VDI Reference Architecture Monthly – October 2016
Dell Precision Appliance for Wyse - VMware Horizon 7.0 with NVIDIA M10 Test Results
by Dell Wyse Solution Engineering
Looks like DELL has released an update to their hyper-converged Precision Appliance for Wyse to now include NVIDIA’s M10 GPU. Graphics rich experiences are becoming all the buzz lately and if you ask me, that is where I would want my desktop VM to live because the benefits for the end-users are massive. Typically, GPUs are seen as the elixir to what has been commonly referred to as a “laggy” experience, which basically expresses the feeling we have when there is time between a user action and the desktop (like clicking on the Start Menu in Windows) or seeing videos get all choppy.
If you are going to check out this RA, here are some things to look for:
- Take a look at the updated appliance hardware. M10s from NVIDIA, v4 Processors from Intel and SSDs are all pretty cool when it comes to performance-oriented solutions.
- The system utilization for 64 users running the Power Worker workload and a resolution of 1920 x 1080 is great. There are details about this workload in the paper. As far as system utilization here are some quick numbers:
- Processor Utilization was 73% on average during the stead-state part of the workload.
- Memory reached 54% max during steady-state.
- Network utilization reached 220 Mbps.
- Storage utilization looked great and I noticed that during the aggressive boot storm, IOPS spiked up to 23,000 IOPS @ 13ms, then settled down to fewer IOPS and 1ms or less for the rest of the simulation.
- GPU Processor Utilization hit about 70% during steady-state, and GPU Memory stayed around 28%.
- Since it can be tough to objectively measure the quality of a rich-graphics user experience I was intrigued to read the following in the paper:
“At different times during testing, the testing team will complete some manual “User Experience” testing while the environment is under load. This typically involves a team member logging into a session during the run and completing tasks similar to the User Workload description. While this experience will be subjective, it will help provide a better understanding of the end user experience of the desktop sessions, particularly under high load, and ensure that the data gathered is reliable.”
- Users: 64 Power Worker VMs (2 vCPU and 4GB) with Pro Library @ 1920x1080 resolution
- Provisioning system: VMware Horizon 7.0
- Desktop OS: Windows 8.1 with Office 2013
- Hypervisor: VMware vSphere 6.0 Update 2
- Hardware: A DELL Precision Appliance for Wyse (R730 with Intel E5-2698 v4 CPUs and 512GB DDR-4 RAM) was used with 2ea. NVIDIA M10
Login VSI Results
The VSIbaseline was 1068 and VSImax was not reached during this run of 64 Power Worker users. Remember that not reaching VSImax is in most cases 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.
Something that is really fun to look at in the way they present the performance data is how they highlight phases of test with Boot, Settle, Login, Steady State, Logoff… I love that!
GPU Utilization – M10 graphics card 1 on DELL Precision Appliance for Wyse
Lenovo Client Virtualization with VMware Horizon 7
Wow! I’ve not seen such fully comprehensive documents like these in a very long time. They are reminiscent of the IBM Redbooks. If you are looking for a reference architecture to (1) read and learn about how to build a VDI, (2) understand all of the components necessary to build a VDI and (3) find the perfect recipe for the VDI you have in mind, all using IBM and Lenovo products from thin-client to backend storage, then this is your guide.
The two documents on our reference architecture page are as follows:
- The Lenovo Client Virtualization reference architecture: This is a fundamental level (think a dummies guide to VDI using Lenovo goods) document that provides a list of all ingredients needed to bake a great VDI or hosted desktop/application solution, as well as great working knowledge of all that you need to know about these solutions. It also includes information on their test methodologies, which I’d recommend you adopt when you test your new VDI or hosted desktop/application solution.
- The other Reference Architecture applies the principle more in an in-depth review with VMware Horizon. This document is very thorough and provides extensive information about all of the different variations of solutions one could adopt. Whether it is an enterprise multi-component architecture or hyper-converged, stateless or persistent, virtual desktop or hosted desktop, etc…
There is really just too much information to briefly summarize in a blog, and I would highly encourage you to check out these documents as they are more than just a proof-of-claim, they are really a true reference guide to client virtualization.
Here are a few highlights that I found particularly interesting.
- Super helpful sizing guidance including VSImax numbers for all different sorts of server configurations, provisioning types and worker types.
- Detailed VMware VSAN performance comparisons between Hybrid and All Flash storage.
- Sizing guidance for storage capacity with a breakdown of VSImax numbers per storage type and policy.
- A detailed review of different graphics acceleration options using NVIDIA GPUs using vDGA, vSGA and vGPU.
- Seriously, this blog isn’t long enough to cover all of the details in this paper… go check it out to learn more.
- Users: Sizing guidance from 200 CAD users to 10,000 Stateless, and everything in between including mixed stateless and persistent
- Provisioning method: Both Linked-Clone and Full-Clone
- Provisioning system: VMware Horizon 6 for Stateless and Persistent VDI and Hosted Desktops and Apps
- Desktop OS: Windows 7 and 8
- Hypervisor: VMware vSphere 6.0 Recommended
- Hardware: You’ll have to check out the RA for more details about what was used. They provided operational models for both enterprise, built-to-spec architectures and hyper-converged VSAN using both Broadwell (v4) servers and Haswell (v3) servers
Login VSI Results
There are no specific full-coverage Login VSI test results in this document. Since this is more of a cookbook style of reference architecture, the information they provide has more to do with where they hit VSImax with a specific solution architecture. The guidance they provide shows that these guys really know how to use Login VSI and I would expect this can really help an IBM shop figure out the sweet spot when it comes to the amount of equipment needed for any given solution using their products. Of course, I would still recommend that if you do implement using these excellent documents, that you use Login VSI to make sure it works with your enterprise use cases.
VMware Horizon 7 with SimpliVity OmniStack
I was visiting SImpliVity recently and they told me there would be another reference architecture coming out soon. So happy to see this. As you may know SImpliVity is passionate about their testing and proofs-of-claim when it comes to VDI. They are a company we have worked jointly with in the past to do “Validated By Login VSI” reference architectures. On top of being a very detail-oriented and thorough document, they have also included some great information on VMware App Volumes and DR.
- Application layering is becoming a hot-topic and SimpliVity has done some testing to compare the differences between VMware App Volumes in vCenter Mode and VHD Mode.
- Updated performance information using the latest Broadwell-based processors from Intel.
- In typically SimpliVity fashion they hit the systems hard to make sure they stand up to the aggressive nature of VDI using Login VSI steady-state testing, recompose tests, rebalance tests, refresh tests and reset tests. If you know storage, then many of these are going to push the IO to its limits. See how they performed.
- Active/Active disaster recover scenarios including failover and failback for linked-clones and full-clones and DR with a tertiary site.
- Users: 3000 Knowledge Workers and 4000 Office Worker
- Provisioning system: VMware Horizon 7.0 Linked Clones
- Desktop OS: Windows 7 with Office 2010
- Hypervisor: VMware vSphere 6.0 Update 2
- Hardware: OmniStack hyperconverged infrastructure solution. Intel E5-2697 v4 CPUs, 455GB Memory in an 8+8 clustered configuration. Storage was hybrid SSD+HDD with Compression and Deduplication
Login VSI Results
SimpliVity conducted 2 large scale performance experiments for 3000 and 4000 users:
- For the 4000 Office Worker workload the VSIbase was an impressive 817ms. VSImax was not reached on the single 8+0 cluster, with plenty of headroom.
- For the 3000 Knowledge Worker workload the VSIbase was a remarkable 649ms. VSImax was not reached on the single 8+0 cluster, and the VSImax average at max sessions was only 801ms. That’s low and makes for a happy end-user.
Also worth mentioning is the App Volumes testing that was conducted using the Login VSI Office Worker. Which do you think performed better and allowed for more sessions, vCenter Mode or VHD Mode per OmniStack system? Download the paper and find out ;-).
Cisco HyperFlex HX220c M4 Node for Virtual Desktop Infrastructure with VMware Horizon View
Cisco has long been a part of the converged infrastructure game, but with the HyperFlex systems they are now bringing hyperconverged infrastructure appliances to the game (HX220c and HX240c), which I believe has been made in part by their partnership with Springpath. There have been a number of reference architectures published by Cisco using this new product offering. Let’s check out some of the details around the HX220c M4 being used to support VDI with VMware Horizon View.
- Cisco is using 8x of the 1U HX220c systems to support 1000 linked-clone desktops, and all can be managed with UCS Manager framework.
- The HyperFlex system consolidates distributed storage across the nodes in the cluster using a handful of VMware mechanics like IO Visor, VAAI and VMDirectPath, and Cisco provides some deeper technical information about how it works.
- Of course, when it comes to networking, Cisco provides a very detailed run down of equipment and configurations, including a section on Cisco UCS Manager QoS settings.
- During steady state the storage reached 12,500 peak IOPS with 6ms average latency and 85% CPU utilization.
- Through the use of deduplication and compression, storage capacity saw a 97% savings.
- Users: 1000 Knowledge Workers
- Provisioning system: VMware Horizon 6.2 Linked Clones
- Desktop OS: Windows 7 with Office 2010
- Hypervisor: VMware vSphere 6.0 Update 1b
- Hardware: Cisco HyperFlex HX220c M4 Nodes (Intel E5-2680 v3 CPUs and 384 GB Memory), Storage was hybrid SSD+HDD with Compression and Deduplication
Login VSI Results
One of the things I really like about Cisco is how they use Login VSI for what we call soak testing, or testing that tries to flush out any issues that may arise when the system is under stress for an extended period of time.
- For the standard test run using Login VSI Benchmark Mode with 1000 Knowledge workers. Which runs for 48 minutes, Cisco saw a VSI baseline of 702ms, which is a very good number. The VSImax threshold was not reached and users on this system would have been happy with an average VSImax response time of 1115ms.
- As for the 8 hour soak test, the VSI baseline was also a very good 706ms. Again the VSImax threshold was not reached and the VSImax average response time reached 1128ms. During this test all of the resources held up as normal and storage latency only went up to 8ms.
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!
Start using Login VSI today
Our industry-standard software is built to help you avoid problems, lower costs and improve performance. Request a Quote or get your free Trial below, and benefit from our award-winning services.
About the company Login VSI
The company Login VSI provides end-user performance insights for virtualized desktop and server-based computing environments. Enterprise IT departments use flagship product Login VSI (for load testing) and Login PI (for continuity testing) in all phases of their virtual desktop deployment—from planning to deployment to change management—to build and safeguard a good performance, a high availability, and (as a result) a good and consistent end-user experience. For more information about Login VSI or for a free test license contact us.