VDI Reference Architecture Monthly – September 2016
VMware Horizon 7 and VMware vSphere 6.0 with VCE VxRail 160 Appliance by EMC Global Solutions Engineering and EMC End-user Computing Summary
As I’ve come to expect, this 68-page reference architecture (RA) was very detailed. Not only did the document go through a number of different and relevant scenarios but it also includes clear details on best practices and design considerations, offering a great recipe for setting something like this up for yourself. The focus of this document was on a relatively new hyper-converged platform from EMC and VMware called the VxRail, as well as a comparison between Windows 10 and Windows 7 desktop environments.
There are a lot of exciting things in this RA that caught my eye. Here is my quick list:
- The VxRail is a fully integrated, pre-configured and solution-tested hyper-converged appliance which leverages server nodes that provide the compute and storage layers. Of particular note is the sub-5 millisecond storage latency in a VMware Virtual SAN configuration, which should offer a snappy end-user experience.
- My background is heavily footed in high-performance storage, so I really like to see reference architectures which highlight the types of tests that some of my virtual administrator or architect friends will conduct to prove out their storage choices. These typically include boot storm, refresh storm, recompose and virus scan. If you’re curious how well the tests went, here is a little proof.
- Microsoft is moving quickly with their Windows 10 push, so any information that helps provide insight around performance comparisons with Windows 10 is a big win in my book. In this paper they compare 230 Win 7 vs Win 10 linked-clone using Office 2013.
- Users: Between 230 and 290 Knowledge Worker users depending on desktop OS.
- Provisioning method: Both Linked-Clone and Full-Clone.
- Provisioning system: VMware Horizon 7.
- Desktop OS: Windows 7 and Windows 10 with Office 2013.
- Hypervisor: VMware vSphere 6.0.
- Hardware: VxRail 160 Hyper-converged Infrastructure with 4 servers (64 cores total of Dual Intel Xeon E5-2630 v3 2.4 GHz, 256GB) which were used for all roles. VMware Virtual San was used for storage.
Login VSI Results
Let’s take a look at the Login VSI test results quickly. One thing to note is that VSImax was not reached in these performance tests. In most cases this is a good thing and means there is additional compute, network or storage capacity to host more users and that user experience is optimal. The VSImax baseline represents the responsiveness of the system under minimal load, or in other words the best case user experience on that system.
- Windows 10
- 230 Concurrent Linked-clones: VSImax Baseline: 951 ms, VSImax Avg: 1503 ms, VSImax not reached.
- 230 Concurrent Full-clones: VSImax Baseline: 956 ms, VSImax Avg: 1642 ms, VSImax not reached.
- Windows 7
- 290 Concurrent Linked-clones: VSImax Baseline: 785ms, VSImax Avg: 1557, VSImax not reached.
ESXi host CPU utilization comparison between linked-clone Windows 7 and Windows 10
WINDOWS 10 OPTIMIZATION TEST RESULTS, View in VMware Horizon 7
Sweet! Another document comparing Windows 7 and Windows 10. Are you catching on yet that Windows 10 is quickly becoming the pervasive choice for desktop OS for many businesses. This 17-page paper is pretty basic and quickly covers performance advantages that come with using their OS Optimization Tool (OSOT). It looks like this paper focuses very much on desktop performance using VSI baselines and not VSImax, which means you’ll have to draw some conclusions about scalability. Of course you can also refer to our test results when it comes to OSOT and scalability.
Here are a few highlights that I found particularly interesting:
- This paper compares the performance benefits of using VMware’s OS Optimization Tool (OSOT). If you recall Login VSI also created a Windows 10 template for the OSOT.
- This paper details some of the overlooked but very important optimizations within server BIOS dealing with the High Performance mode set in Power Management Policy. There are other settings you may want to look at in this paper.
Given that this paper was more of an executive summary, it didn’t go into great detail about the test environment. Here is what we know:
- Desktop OS: Windows 7 and Windows 10.
- Provisioning system: Horizon 7.
- Hypervisor: VMware vSphere.
- Hardware: DELL PowerEdge R730 (28 cores of Intel Xeon E5-2695 v3 2.3 GHz, 262 GB); XtremIO storage array.
Login VSI Results
Instead of focusing on the Knowledge Worker workload, this paper emphasized the benefits of OSOT by looking at the Virtual Machine baseline response scores. Basically this helps to demonstrate how much of a performance difference there can be when the desktops have plenty of compute and storage resources and allow the desktop to run at full performance.
Optimization Test Results
The scalability and economics of delivering Citrix XenApp services from Microsoft Azure
If you recall, one of the big announcements at Citrix Synergy this year was about Microsoft and Citrix working together to deliver XenApp 7.11 Services on the Azure cloud. This paper takes a brief look at the scalability costs of running XenApp in the cloud using different Azure compute packages. What’s super cool is that Citrix used Login VSI to do some performance validation with this configuration… yeah Login VSI testing XenApp on Azure. Gotta love it!
- Running XenApp in the cloud makes sense because it leverages the cost benefits of RDSH. It is interesting to see how the cost per user, per hour, breaks out when run on the different instance types… and yes we are talking about pennies ranging between $0.015 and $0.06. Also note that costs vary depending on region.
- Something you should check out is that the most expensive option did not provide the best results in density and performance.
- Another interesting note is the disk I/O latency for the Knowledge Worker workloads were only 3 ms for the writes and less for the reads. Write performance is typically a place where I am most critical on storage performance, and this result looks promising.
- Lastly, note that everything in the cloud typically has a cost, and the paper ends with a close look at the associated network and storage costs.
- Number and type of users: Task and Knowledge Workers. Depending on the Azure instance type, this solution tested a single server hosting between 3 and 70 users. Take a look at the paper to see how this looked per instance type.
- Pool Type: RDSH on Windows Server 2012 with Office 2013.
- Provisioning system: XenApp 7.11 Shared Desktops – sessions through StoreFront.
- Hypervisor: Azure cloud instances (A-Series, D-Series and Dv2-Series Machines). Login VSI Infrastructure on D2v2, Citrix Infrastructure on D2v2, AD and Licensing Server on D2v2, NetScaler VPX on A2, RDSH on A, D and Dv2. For more information, see this article.
Login VSI Results
- Task Worker on D2v2: Baseline 860ms, VSImax: 19 @ 1861 ms.
- Knowledge Worker on D2v2: Baseline 873ms, VSImax: 15 @ 1873 ms.
This was a fun opportunity to learn just how much great information is contained in these documents. Please understand that the information that I found exciting is only part of the story and I highly encourage you to dig into anything that may catch your attention.
About this Publication
We are proud of the fact that technology vendors use Login VSI software to test their VDI solutions on a regular basis. Often, this testing results in a Reference Architecture with Login VSI test results. In this monthly newsletter, we review new reference architecture whitepapers published by technology vendors. Many of these documents can be long and time-consuming to review, so in this newsletter we try to save you time by distilling the highlights for you. Please note that the views expressed in this newsletter are solely those of Login VSI, and this newsletter is no substitute for a more thorough review of each reference architecture document on your own. We hope you find this newsletter useful!