I finally found sometime to finalise the third and last part of this article. In the first part of this article I discussed the Business Drivers and Operational Benefits, Technology Savings and Operational Improvements of a properly deployed Desktop Virtualization project - and in the second part I talked about the underlying technology, design considerations and the site-wide architecture for a multi-site VMware View rollout.
In the meantime VMware released it’s new VDI solution, VMware View 4.0, that promises to solve latency problems that represented an obstacle for deployments over WANS and Internet. The new PCoIP remote display technology is a dynamic solution that adjusts the compression/quality level depending on the available network resources.
The architecture provided in my article can also be validated for VMware View 4.0 with few changes and I will soon work on it after the View 4.0 is released to the general public.
As discussed in second part of this article VMware View works integrated with Virtual Infrastructure 2.5 APIs. The VI3 technology has limitations and some requirements can only be achieved adding an additional parallel VI3 infrastructure or with vSphere 4.0 and VMware View 4.0.
Cluster and Server Sizing
At the time this solution was designed, the numbers of users per CPU core could range from 3.8 to 4.2, however for most VDI deployments using new processors (Intel Nehalem 5500 and AMD Phenom II) this number can be around 6.0 per CPU core, allowing up to 100 virtual desktop machines in a single dual-quad server.
This solution is based on servers with 4 sockets, 24 cores and 96GB RAM.
Note. The numbers presented in this article can change based on user profile utilisation and type of workload. I’m assuming heavy task oriented business users using applications such as Internet Explorer and Microsoft Office suite.
The solution scales beyond 2000 VMs per site/cluster, however the new vSphere and vCenter 4.0 will be required due to VI3 limitations. 2000 VM’s is the maximum number of virtual desktop machines that a single Virtual Infrastructure 2.5 instance can handle. If you are planning to have more than 2000 virtual desktops in VI3 a new Virtual Infrastructure will have to be added to the solution.
As commented by T-Rex in in the second part of this article – Probably best to have a plan for that day 1, as it appears there could be 4000 desktops one day ?
VMware View 3.x is not compatible with vSphere you will need VMware View 4.0 to achieve higher numbers. In any case I would not recommend going further than 2500 virtual desktops per cluster without a good performance tunning.
For the presented configuration and assumptions with >1680 VMs, a third cluster will have to be added to the infrastructure allowing each cluster to withstand 1 host failure.
The table below shows the #ESX hosts and Clusters required, given the number of users.
Some important considerations:
- At 20,000 VM’s mark, the design will result in hundreds of servers.
- A host failure means 100 desktops will reboot.
- Storage Sizing
Different types and the number of golden images will directly affect how storage is sized and performance tunned. It’s also important to understand how virtual desktop pools are configured and if user data disks will be used to host profiles or a roaming solution works better for your environment.
More recently, I have been seeing deployments of the different components over different storage tiers allowing virtual desktops to be provisioned and run on tier one storage while profiles and data disks in tier 2 or 3 storages. This approach could save thousands of dollars in storage for your organization.
VMware View Composer comes bundled with the Premier version and provides advanced image management and storage optimization, reducing storage requirements for virtual desktop machines by up to 90 percent and enabling organizations to more effectively manage their desktop images.
VMware View Composer and Thin-Provisioning (VMware View 4.0) technologies provide greater storage savings and should be used when possible.
VMware and partners have done a good job in documenting storage and server sizing and these sources should always be used when sizing your environment.
Summary and Final Thoughts
In this article I tried to cover the most important aspect of a VDI implementation in regards to the strategy and technology definition.
There are a number of other subjects such as High availability and business continuity, Disaster recovery, Portability and flexibility, Networking, Security, Monitoring & management and Operational considerations that should be considered when designing your VDI solution however they are not the focus in this article.
I hope this article helps you with an initial roadmap for your VDI strategy. Please feel free to provide a feedback or get in touch should you have any questions or require assistance for your solution or deployment.
“There is no right or wrong when designing a solution, just different ways to achieve the same outcome. The design need to be catered for your organisation needs.”
Other related contents:
NetApp and VMware View (VDI) Best Practices for Solution Architecture, Deployment and Management
Whitepaper: Designing an Enterprise XenDesktop Solution (for 10,000 VDI seats) –> Different technology but interesting to compare