, what is the size of the delta when your users logoff at the end of the day or shift. Play with “Full Size of Parent VM” and “Refresh OS disk on logoff at” settings.

    Andre Leibovici

  101. Andre Leibovici

    @Bob
    Frontend IOs are the requeste that arrive at the array storage processors coming from the hypervizor. Backend IOs are the requests coming coming from the array storage processors and hitting the disks. Depending on the RAID configuration you have a different backend number of IOs.

    Total IOPs = Read IOs + Write IOs

    I hope that helps.

    Andre Leibovici

  102. Dzmitryj

    Dear Andre

    excellent job

    Can you please explain the meaning of Linked Clones (LC+Disposable+Log). Am I right LC means delta disk ?
    will it be possible to set maximum HOST Core MHz in future release ?

    Thank you

  103. Dzmitryj

    Dear Andre

    excellent job

    Can you please explain the meaning of Linked Clones (LC+Disposable+Log). Am I right LC means delta disk ?
    Will it be possible to set maximum HOST Core MHz in future release ? Sometimes ) get 4400 and even more MHz per core

    Thank you

  104. Markus

    A Question about Linked Clones vs. Full Clones: In real Life, I see that a deployment using Full Clones consumes less IOPs than a deployment with linked clones (in steady state production, when stressed with Login VSI Workload) That seems correct to me, because Overhead IOP’s are needed to determine which *.vmdk has the acual Block that is demanded (is the one in the replica the newest, or is it served from the delta disk). This is the Same as with VMWare Snapshots: stressing a VM that has Snapshots leads to more IOP’s in the Storage than on a VM without Snapshots.
    However: your Calculator seems to not reflect that, right? Are you aware of that, and planning changes? Besides that: great Work :)

  105. Andre Leibovici

    @Dzmitryj
    Yes, LC (Linked Clones) is the Delta, also called REDO Log.

    The calculator work in an inverse order. You define the number of desktops and number of vCPU and MHZ – and the calculator will give you the required hardware you should buy. The other way around would deprecate the ability to set essential parameters from a sizing perspective.

    Andre

  106. Andre Leibovici

    @Markus
    Markus, you are absolutely spot on.

    The calculator does not reflect that because it’s not possible to know the exact IOPs difference between Full Clones and Linked Clones unless you run parallel tests with the real/production workload. Instead, to overcome the IO difference, I give you the ability to specify the number of IOPS per VM in the calculator. So, if you know that you will be using Linked Clones and that will use more IO, you should feed that into the calculator.

    I hope that makes sense to you.

    Andre

  107. PhilS

    This is a really cool tool as noted by all of the previous commentors! :-) I’m having trouble getting the storage results to match up with our real-life deployment. Perhaps you can contact me offline and we can figure out where the trouble lies: either my inputs, or the calculations themselves. Currently, I’m being presented with a result on the calculator roughly 3x the space than what we’re actually using in View 5.x. Looking forward to your response! Keep up the great blog! -PHIL

  108. Andre Leibovici

    @PhilS
    If you provided the correct number to the calculator the output will be the maximum storage/cpu/memory commitment. If you have less capacity utilized than the calculator is providing you it’s because clones are still growing as an example. Print out the calculator to PDF and send me (andre_leibovici@yahoo.com) the file along with detailed explanations about your environment.

    Regards,
    Andre

  109. shenawy

    The calculator is trimmed, I can’t see the whole 3 buttons at the end.

  110. Robs

    Andre,

    First of thank you for such an informative site. I was looking at your VDI calculator and was wondering a few things that I hope you can clarify.

    The VMs per core only goes up to 16 on the calculator. According to the VMware Configuration Maximums for 5.1 the maximum Virtual CPUs per core is 25. (page 2.) In your article “A Review of VMware View 5.1 Limits and Maximus” you state that the limit of 16 VCPUs per core did not change. So I am not sure if it should be 16 or 25.

    Next issue I have with the calculator on the number of clients per host. I plug in 16 VMs per core and with a 2 processor server with 8 cores each (16 cores total) that limits me to 256 VMs per host. However that number does not change if I put 2 or more vcpus on each client. Should that number of virtual machines per host reflect the number of vcpus that I configure ultimately?

    The reason I ask was that I have a 2 processor box with 8 cores each for a total of 16 cores and using the limitation of 25 vcpus from Vmwares guide I can have 400 vcpu’s. Since I am placing dual vCPU’s on each client I was only able to power on 200 since any more gave me an error about the number of vcpus that I could have.

    Can you clarify since I do not think I can put 256 clients on a host with 2 processors with a total of 16 core. I think that number would be more like 128.

  111. Andre Leibovici

    @Robs
    The calculator is based on Limits and Maximums for VMware View. The supported number of desktops per core with VMware View 5.1 is still 16. However, this is just the VMware validated number and doesn’t mean you are not allowed to have more desktops/core if you use case and environment supports it. I will increase the number of desktop/core for the next release of the calculator, giving the ability to create solutions beyond VMware support statement.

    For your 2nd question – the number of desktops per host will not change when you add a 2nd CPU. However, pay attention to the Host Core Mhz, that will change indicating that you will need more CPU for that given environment. You need to make sure the amount of Host Core Mhz fit into your environment. If you want to run 2vCPU you could reduce the amount of desktops/core or reduce the amount of MHZ per desktop. There are multiple ways to play with the values.

    I recommend you watching the video manual I have created http://myvirtualcloud.net/?p=2551

    Regards,
    Andre

  112. Andre Leibovici

    @shenawy
    Thanks for letting me know. What browser and OS are you using?
    I’ll try to fix for the next release of the calculator.

    Andre

  113. Thomas

    In result, What is host core MHZ and how is it calculated? If I need VM’s to have 3 GHZ how many Vm’s can I have per core? How is this calculated?

    1. Andre Leibovici

      Thomas,

      When you specify that each desktop must have 3GHz vCPU, the number of vCPU and the number of dekstops per pCPU/core, than the calculator will provide you with the server hardware required for your solution. Sometimes you need to play with the numbers until you are able to fit the solution in a server you already have in-house.

      I recommend you watching the video manual I have created http://myvirtualcloud.net/?p=2551

      Cheers
      Andre

  114. Patrick COMO

    Hi,
    Concerning CBRC Base, could you explain more about this? I cannot find anything about this percentage and i cannot find what it is.
    Thanks a lot for your good tools.

    Patrick

  115. Chris MacDonald

    Has the tool been removed? I can’t find a means to download it. Help.

    1. Andre Leibovici

      Chris, I see no problem with the calculator. The calculator is to be used online, not downloaded. Make sure you have Flash installed.

      Andre

  116. Tom Brown

    Hi Anrde,

    Awsome tool, have a look at the View 5.2 option with replicated datastore sizing. Having a dedicated replica LUN does nothing to the Replica Total Usable capacity. Is this an issue with the calculator or has the View 5.2 replica requirements changed?

    Thinking about deploying 5.2 in my project. Do you have a recommended cluster size? If you had twenty hosts in a block would you put them in one cluster or split them up? Do larger clusters add any overhead or LUN contention issues?

    Cheers

    Tom

  117. Oscar Chacon

    Hi Andre,

    Great tool you’ve implemented, it’s exactly what I’m been looking for: not just a calculator but a design tool!

    I’m been reading and testing it for about 2 days (I’m a newbie) and I’ve been working in VDI solutions design for 5 years; so far this is the best tool I’ve found. VMware itself doesn’t have a tool like this. Thank you so much.

    I’ve found something interesting about this tool: when you try to input a VM “Memory Size” of more than 2048MB (lets say 2049) and leave any other parameter with default values, the output in both “Host Memory” fields is: “NaN”; and if I try 2048 or less, then both “Host Memory” fields are populated with some numerical values. I’ve try different combinations when populating all other fields, but so far, every time I’ve tried any value above 2048 in the “Memory Size” field, the outcome is the same: “Host Memory” = NaN.
    I don’t know if this is a “user problem” or a minor bug; most probably, a “user problem”; any way just wanted to share this behavior I’ve found in this excellent tool of yours.

    Regards,

  118. Andre Leibovici

    @Tom,

    In relation to the dedicated replica datastore, if you load the calculator and simply enable use od dedicated replica datastore you will notice that the total usable capacity goes from 0.45TB to 0.04TB. The reduction is because the calculator is not placing a replica disk in each datastore for each desktop pool.

    In regards to Cluster size I recommend going large if you can, because or the reduced number of servers for N+1 high availability. In saying that it all comes doesn’t to use case, RPO, RTO etc. If you follow VMware tested guidelines and observe LUN latency thresholds you should not face issues.

    Andre

  119. Andre Leibovici

    @Oscar,

    The calculator is based on real number tested and provided by VMware. When talking about CPU overhead (memory has impact on it too) VMware did not specify the overheads for uncommon memory sizes. For this reason the calculator will not give you any results.

    Andre

  120. Daniel

    Excelent!

  121. Julian Botero

    I know you are no longer supporting the XenDesktop calculator but this seems to work fine for calculating that as well. My question is related to the “Number of VMs per Datastore” variable. Your other manual says “VMware recommends a maximum of 64-128″. Can you point me to that actual reccomendation. I havent found a concrete number but i have found in Cirtix XenDesktop documentation where they state “it is a best practice to implement no more than 20-30 virtual machines per LUN on average.” If you could point me to a different statment from VMware it would be appreciated. Link to Citrix paper below.
    http://support.citrix.com/servlet/KbServlet/download/17979-102-671614/Introduction%20to%20Storage%20Technologies.pdf

  122. Andre Leibovici

    Julian, the soft imposed limits relate the number of concurrent SCSI reservations in a FC or iSCSI LUN. That does not apply to NFS protocol. Those are safety numbers that have been tested and approved by vendors. Can you go beyond those numbers? Yes, absolutely. However, when doing so, make sure you keep track of storage metrics, such as latency, ensuring the users still have a good experience when using the solution.

    -Andre

  123. Sittichai Palanisong

    Hi Andre,

    I have a partner who used the VDI calculator for sizing of View 5.1. When they changed the version to Horizon View 5.2 somehow the storage usable capacity is increased from 16.25TB to 22.7TB even all parameters are the same.
    What is the change in 5.2 that cause this difference?
    -Sittichai

  124. Andre Leibovici

    Sittichai Palanisong, yes, things have changed from View 5.1 to View 5.2 due to additional storage resources utilized for advanced video processing. I recommend reading the article below. Thanks.

    3D Software Rendering with View 5.2… What’s Changing!
    http://myvirtualcloud.net/?p=4927

  125. Sittichai Palanisong

    @Andre.

    In this case if I disable 3D rendering then the storage sizing should be the same as in 5.1, right?

  126. Julian Botero

    @Andre in regard to my question which you replied to on 05/03/2013, I found this quote in Citrix XenDesktop Best Practices document. Just thought I would share…
    LUNs accessed by means of Fibre Channel or iSCSI should not host disks for more than 20-30 virtual machines, in order to avoid read/write latency caused by SCSI reservations or SCSI queuing. For environments based on VMware vSphere 5.x and a VAAI-capable storage array, the maximum number of virtual machines per LUN should not exceed 75.

  127. Filipe

    Hi Andre,

    When using linked clones, replica storage is higher than I expected for View 5.1 . For example I’ve used 15 Desktop Pools, 100 VM per datastore (200 total) , 1 snapshots per pool and thin parent VM size of 50 GB to end up with 1.57 TB replica storage. By changing the number of VMs per datastore makes replica storage change (64 VMpDS results in 3.15TB). Correct me if I’m wrong but I was not expecting this for versions after View 4.5 because apparently replica storage size should not depend on the number of datastores.

    Based on http://www.vmware.com/files/pdf/view_storage_considerations.pdf and http://myvirtualcloud.net/?p=4093

    Regards,

  128. Tristan

    typo:

    Deskop Pools Type

    :)

    But cool calculator

  129. Pantana Na Takuathung

    Hi Filipe
    A replica of the master image is automatically placed into each LUN that you select to host the linked-clone virtual machines in
    your Horizon View desktop pool. Changing number of Datastore will effect total replica storage size, correct me if I am wrong

  130. Filipe

    I agree with your comment and there is info to support it.

    “The replica size, equal to the master image size, is taken into account on a per-LUN basis. A replica of the master image is automatically placed into each LUN that you select to host the linked-clone virtual machines in your Horizon View desktop pool.” (page 18)
    http://www.vmware.com/files/pdf/view_storage_considerations.pdf

    In the end it’s all about desktop pools, datastores and LUNs…
    On Andre’s VDI calculator I suppose it follows the document http://www.vmware.com/files/pdf/view_storage_considerations.pdf, to estimate the number of Luns by using the formula nLUN = lcVMs/64. According to vmware best practices 64 is a good value of lcVM per LUN. I’m not going thru all details but vmware example for 500 VMs results in 384 GB for Replica (8 LUNS) which is far different from the TB I get with a 200 VMs example.

    From several case studies I found replicas volume to be significantly less than linked clone’s storage (not a rule!).
    After reviewing my parameters for the calculator, the only way I can explain the large storage for replicas is the high value of desktop pools I used (15!). Probably something to rethink…

    Thanks
    Filipe

  131. Filipe

    java version typo?

    At storage section description “Clone Frontend IOPS” is duplicated, 2nd line should be “Clone Backend IOPS”.

    Fil.

    1. Andre Leibovici

      Filipe, thanks for that. I noticed that while working on the new release and have already fixed. The new release should be out soon.

      Andre

  132. Tomas

    One question related to Replica Frontend/Backend IOPS. If we change number of concurrent desktops for example from 200 to 1000,it doesn’t change the Replica Frontend/Backend IOPS requirements. Is the number of concurrent users does not affect replica IOPS? I noticed what it starts increases only from 8000 desktops. Thanks

  133. Tom Queen

    Andre – maybe missing something here, but seems when HA is selected, I’m still seeing 1 host selected. In previous editions, when HA was selected, I always started at 2 hosts, no matter how small (or so memory seems to indicate).

  134. Andre Leibovici

    Tomas, the replica is mostly utilized only during boot storms and the number of IOs is directly related to the number of desktops booting at the same time. This two part series article will provide you with the backgournd on that http://myvirtualcloud.net/?p=2084. Additionally, if CBRC is enabled you will get even less IOs hitting the replica disk.

    Andre

  135. Andre Leibovici

    Tom, in this new version HA is a function of the cluster size. If you reduce the cluster size to 2 or 3 you will get the desired outcome. I understand what you say and I will see if there is anything I can change to ensure that we have N+1 even when cluster size is set larger than the actual number of hosts.

    Andre

  136. gothicreader (@gothicreader)

    Hello,
    Do you think this calculator is good to use for VMs operating as a system sizing and not necessarily for VDI?

  137. sam

    Guys,

    have anyone doing some research on the host connectivity required to x number of users.

    example host server 4x1G vs 4x10G

    thanks

    Sam

  138. Andre Leibovici

    Sam,

    The first question that comes to mind is related to your storage fabric. Are you running storage over ethernet (NFS/iSCSI) or SCSI?
    Independent of the option, 10GB with a redundant 10GB should give you more than enough bandwidth for your workload. In saying that, you need to look at application behaviour.

  139. Andre Leibovici

    gothicreader,

    The calculator is focused on VDI and vSphere features only enabled when used in conjunction with View. However, yes, you may use for general workload calculations.

  140. Sam

    Andre,

    have few remote sites that is hard to get a solution going for cheap 100-200 users, on the main site with 1000+ users I”m ok with full SAN and 10G connectivity.

    for the remote sites have choosen a couple of DL380 2x8core 256GB fusion io card, the server right now will be using 4 1G’s links with vdswitch setup for physical load – load balancing, will play around with that and see what kind of performance I can get out of them (my next possible upgrade on the servers will be hooking a pair of 10G but though to start with that it would be overkill) oh and the goal is to have around 100 users per server.

    thanks again for the great calculator :)

    Sam

  141. Davoud Teimouri

    Thanks for nice tools.
    Please increase CPU MHz average.

  142. Davoud Teimouri

    Also number of vCPU.

  143. suliman

    Unable to open it.. plz help…

  144. Andre Leibovici

    Davoud,

    As for the CPU MHz you can change according to your requirements.
    What other option around vCPU would like to see? 4 vCPU?

  145. Andre Leibovici

    Sulivan,

    You need Java 7 installed in your machine in order to launch the calculator.

  146. Oswald Luraghi

    Andre,
    Love the calculator, great job! A question though about the vCPU calculation and how many VDI VM’s are calculated per host. In the calculator, it does not seem to matter if I choose 1 vCPU or 2, the number of VM’s per host does not change, nor does the amount of hosts required.
    My Host configuration is 2 CPU / 8 core
    VDI VM Configuration is 2 vCPU / 4GB Ram
    I would expect the number of VM’s per host to decrease if I increase the number of vCPU’s in each VDI. Is this by design? If you could shed some light on this I would appreciate it.

    Thanks,
    Oswald

  147. Andre Leibovici

    Oswald,

    Adding vCPUs will not change the number of VMs per host; instead it will increase the CPU clock rate required for that VM and host.
    You must set the number of vCPUs you need and then look at Host CPU MHz required. In many cases you will end up with something like 6000 MHz and the you will need to scale down the number of VMs per core to a reasonable configuration that can be obtained on the market.

    The reason for that is because there are new CPU clocks almost every 3 months.

    -Andre

  148. Oswald Luraghi

    Thank you for the reply. I looked at the calculator again and saw that if I changed the average vCPU Mhz that this also affected the configuration of the host. It looks like for this config, I need to set it to around 250 per vCPU, which seems a little low but it gives me a more real world configuration of what I have for CPU’s on the host. Also, adjusted the VMs per core down to 5.

    Thanks again,

    Oswald

  149. Davoud Teimouri

    Thanks Andre, just “Desktop per Host” unit on PDF file is wrong, GB will be printed.

  150. Adam Cook

    When I change the number of vCPUs in the desktop from 1 to 2 neither the number of hosts or the cores per host change. Is this a bug or if you double the vCPUs does the number of cores you need not change?

  151. Adam Cook

    Sorry I didn’t read the above comments. I am getting push back from my vSphere guys about how the calculator says you do not need more physical cores for double the number of vCPUs. Is there any way you can go into even more detail on this? the calculator says that 1500 desktops with 1 or 2 vCPUs still only needs 13 Hosts at 2 procs X 8 cores.

  152. Jim Nickel

    It’s because it’s not constrained by CPU. You will run out of RAM before CPU. You can likely fit 150 desktops per server quite easily, so 13 servers is quite reasonable. You didnt tell us how much RAM you have in each server, but I can tell you from experience that this calculator is a good guide. I would also encourage you to try single vCPU machines first and see what the experience is like. It is easy to add vCPUs but hard to go the other way.

    Jim

  153. Andre Leibovici

    Adam,

    Adding vCPUs will not change the number of VMs per host; instead it will increase the CPU clock rate required for that VM and host.
    You need to look at CPU clock and reduce the amount of VM’s per core until you reach your CPU clock number. The calculator doesn’t increase/decrease number of cores required; instead you tell the calculator how many cores you have. It’s the opposite.

    -Andre

  154. Carlos Luna

    Hi Andre, is there some updated documentation of the description of each field on the calculator? The currently Manual link is for previous version of the calculator. Thanks CL.

  155. Andre Leibovici

    Carlos, I am working on a new video where I explain the features. The video should go online next week.

    -Andre

  156. JD

    Noticed ‘Sponsored Calculations’ field is greyed out. What is the impact of this on the calculations.

  157. Douglas

    Hi,

    It doesn’t allow me to choose 4 vCPU per desktop, only 1 or 2. Can you fix this?

  158. Andre Leibovici

    JD, sponsored calculations is turned off because no storage company is currently sponsoring that feature. There’s no impact to your calculations.

  159. Andre Leibovici

    Douglas, I’m working on a new version that will enable 4vCPU. There are multiple validation tests to be done because of memory and CPU overhead that chafes with the VM configuration.

  160. ILYA

    Hello everyone!

    Somebody knows, why VDI Calculator doesn’t include a parameter like this: HyperThreading is Enabled/Disabled?

    If it really doesn’t include this, what should I do if my servers has a 2 sockets with 6 cores plus HT enabled? Should I set 12 cores instead a 6 cores?

    And what a value a should select for “VMs per Cores”? One tech guy told me that value for this option must be a 6. Is it correct?

    Thank you very much!
    ILYA

  161. Andre Leibovici

    Ilya, for na accurate calculation you should ignore hyper-threading.
    The number of VMs per core is dependent on your consolidation ratio and the results from your Pilot.

    -Andre

  162. Ryan Doe

    The “Persistent Datastore Size (GB)” appears to be wrong. What is the formula for calculating the size?

    When I enter 80 virtual desktops with 1024 MB (1 GB) for the persistent disk, the calculator returns “47186 GB” for the Persistent Datastore Size.
    59 MB persistent disk returns a Persistent Datastore Size of 0 GB, but 60 MB persistent disk returns 5243 GB for Persistent Datastore Size.

    Thank you for the excellent calculator!

  163. Andre Leibovici

    Thank Ryan. I am looking into the problem.
    -Andre

  164. erik

    I’m getting a conversion error on the beta when I hit update, floating point to string default value is 0,7

    Exception in thread “AWT-EventQueue-2″ java.lang.NumberFormatException: For input string: “0,07″
    at sun.misc.FloatingDecimal.readJavaFormatString(Unknown Source)
    at java.lang.Double.valueOf(Unknown Source)
    Only applies when using linked clones

  165. Rick

    will this work for smaller vendors like Virtual Bridges?

  166. Andre Leibovici

    Rick, yes, it can be used for any VDI vendor. However, you need to know how to select the right options that reflect the features available with the VDI solution.

  167. Piotr

    Is it just my PC, or the calculator won’t work anymore?
    When I click “update” after entering all parameters, it just asks me if I prefer to be contacted – and after I click “No”, I can only see zeros within all result fields.

  168. Prince

    It doesn’t work

  169. Prince

    Still Crashing…Does it still work?

    VDI Calculator
    Andre Leibovici

    The world famous VDI Calculator
    VDI Calculator
    Launch VDI Calculator

  170. Andre Leibovici

    Piotr and Prince, is this still the case? Are you using Windows or Mac? What type of calculation, Xen or View? WOuld you be able to provide some additional information as I am not able to replicate the issue? Thanks.

  171. Piotr

    Andre,
    I’m using Windows 7 x64, Java 7 build 1.7.0-b147. Mozilla Firefox 26.0, IE 8.0.
    You don’t have to change any of default settings, just click “update” after calculator starts up. It shows the “partner contact” question, and then – nothing, all result fields are filled with zeroes.

  172. Sergey

    Hello Andre,

    I have XP pro 32 / win 7 home,64 Java 7 build 1.7.0-b13, chrome 32 and i suffer from same problem as Piotr

    “just click “update” after calculator starts up. It shows the “partner contact” question, and then – nothing, all result fields are filled with zeroes.”

  173. Piotr

    I checked it with another PC with almost the same config – and the effect is also exactly the same…

  174. Andre Leibovici

    Servey and Piotr, I am having hard time to replicate the issue. I have tried with both Mac and Windows without problems. I also see people using the calculator via Google Analytics. Do you have a firewall in place that could be blocking something? Are you selecting Xen or View?

  175. Sergey

    Hello, dear Andre

    I have isa client at my xp machine and i have no any firewall at win 7 machine, result was the same. I selected View.

    Here is the screenshot of result https://drive.google.com/file/d/0Bz9lEAe2XauOMGdaRzZ3TDRiM28/edit?usp=sharing

    I changed some parameters, pressed update button and all fields: host, storage, storage detail, view had remained with zeros

  176. Meik

    Hello Andre,

    fresh start of the application, when i click on update no Result is calculated

  177. Piotr

    Andre,

    Any news?
    Could you please share a link to an older version of calculator? It’s really hard for me to do my work without your great calc :)
    Did you try to update JAVA to the newest version to replicate the issue?
    I’m connecting through few different devices from few different locations, still the same effect…

    Regards, Piotr

  178. Igor

    I have the same problem – result is not calculated

  179. Andre Leibovici

    Igot, Meik and Piotr,I have uploaded a new version this weekend. Have you tried tha calculator in the last couple days? Please make sure you launch it from the blog to ensure you are launching the latest version.

  180. Piotr

    I tried today – same issue…

  181. Martin

    I Tried today with different devices and different java version. After click update i always see the following in the java console:

    Exception in thread “AWT-EventQueue-0″ java.lang.NumberFormatException: For input string: “0,07″
    at sun.misc.FloatingDecimal.readJavaFormatString(Unknown Source)
    at java.lang.Double.valueOf(Unknown Source)
    at MainFrame.a(Unknown Source)
    at z.actionPerformed(Unknown Source)
    at javax.swing.AbstractButton.fireActionPerformed(Unknown Source)
    at javax.swing.AbstractButton$Handler.actionPerformed(Unknown Source)
    at javax.swing.DefaultButtonModel.fireActionPerformed(Unknown Source)
    at javax.swing.DefaultButtonModel.setPressed(Unknown Source)
    at javax.swing.plaf.basic.BasicButtonListener.mouseReleased(Unknown Source)
    at java.awt.Component.processMouseEvent(Unknown Source)
    at javax.swing.JComponent.processMouseEvent(Unknown Source)
    at java.awt.Component.processEvent(Unknown Source)
    at java.awt.Container.processEvent(Unknown Source)
    at java.awt.Component.dispatchEventImpl(Unknown Source)
    at java.awt.Container.dispatchEventImpl(Unknown Source)
    at java.awt.Component.dispatchEvent(Unknown Source)
    at java.awt.LightweightDispatcher.retargetMouseEvent(Unknown Source)
    at java.awt.LightweightDispatcher.processMouseEvent(Unknown Source)
    at java.awt.LightweightDispatcher.dispatchEvent(Unknown Source)
    at java.awt.Container.dispatchEventImpl(Unknown Source)
    at java.awt.Window.dispatchEventImpl(Unknown Source)
    at java.awt.Component.dispatchEvent(Unknown Source)
    at java.awt.EventQueue.dispatchEventImpl(Unknown Source)
    at java.awt.EventQueue.access$400(Unknown Source)
    at java.awt.EventQueue$2.run(Unknown Source)
    at java.awt.EventQueue$2.run(Unknown Source)
    at java.security.AccessController.doPrivileged(Native Method)
    at java.security.AccessControlContext$1.doIntersectionPrivilege(Unknown Source)
    at java.security.AccessControlContext$1.doIntersectionPrivilege(Unknown Source)
    at java.awt.EventQueue$3.run(Unknown Source)
    at java.awt.EventQueue$3.run(Unknown Source)
    at java.security.AccessController.doPrivileged(Native Method)
    at java.security.AccessControlContext$1.doIntersectionPrivilege(Unknown Source)
    at java.awt.EventQueue.dispatchEvent(Unknown Source)
    at java.awt.EventDispatchThread.pumpOneEventForFilters(Unknown Source)
    at java.awt.EventDispatchThread.pumpEventsForFilter(Unknown Source)
    at java.awt.EventDispatchThread.pumpEventsForHierarchy(Unknown Source)
    at java.awt.EventDispatchThread.pumpEvents(Unknown Source)
    at java.awt.EventDispatchThread.pumpEvents(Unknown Source)
    at java.awt.EventDispatchThread.run(Unknown Source)

  182. Andre Leibovici

    Martin, thanks for that. I know can start troubleshooting with some knowledge of what is actually happening.
    Did you enter “0,07″ in any field?

  183. Igor

    The problem may be at the authorization proxy. My proxy generates an error when java opening google-analytics.com. But all other java-applications work with proxy correctly.

  184. Andre Leibovici

    Would mind setting your regional settings and language to US to verify if problem is gone? Thanks.

  185. Igor

    It work. Change region settings to US helped! Thank you.
    But how i may use my native region settings with you app, can you fix it?

  186. Jonathan B

    hi, i have introduce data but i dont know how to finish the results. thx

  187. Andre Leibovici

    Igor, what region are you using. I want to replicate the issue.
    For now keep using US language and region.
    -Andre

  188. Prince C.

    It worked for me…however, I got a prompt saying application is blocked by security settings. So I had to go into Control Panel and reduced my Java security from the recommended high to medium and it worked. I am using Windows 8

    Thanks for the update.

    Prince C.

  189. Derek

    Curious to know how the “Refresh on logoff (%)” is being used. Is 1% always or is it 100% always as I interpret it? The reason I ask is that if you use were to calculate with 100% (which in my mind would be refresh the desktop every time logged off) the storage should be lower rather than higher by a lot. This would be using linked clones as well.

  190. Andre Leibovici

    Derek, 100% means it will never refreshed. 10% means the clone (delta) desk will only grow up to 10% of it’s maximum size. You should determine how big the delta is (on average) when users logoff, and based on that select the percentage. I hope that makes sense.
    -Andre

  191. Craig Bramley

    Great work Andre! awesome calculator :)

  192. Juliano Bongiorno

    The calculator is not working, as I have to use the old one?

  193. Tardeli

    hi andre , thanks a lot for calculator , great work ! one question , can u share programs formulas?

  194. Andre Leibovici

    @Tardeli, most of the formulas are spread throughout my blog posts as I discuss each individual technology. I have not plans to share the source code for the calculator at this point in time.

  195. Andrew Mostovoy

    Hey guys, both IE and Chrome are not allowing me to run this Java applet. Can anyone recommend a setting change?

  196. L. Wijnhoven

    Hi The calculator is not working anymore. Tried with Java JRE 6 and 7. Please correct, I love this calculator

  197. Andre Leibovici

    L. Wijnhoven, there is a known bug with non US regional settings. I have already fixed the issue and will be releasing the new version soon. For now, please change your computer regional settings and language to US.

  198. L. Wijnhoven

    Tnx!

  199. Andrew

    My settings are all US and I still get the error message about my security settings have blocked the self-signed application from running. Any suggestions guys?

  200. Andrew Mostovoy

    OK, never mind, I resolved this by adding the site address into the exceptions list on the Java 32 control panel. See link here: https://www.java.com/en/download/help/java_blocked.xml

  201. Glenn

    Hi Andre. I’m using your calculator for the Nutanix nodes and I’m curious about the “Cores per Host” metric. The 3050 and 3060 have 16 and 20 cores according to the Nutanix website but the calculator is only showing 12 and 16. Is the calculator wrong or are you “reserving” 4 pCPU for the NCVM or am I missing something? Thanks, Glenn.

  202. Andre Leibovici

    Hi Glen,

    To make it easy for users I have already included Nutanix recommendations for VDI into the calculations. The calculator assumes duplication and compression are used for any VDI deployment. With these features tuned on Nutanix engineering recommend to assign 4 cores to the NCVM. That is the reason why you see less cores available. The calculator however, will only measure the available cores to fit your virtual desktops. From a de-duplication perspective we pre-assign 70% savings, which is the average in VDI environments. You will also notice that different amount of RAM is assigned when you select Linked (24GB) or Full Clones (30GB).

    I hope that helps.

    -Andre

  203. ahmed

    i have a VDI solution, 3 host servers 2 x 10 core, memory : 144 GB and i want to assign 250 VM with the folowing specs (3GB, win7)

  204. Stef

    Hi Andre. Can you add the possibility to configure the storage for layering solution like Unidesk ?

  205. Andre Leibovici

    @Stef, I am in contact with Unidesk so see what would take me to introduce their tool into the calculator. You should hear from me soon enough.

  206. fil.

    Hi Andre,
    I can’t find a reason for the value I get at replica frontend IOps. I’m using the following settings:

    - 600 boot IOps;
    - 1 replica steadystate IOps;
    - 100 Concurrent boot;

    Assuming that these are all the parameters needed for calculations, I get 21000 replica frontend IOps. This also means 210 IOps per concurrent boot. How it relates with 600 boot IOps initial parameter?

    2nd question
    Why do I get a red negative value at local vswap disk (storage detail tab)? According to http://myvirtualcloud.net/?p=1654 it should be different.

    Thank you.
    Fil.

    1. Sriram

      Hi Fil,

      For you first question, I believe Andre is using 65% Read hit which is managed by CBRC( Content Based Read Cache).

      (Boot IOPS * Concurrent Boot)/(1-Readhit%) = 600*100/(1-0.65) = 21000 IOPS

      1. Fil.

        Now I understand.
        Thank you for your reply Sriram.

        Fil.

  207. Sam

    Andre, reason when I change to nx-3060 it says “cores per host” 16, should it say 20? thanks Sam

    1. sam

      never mind, answer was above, thanks

  208. Ian

    Andre,

    This calculator is amazing, and has been such a helpful tool for me. Kudus. It’s guru’s like you that allow the rest of us to stand on the shoulders of giants.

    On a side note, anyone know of a good calculator for server virtualization (for both converting P2V, or standards for creating new VM’s for various roles)? I know the info is all out there, but it would be awesome to have consolidated like Andre’s work.

    1. Andre Leibovici

      Ian, thanks for that. This is the kind of support that makes me keep doing it. Unfortunately I don’t know of any free calculator for virtual workloads.

  209. Jeff

    Tried opening with Chrome and IE using latest JAVA 7 update 60. I am getting an “Unable to launch the application” with error “Missing required Permissions manifest attribute in main jar: http://myvirtualcloud.net/webstart/vdi_calculator/jars/vdi_calculator_ob.jar

    Help?

    1. Andre Leibovici

      Jeff, this is a security warning. Please follow the instructions here:
      http://community.spiceworks.com/how_to/show/65737-java-error-missing-required-permissions-manifest-attribute-in-main-jar

  210. Ian

    Just open Java properties and turn security down. I had to do the same this afternoon.

    1. Jeff

      Thanks, I just added the URL to the allowed area and it works now. Thanks a bunch…

  211. Nick

    After getting through the security messages, I’m now getting an “application expired” error. Any tips?

    1. Andre Leibovici

      Nick, you should not be getting ‘application expired’. Please make sure you launch the app from the website to load the most recent one.

    2. Andre Leibovici

      Nick, you are actually correct, the application is expired. I’ll be publishing a new release tonight.

      1. Nick

        Thank you, Andre :-)

  212. Kir Grig

    Nick, ?hange time on PC a month ago as temp fix

  213. banc

    Hi Andre

    It seems that with the current version of the calculator there is a wrong number of physical cores (Nutanix) used during the calculation. To my understanding you divide the number of vCPUs through the ratio of vCPU to physical cores which should tell how many nodes you need of the chosen Nutanix model. Could it be the numbers used are not according to the spec sheet?

    1. Andre Leibovici

      Banc, if you read previous comments you will find the same question. Nutanix uses 4 vCPU and those are automatically excluded from the total amount of usable cores available to the VDI virtual desktops.

      1. banc

        Thanks for the reply! One more question: When I chose the NX1050 model it shows 20 cores per node. That would mean it has 24 cores all together which would be double from what the spec sheet says. Is that based on the usage of HT or some similar technique?

        1. Andre Leibovici

          I have already updated the code and will be uploading a new release tonight. Thanks for noticing. The NX-1050 was a last minute addition to the calculator.

  214. banc

    Andre, according to Nutanix best practices (Prism Online Help) it is not recommended to use dedupe or compression for linked clones as there is minimal space savings. Can you tell me more about your experience and why you automatically assume these features for all VDI environment no matter if its full clones or linked clones? In regards of dedupe: Do VDI environments generate many 64k IOPs? Thanks for the clarification, it’s very much appreciated!

    1. Andre Leibovici

      Banc, good observation. For Linked Clones de-dupe should be 0 since it’s all managed at the app tier. I’ll fix that soon.

      1. Andre Leibovici

        Banc thanks for the feedback. I have updated the calculator. De-duplication is now only applied to Full Clone pools.

        1. banc

          Great, thanks for the update! I opened the calculator today but I still see 70% dedupe ratio when I select Nutanix and linked clones. How much RAM is assigned to the CVMs with linked clones in this new release?

          1. Andre Leibovici

            Yes, it will always show 70% because it’s a Nutanix wide-setting for the calculator, however it is only being applied to full clones. To validate that, select Linked Clones as 100% retention and see the storage consumption results. Next, change it to full clones and see the storage consumption results.

            CVM for Linked Clones will be assigned 24GB, for Full Clones 32GB. For a mixed cluster with both Linked Clones and Full Clones 32GB CVM is used for all nodes.

  215. Sam

    good info on the compression ded-dupe option
    so these would be general good suggestions?

    I would think perf tier would help in most cases including linked-clones as the ram/ssd is being used more efficient through de-dupe, please let me know what you think on these settings, thanks

    Sam

    Linked-Clone
    (disable) COMPRESSION – DELAY
    (enable – delay 0)PERF TIER (RAM/SSD) DEDUPLICATION (FINGERPRINT ON WRITE) – DELAY
    (disable) CAPACITY TIER (HDD) DEDUPLICATION (ON-DISK)

    Full-Clone
    (enable – Delay 4) COMPRESSION – DELAY
    (enable – delay 0)PERF TIER (RAM/SSD) DEDUPLICATION (FINGERPRINT ON WRITE) – DELAY
    (enable) CAPACITY TIER (HDD) DEDUPLICATION (ON-DISK)

    1. Andre Leibovici

      Sam, I am assuming you are sizing with Nutanix. Correct? At this point in time you cannot have both compression and de-duplication enabled on the same container. The only change for your config would be that for full-clones only uses de-duplication; no compression. For user data you could create a persistent disk in a different container with compression enabled, or even offload data to a windows server in a container with compression enabled.

      1. Sam

        Andre,

        yes correct, this is for Nutanix, thanks for the suggestions, as far as de-duplication + compression, the gui allows me to select both, but I see what you mean, only one would actually be working, I’m glad that you agree that perf tier deduplication works with linked clones as I see it technically helping the performance of the box, thanks again

        Sam

  216. Sriram

    Hi Andre,

    I have 2 questions
    1) How the link clone capacity is calculated.. If we don’t add any overhead and not considering vswap capacity may I know how much percentage of parent capacity is considered for this link clone capacity.. Could you please let me know the best practice which is used in the calculator to derive the link clone capacity and the formula to have a clear understanding.

    1. Sriram

      2) Why do I get a red negative value at local vswap disk (storage detail tab) if I need to add the local vswap disk?

      1. Andre Leibovici

        Sriram, vswap is negative because it is being deducted from the total storage requirement, since you are saying that it will be stored in each host instead of shared storage.

  217. Stef

    Hi Andre,

    some post above you write: “…Nutanix engineering recommend to assign 4 cores to the NCVM.”
    Where did you find this info ?
    The upgrade guide (https://portal.nutanix.com/#/page/docs/details?targetId=Upgrade_Guide-NOS_v3_5_4_5:upgrade_cluster_upgrade_complete_vsphere_t.html) in the section “COMPLETING UPGRADE TO NOS 3.5.4.5 (VSPHERE)”, point 3b writes vCPUs: 8

    Stefano

    1. Andre Leibovici

      Stef, you are right, they have increased the vCPU count for compression and de-duplication. I’ll provide an option in the calculator for the next update.

    2. Andre Leibovici

      Stef, the number of vCPU in use for Nutanix nodes has been increased to 8.

  218. Rachit Srivastava

    Unable to use the calculator on Macbook.. Some java error props in :(

    Was working on my windows machine though.

    1. Andre Leibovici

      Rachit, what error are you getting? I have no problems running the app in my Mac. Also, check previous comments and answers about Java security.

  219. Keath Marx

    I cannot run it on any of the 3 browsers on my windows 7 machine

Leave a Reply