I, just like my colleague Tony Paikeday, am somewhat preoccupied these days with the fast changing world of the desktop and its impact on data center infrastructures. I wanted to pick up on Tony’s desktop virtualization “just another workload” blog back in November because it is a subject of growing discussion, especially with “cloud” being all the buzz. While desktops are an increasingly popular workload to get started with private cloud initiatives, does that mean that data center architects are mixing desktops with more traditional data center workloads?
Talking to our system engineers who are helping plan and design desktop virtualization deployments day in day out…..the more I learn there are very good reasons for treating this workload as special and separate.
The first thing I hear about is sizing of the desktop workload. A “desktop” is not a “desktop”. You can’t just characterize a generic Win 7 desktop for compute, memory, I/O and storage IOPS. You need to be able to customize the infrastructure profile to the specific user type being deployed. Therein lays the danger of mixing these virtual desktops with production workloads, where desktops could end up capturing valuable resources of mission critical services. For example consolidating a company procurement application on the same compute pool as your desktop workloads could result in a lot of unproductive – or even worse –unhappy employees.
The next thing they talk about, and perhaps the most important from my perspective is the operational model. Typically, data center operations managers like to lock down their application hosting environments with very controlled maintenance windows. No one touches it -- no one even looks in its direction without the necessary process being followed. On the other hand desktops are prone to change….Windows patches, virus updates, application updates, new desktop deployments…..one IT executive recently stated they were pushing out 95 million desktop updates per year!!….While operations managers absolutely will leverage systems management and cloud orchestration tools and processes across enterprise app and desktop infrastructures, desktops present a much more fluid environment that requires more operational interaction and therefore makes it challenging to fully consolidate into the same pod.
And then there is of course storage IOPS and storage capacity requirements under “steady state” and “boot storm” conditions which are very unique to the desktop workload. Trusted storage vendors, like EMC and NetApp know this better than anyone and have innovative approaches that are very much tuned to deal with this. For example take a peek at the recent post by Vaughn Stewart on how storage tiering can provide some relief. The point is that it often makes sense to create a separate storage pool, optimized for desktops.
The other big concern the engineers I am talking to also invariably bring up is security. One asked me – “do you really want a temporary contractor’s desktop running on the same blade as your HR systems?” From their point of view they want to simplify control of out-breaks of viruses by creating barriers between the different desktop workgroups themselves and also between the desktops and the production workloads in the data center. There are technologies like Cisco’s Virtual Security Gateway running on UCS that allow you to do this.
So can virtual desktops and server workloads leverage the same infrastructure? I believe that there ultimately is a path to a consolidated, shared cloud. However, with a short to medium term focus on getting your virtual desktop initiative quickly up and running there’s certainly an opportunity to de-risk the initiative by designing the optimized ”desktop cloud” to begin with.
Let me know what you think ………and come and look us up at Citrix Synergy in San Francisco www.cisco.com/go/citrix11
Learn more about Cisco VXI desktop virtualization solutions at www.cisco.com/go/vdi