It’s been said that a wound gets worse when it’s treated with neglect. The same can be said about the performance of your computer environment. Even the best technology cannot do much if it’s neglected or not optimized to perform at its best. This is undoubtedly true of Virtual Desktop Infrastructure (VDI). For all the discussion about VDI solutions, the published applications, or shared hosted desktops for that matter, good or bad, tuning and optimization are essential elements of any application or VDI implementation – be it Citrix Virtual Apps and Desktops or VMware Horizon.
A recent two-week VDI Assessment we completed for a national insurance carrier illustrates the point very well. While we were there to look at vGPU performance, what we uncovered was far more critical in terms of maximizing the value of the VDI infrastructure and its role for the organization.
Let's start with the simple stuff. Login times at baseline for this client were very good – 16.37 seconds. After optimization, however, there was a 49 percent improvement to 8.27 seconds. We’re only talking about a difference of eight seconds, so let's keep going.
Storage performance – One of the big three in terms of its ability to impact the user experience and the storage here was no exception. Initially baselined at 12 megabits per second (MiBPS), performance throughput or the amount of data the storage could move across the environment increased by 208 percent to 37 MiBPS.
Whitehat Virtual was able to reduce the maximum IO response time from 718ms to 26ms. A 96 percent decrease to achieve a 69percent improvement in IO response times, from 5.3ms to 1.6ms. Total storage IOPS improved from an average of 3,000 IOPS to 9,440 IOPS – a 211 percent performance increase.
The number one bottleneck limiting the number of end-users that can share one VDI host is, of late, always capped by CPU performance. That means the ultimate cost of VDI (per user) is going to be calculated by understanding how many employees can share a server and still have an excellent end-user experience.
The more, the better, up and until the end-user experience starts to degrade, slowing down the end-users’ ability to be more productive.
So how did this client environment perform?
The Central Processing Unit (CPU) – At baseline, this CPU bounced between 90-100 percent utilization, generating many complaints from end-users as the environment would slow down or hang for a bit and prevent them from being productive. After optimization, with the same user load, CPU utilization decreased 77 percent to 20 percent, leaving a lot of room for additional end users to be added to the environment while maintaining an excellent end-user experience. CPU consumption when at idle was reduced from 25-50 percent CPU to less than 2 percent – a 92 percent decrease.
NVIDIA vGPU – The client's original request was to review NVIDIA vGPU performance and help them understand what the challenges were. While many VDI environments don’t have vGPU yet, wider adoption occurs as Microsoft Teams with webcams, training content, and multi-media continue to become more critical to the way people get work done.
NVIDIA vGPU bandwidth consumption dropped by 50 percent upon completion of the VDI optimization work done as part of the assessment engagement. The takeaway here is that in a VDI environment thought to be performing well, with reasonably fast login times, significant improvements in VDI workspace performance can still be realized. This "well-run" VDI environment left a lot of capability and performance unknowingly on the sidelines and it was all uncovered through a simple assessment. The minimal improvement in this environment was 49 percent, with some optimizations improving VDI environment performance by over 200 percent.
Looking beyond the numbers at the desks of the end-users, where the data converts into dollars, there were immediate reports of happier end-users reporting performance improvements. With the CPU bottleneck removed, support calls related to random slow downs stopped. The cost per user of the VDI environment decreased as more end-users could be in the environment having a great VDI experience at the same time.
What did Whitehat do to find so many opportunities for VDI performance improvement? Whitehat's assessments are derived from what Whitehat learns tuning, optimizing, managing, or co-managing VDI environments worldwide. Every lesson learned, every optimization discovered, is cataloged to the benefit of every Whitehat customer.
However, the most important thing you can do to find these performance improvements is to look for them. Not just in the Citrix CVAD and VMware Horizon environment specifically, but in all the systems that support VDI as well. I can all but guarantee there are performance gains there awaiting discovery. Take the time to look at these areas that may be neglected today to make sure Citrix or VMware VDI is delivering maximum performance. Devise an assessment or contact Whitehat Virtual to get the correct assessment product to meet the need. For example, Whitehat has a “30 Second Challenge” that guarantees its customers’ login times will be 30 seconds or faster in eight hours or the work is free. The next step up is the optimization service Whitehat offers. It’s not a full assessment but an optimization of Citrix/VMware and the related systems.
While it is true that not all assessments are created equal, those delivered by teams in VDI operations, not just consultants, tend to provide more meaningful results. Find the edges of your VDI environment, optimize it, and push it to its maximum performance. Good luck!