DaaS is a Non-Starter

A while back I published articles describing guidelines to what DaaS architectures should observe. My articles focused on Secure Multi-Tenancy, Elastic Resources, Automated Provisioning, Identity Management, Federation and Role Management, Self-Service Portals, and Reporting and Metering.

I was mostly focused on DaaS aspects that relate to large service providers’ implementations. However, the reality is that any enterprise can have a successful DaaS implementation simply creating a self-service portal where users can provision virtual desktops, as needed, automatically, without requiring human interaction.

DaaS self-service portals should be an easy-to-use web interface that abstracts infrastructure from end users. Underlying information such as datastores, resource pools and desktop images should not be something that users should be required to have knowledge about.

Assuming this convention is correct, DaaS is nothing more than VDI with a pretty self-service (amazon like) portal. Yes, I am oversimplifying as I am not considering aspects mostly applicable to service providers, such as multi-tenancy and charge back.

However, while VDI has been a success in terms of adoption and growth with a total market forecast of $6B over the next 3 years, I don’t believe the Cloud-based DaaS hype will survive much longer.

I see vendor trying to capitalize on the idea and create a false sense of hype, but the reality is that while VDI is complex to implement from an infrastructure perspective, DaaS is almost impossible to implement from an applications and usability standpoint; and the reasons are simple to understand.


Enterprise Applications

The simple fact of being able to provision and access desktops instantiated in a cloud provider datacenter doesn’t mean much for the large majority of enterprise users given that they will not have access to their corporate applications.

Virtual desktop images must be customized with enterprise applications that are specific to certain users, departments or branches, and the majority of DaaS providers do not offer this type of service today. Service providers are not interested in providing this type of service because they are infrastructure players, and they will certainly offload the work to companies specialized in PC life-cycle management. I believe overtime we will see more of this value added service, but will be vital to the success of any DaaS initiative to have a gigantic knowledge transfer to enable service providers and partners to execute on enterprises’ PC life-cycle management methods and processes.

A positive thing is that many applications used by enterprises today are SaaS based. However, there is still a massive number of enterprise applications residing on corporate datacenters that have never been designed to cope with high internet latencies, VPNs and low-bandwidth. Many of those applications would have to be presented to user’s cloud-based desktops using RDS type technologies, forcing organizations to expend dollars and cycles to make that happen.

Some of the organizations I talk to have more than 500 or even 1000 enterprise applications and moving all those applications to a presentation layer is a non-starter. Furthermore, some applications may never be suitable for this type of technology.

The other possibility is to effectively move enterprise applications closer to cloud desktops themselves, and that’s what service providers are dreaming about when they offer cloud-based desktops at such low prices ($35/month). Moving applications to their datacenters would ensure the lowest possible latency to access databases and backend services, but for large enterprises this is also a non-starter.


Resource Access

Where are user’s files located? They are likely seating in a NAS or Windows CIFS shares in corporate datacenters or branch offices with backups across the organization. Organizations have moved user files from corporate datacenters to branches because users complained about access times when retrieving data.

Unless organizations move all user data to the cloud a sizable amount of bandwidth would be required to allow users to have a good experience when accessing corporate files and data. It’s not only about bandwidth, but also about latency. This is also part of service providers’ dream to make sure organizations shift their workloads to their datacenters.

Fortunately there are some good solutions to minimize the problem, such as replication or maintaining multiple copies of the content in different locations, but ultimately this is only augmenting complexity.

How about printing? Printers are not located at the DaaS provider’s datacenter. There are solutions today to minimize the impact of printing and the huge PCL bandwidth consumption, but again, this will just add more complexity to the solution.



Guaranteeing a good user-experience for the display protocols, be PCoIP, ICA or HDX is just the starting point. Organizations embracing DaaS need to ensure that there is enough bandwidth available between DaaS providers and corporate datacenters for application presentation and backend services access; and between DaaS providers and user locations for desktop access using efficient display protocols, file transfers and printing.

Ultimately, there is also the need to completely re-architect network infrastructures to make sure user experience is not affected when shifting to cloud-based desktops.


DaaS vs. VDI

VDI is also complex and many of the DaaS challenges also exist in VDI deployments. However, the majority of VDI deployments are found in corporate datacenters where enterprise applications are hosted, not requiring to re-evaluate applications to run over high latency, low bandwidth networks, or to be presented with RDS type technologies.

VDI desktop management is still controlled by IT, reducing friction and knowledge transfer. In many cases VDI is just managed like physical desktops using existing PC life-cycle management tools that would not work well with DaaS deployments due to network consumption restrictions. In other cases VDI can use tools to enable layering and other image management methods that may not be available trough a DaaS provider.

Bandwidth for printing and display protocols are an issue for both VDI and DaaS and may be mitigated through the use of additional technologies, but with VDI organizations don’t have to worry about application level bandwidth since all applications servers and backend databases live for the most part in the same datacenter. Imagine trying to figure out bandwidth requirements for all 1,000 enterprise applications in your datacenter.



I might be oversimplifying, but I actually think shifting to cloud-based desktops can be even more complex than what I am describing here since I have not even touched aspects like security and control, service levels and compliance.

Unquestionably there are organizations that may be able to embrace cloud-based desktops and I know few of them, but I would say that they are just a very small subset of the organizations that would deploy VDI, or small companies and startups that do not want to invest in VDI infrastructure and operations.

All in all I hope I am wrong and that the DaaS market find a way to strive.


Disclaimer: Any views or opinions expressed in this article are my own, not my employer’s. The content published here is not read, reviewed, or approved by Nutanix and does not necessarily represent or reflect the views or opinions of Nutanix or any of its divisions, subsidiaries, or business partners.


This article was first published by Andre Leibovici (@andreleibovici) at myvirtualcloud.net


1 ping

Skip to comment form

    • forbsy on 03/12/2014 at 2:08 am

    DAAS such as VMWare’s offering has the option of on premises DAAS – thus eliminating the concerns you’ve listed. It would be managed VDI

  1. forsby,

    Yes, I that would be just like putting a self-service portal in front of VDI; perhaps some multi-tenancy. My article is really focusing on cloud based desktops because that’s what the market understand for DaaS.


  2. Beautifully written. I must say that if I establish my new Cloud platform with keeping the above instructions in mind, it would be a grand success and easy to work with.
    However, I’m still apprehensive about the bandwidth that you are talking about and also inquisitive to know about what are you exactly trying to compare between nutanix and vcenter.

    • Chris Keim on 03/13/2014 at 8:25 am

    I agree with you that implementing DaaS in the cloud is more complex. The company I work for provides DaaS on a small scale to customers. The issues that I have seen that department go through mirror what you have said. Properly architecting the infrastructure and setting limits a.k.a. expectations for customers is huge and can help mitigate some challenges. Also, when onboarding a new client with specific requirements (data, apps, vpn, etc…), it is best to not rush through the beginning stages even at the customers request. Every good project manager knows that a lot of work a.k.a. planning/architecting is done in the beginning of a project.

  3. @Viola Allen, I am not sure what you meant about comparing Nutanix and vCenter. There’s nothing related to that in this article. I suppose you are referring to this article? http://myvirtualcloud.net/?p=6030

  4. @Chris Kleim, well said and done! Managing expectations and a proper consulting service for the migration is the key for success with DaaS. I would love to learn more about your solution and how you onboard those customers.

  5. We are actually rolling out DaaS for a F1000 customer who wanted to go to a cloud based DaaS but for the reasons you listed decided it wasn’t going to go well despite vendor promises (DB2 database tied to the i-Series and not ready to move the whole data center with it). Ended up doing Desktop as a Service by deploying View in their same data center using cross connects. They get the agility/expertise/outsourcing of us doing it, while they get the benefits of on prem without loosing a single rack unit.

    • forbsy on 03/14/2014 at 8:22 am

    I guess it depends on your definition of cloud based desktops. DAAS is still VDI. It’s just hosted and managed in the “cloud” by service provider. The “cloud” can be private or public or even hybrid.
    Desktone (VMware Hybrid Cloud DAAS) offers the service via managed provider cloud services (i.e. offsite from customer premises), but they also give the option of hosting and managing the service on customer premises (i.e. private cloud).
    So, the on-premises service isn’t a mish-mash VDI with some multi-tenancy sort of thing. It’s the actual Desktone solution sitting on the customer datacenter.

    • forbsy on 03/14/2014 at 8:24 am

    @John – Exactly what I was talking about 🙂

  6. if your “DaaS” (I don’t really call us that) provider is willing to setup shop in your DC, and they can deliver a “per user per month” billing model (We can even do this with equipment sold to customer getting really creative with contracts). It sounds weird, but in a company where purchasing hardware takes a 6-12 month through procurement, and they need dynamic scaling, a creative financial vehicle and a skilled service deliver team can provide a lot of value. Its amusing having a conversation with the customer about scaling and responding with “how quickly can I put servers on the company AMEX, get the un-boxed, and run down the rack and speed rail them in” is as fast as I can provision. In environments where the customers internal procurement and management will ALWAYS over subscribe a server 20x what it was designed for, this model protects a customer fundamentally from them self. (The foundation for every terribly slow Citrix farm) SLA’s are sadly easier to enforce against an external entity I guess.

  1. […] DaaS is a Non-Starter […]

Leave a Reply