Cisco Blogs

# VDI “The Missing Questions” #4: How much SPECint is enough

In the first few posts in this series, we have hopefully shown that not all cores are created equal and that not all GHz are created equal. This generates challenges when comparing two CPUs within a processor family and even greater challenges when comparing CPUs from different processor families. If you read a blog or a study that showed 175 desktops on a blade with dual E7-2870 processor, how many desktops can you expect from the E7-2803 processor? Or an E5 processor? Our assertion is that SPECint is a reasonable metric for predicting VDI density, and in this blog I intend to show you how much SPECint is enough [for the workload we tested].

You are here. As a quick recap, this is a series of blogs covering the topic of VDI, and here are the posts in this series:

Addition and subtraction versus multiplication and division. Shawn already explained the concept of SPEC in question 2, so I won’t repeat it. You’ve probably noticed that Shawn talked about “blended” SPEC whereas I’m covering SPECint (integer). As it turns out, the majority of task workers really exercise the integer portion of a processor rather than the floating point portion of a processor. Therefore, I’ll focus on SPECint in this post. If you know more about your users’ workload, you can skew your emphasis more or less towards SPECint or SPECfp and create your own blend.

The method to the madness. Let me take you on a short mathematical journey using the figure below. Starting at the top, we know each E5-2665 processor has a SPECint of 305. It doesn’t matter how many cores it has or how fast those cores are clocked. It has a SPECint score of 305 (as compared to 187.5 for the E5-2643 processor). Continuing down the figure below, each blade we tested had two processors, so the E5-2665 based blade has a SPECint of 2 x 305… or 610. The E5-2665 blade has a much higher SPECint of 610 than the E5-2643 blade with just 375. And it produced many more desktops as you can see from the graph embedded in the figure (the graph should look familiar to you from the first “question” in this series).

And now comes the simple math to get the SPECint requirement for each virtual desktop in each test system:

# Cisco and VMware Teaming-Up to Accelerate VDI Uptake in 2013 – Join us at VMware Partner Exchange to Find Out More

Are you a Cisco/VMware partner going to VMware Partner Exchange (PEX)?  In Vegas!  Next week!  If so, and you’re focused on growing your VDI practice, there’s some great content for you to take in while there.  Before I get into PEX, let me remind you about our on-going blog series, “VDI –  The Questions You Didn’t Ask (But Really Should)”.  We’re up to Question 4 (coming soon), and if you’re looking for some great insights into the mystique of processor selection and impact on VDI performance/density/etc, this is the series for you!  Now onto PEX…

Improving the ROI of VDI within small and medium-sized organizations

Next week, we’ll be updating our VMware partner community on new solutions that offer an accelerated path to growing their VDI practice, especially for smaller deployments, as found in small and medium sized businesses (or pilot / proof-of-concept environments), where the up-front CAPEX hurdle is often too much of a barrier to make VDI cost effective.

New Ecosystem Solutions Portfolio

In tandem with VMware, we’ll be announcing a new portfolio of solutions built with ecosystem partners helping to deliver better VDI price-to-performance ratios, greater operational simplicity, and uncompromised user experience, built on Cisco UCS with VMware Horizon View.

Delivering the Tools to Make Our Channel Partners Successful in 2013

Next week we have good news for Cisco/VMware channel partners who want to grow their VDI practice, and deliver unprecedented value for their customers.  Join us at PEX to learn about how Cisco and VMware are accelerating our partners’ path to success in 2013

So Are You Ready for PEX?  Here are some key activities you don’t want to miss:

Cisco Partner Bootcamp (#SPO2400)

Monday, February 25, 8:30 a.m.-5:30 p.m.

The Cisco Boot Camp is dedicated to educating and enabling partners to sell and deploy Cisco solutions successfully. Here’s the best part J VDI is up first at 8:30am!  I’m pretty sure we’ll have food and non-alcoholic beverages (c’mon you’re in Vegas, I really don’t think that will pose a problem) to make it worth your while.  You will:

 Expand your technical depth and understanding of key Cisco solutions for VDI, Cloud, Branch/Remote-Office IT, Unified Management and more Gain insights to identify your customer needs effectively and acquire new ones Find out how to expand business by cross-selling Cisco solutions and services Network with other partners, Cisco experts, and executives Come away with go-to-market selling strategies that enable you to accelerate your business

Cisco’s Breakout Session (#SPO2421): Cisco Unified Data Center – From Server to Network
Wednesday, February 27, 12:30-1:30 p.m.
Presenter: Satinder Sethi, VP, Server Product Management and Data Center Solutions, Cisco

Attend the Cisco breakout to understand why today’s data center architecture must support a highly mobile workforce, proliferation of devices, and data-driven business models and be capable of transparently incorporating cloud applications and services. Satinder Sethi will present these diverse requirements and discuss how the Cisco Unified Data Center platform addresses these challenges.

You will learn about the Cisco Unified Data Center architecture, which combines compute, storage, network, and management into a platform designed to automate IT as a service across physical and virtual environments, resulting in increased budget efficiency, more agile business responsiveness, and simplified IT operations.

Demo’s!  Stop By Booth #1015

Eight solid demos await you at our PEX booth this year, including VDI with VMware Horizon View and our UCS Storage Accelerator (using Fusion-io), Unified Computing System (UCS), Cisco Office in a Box with UCS-Express, Cisco Intelligent Automation for Cloud, Cisco Cloupia, and Cisco Nexus 1000v to name a few.  Experts on hand will answer any/all questions!

It will be a busy week – mark your calendars with the activities above, and see you there!

# VDI “The Missing Questions” #3: Realistic Virtual Desktop Limits

So this is the Million Dollar Question, right? You, along with the executives sponsoring your particular VDI project wanna know: How many desktops can I run on that blade? It’s funny how such an “it depends” question becomes a benchmark for various vendors blades, including said vendor here.

Well, for the purpose of this discussion series, the goal here is not to reach some maximum number by spending hours in the lab tweaking various knobs and dials of the underlying infrastructure. The goal of this overall series is to see what happens to the number of sessions as we change various aspects of the compute: CPU Speed/Cores, Memory Speed and capacity. Our series posts are as follows:

But for the purpose of this question, let’s look simply at the scaling numbers at the appropriate amount of RAM for the the VDI count we will achieve (e.g. no memory overcommit) and maximum allowed memory speed (1600MHz).

As Doron already revealed in question 1, we did find some maximum numbers in our test environment. Other than the customized Cisco ESX build on the hosts, and tuning our Windows 7 template per VMware’s View Optimization Guide for Windows 7, the VMware View 5.1.1 environment was a fairly default build out designed for simplicity of testing, not massive scale. We kept unlogged VMs in reserve like you would in the real world to facilitate the ability for users to login in quickly…yes that may affect some theoretical maximum number you could get out of the system, but again…not the goal.

And the overall test results look a little something like this:

 E5-2643 Virtual Desktops E5-2665 Virtual Desktops 1vCPU, 1600MHz 81 130 2vCPU, 1600MHz 54 93

As explained in Question 1, cores really do matter…but even then, surprisingly the two CPUs are neck and neck in the race until around 40 VM mark. Then the 2 vCPU desktops on the quad core CPU really take a turn for the worse:

Why?

Co-scheduling!

When a VM has two (or more) vCPUs, the hypervisor must find two (or more) physical cores to plant the VM on for execution within a fairly strict timeframe to keep that VM’s multiple vCPUs in sync.

MULTIPLE vCPU VMS ARE NOT FREE!

Multiple vCPUs create a constraint that takes time for the hypervisor to sort out every time it makes a scheduling decision, not to mention you simply have more cores allocated for hypervisor to schedule for the same number of sessions: DOUBLE that of the one vCPU VM. Only way to fix this issue is with more cores.

That said: the 2 vCPU VMs continue to scale consistently on the E5-2665 with its double core count to the E5-2643. At around the 85 session mark, the even the E5-2665 can no longer provide a consistent experience with 2vCPU VDI sessions running. I’ll stop here and jump off that soap box…we’ll dig more into the multiple vCPU virtual desktop configuration in a later question (hint hint hint)…

Now let’s take a look at the more traditional VDI desktop: the 1 vCPU VM:

With the quad-core E5-2643, performance holds strong until around the 60 session mark, then latency quickly builds as the 4000ms threshold is hit at 81 sessions. But look at the trooper that the E5-2665 is though! Follow its 1 vCPU scaling line in the chart and all those cores show a very consistent latency line up to around the 100 session mark, where then it becomes somewhat less consistent to the 4000ms VSImax of 130. 130 responsive systems on a single server! I remember when it was awesome to get 15 or so systems going on a dual socket box 10 or so years ago, and we are at 10x the quantity today!

Let’s say you want to impose harsher limits to your environment. You’ve got a pool of users that are a bit more sensitive to response time than others (like your executive sponsors!). 4000ms response time may be too much and you want to halve that to 2000ms. According to our test scenario, the E5-2665 can STILL sustain around 100 sessions before the scaling becomes a bit more erratic in this workload simulation.

Logic would suggest half the response time may mean half the sessions, but that simply isn’t the case as shown here. We reach Point of Chaos (POC!) where there is very inconsistent response times and behaviors as we continue to add sessions. In other words: It does not take many more desktop sessions in a well running environment that is close to the “compute cliff” before the latency doubles and your end users are not happy. But on the plus side, and assuming storage I/O latency isn’t an issue, our testing shows that you do not need to drop that many sessions from each individual server in your cluster to rapidly recover session response time as well.

So in conclusion, the E5-2643, with its high clock speed and lower core count, is best suited for smaller deployments of less than 80 desktops per blade. The E5-2665, with its moderate clock speed and higher core count, is best suited for larger deployments of greater than 100 desktops per blade.

Next up…what is the minimum amount of normalized CPU SPEC does a virtual desktop need?

# VDI “The Missing Questions” #1: Core Count vs. Core Speed

Choosing the right compute platform for your VDI environment requires both science and art. You have to balance CPU and memory characteristics against your expected workload profile and your desired density. At the end of the day, VDI has to meet some cost criteria in order to go from a fun science project to a funded program in your company. That means you can’t just throw the top bin CPU at the problem; you have to pick the right CPU. This is further complicated by the fact that there is not one CPU that is ideal for all VDI workloads. There is no magical bill of materials at the end of this series of blogs, but we will attempt to make your VDI decisions based more on science than art.

Strength in numbers? Or strength in speed? As Tony said in his introduction, we had several involved questions related to VDI that we honestly couldn’t answer… so we decided to start testing. This will be a series of blogs that attempts to answer practical questions like “when is processor A better than processor B?” And of course you then have to ask “when is processor B better than processer A?” In this first installment in the series, I will tackle the question of whether the number of cores or the core speed is more important when the goal is to achieve the best desktop density per host. Here is a handy guide to the other posts in this series:

The usual suspects. Throughout this series, we will focus on two processors. We picked them because they are popular and cost effective, yet quite different from each other. They are not top bin processors. Take a look at the table below for a comparison.

Note: Prices in this table are recommended prices published by Intel at http://ark.intel.com and may vary from actual prices you pay for each processor. The SPEC performance numbers are an average of SPEC results published by many OEMs (at http://www.spec.org/) across many platforms. These are not Cisco-specific SPEC numbers.

# VDI – The Questions You Didn’t Ask (But Really Should)

There’s no shortage of content out there (a quick Google search easily confirms this) when it comes to looking for vendor-originated material touting the latest server performance benchmarks for hosted virtual desktops.  Being part of that community, I’m pretty sure I have my fingerprints on more than one such piece of collateral – and I’m constantly reminded of this, when we run into questions along the lines of “yeah, {xxx} desktops on a blade is great, but c’mon, you and I both know we’d never do that in practice”.  It’s a balancing act of demonstrating solution performance, intersected with the practical reality of what IT managers would reasonably support in a production environment.

So what really matters?  If I’m implementing VDI for the 1st time, and I’m trying to make intelligent decisions around CPU, memory speed, IOPS, etc., where do I go?  VDI is unique in its consumption of compute, storage and network resources, when compared to other workloads hosted in the data center.  Much of the performance benchmarking info put out by server manufacturers is not specific to VDI performance, or how user experience might be impacted by simple decisions like choice of clock speed or # of vCPU.

Thankfully, there are folks in my company that care a LOT about such questions.  So much so, that a small, VDI-proficient group of them took it upon themselves to design and build an in-house lab environment with one express purpose – exhaustively exploring and documenting the performance and scalability impacts seen when configuring your compute platform for VDI.  No stone left unturned – things like CPU cores, clock speed, memory speed,  vCPU, memory density and more – all fair game.

The findings are extremely valuable to anyone deploying VDI, and what this team discovered is a set of real-life “questions”.  The “Missing” questions if you will – those questions that are noticeably absent or never sufficiently exposed in marketing materials, when it comes to the practical choices you can make that most significantly impact the cost, scalability and performance of your virtual desktop implementation.

So let me start with an introduction.  Over the next few weeks, you’re going to hear from some peers of mine – Doron Chosnek, Jason Marchesano, and Shawn Kaiser.  They’re Cisco Consulting Systems Engineers, and they live and breathe VDI (I know, melodramatic), as implemented in their customers’ data centers around the world.

They undertook this journey with the express purpose of answering the “missing” questions, by assembling a test platform in their lab, built on Cisco Unified Computing System (UCS), using readily available components including:

• Various UCS B200 M3 configurations
• VMware View 5.1.1
• Microsoft Windows 7 SP1 32-bit virtual desktops
• Pure Storage FlashArray with Purity version 2.0.2.

Keep in mind that their goal was not to explore maximum scalability, or prescribe a preferred design/architecture, or even what kind of server blade or processor you should use for VDI.  Instead they relied on commonly available gear easily found in our customer’s data centers.  If you want prescriptive design guidance, Cisco CVD’s are ideal for that, and you can find them here.

So let’s talk about their test environment.

Physical Lab

The physical environment shown below is a highly overprovisioned system.  Only one B200 M3 blade was tested at any one time, yet every logical link between elements shown consists of multiple 10-GbE links or multiple 8-Gb Fibre Channel links.

The storage array has 24 flash disks and is capable of substantially higher IOPS than used for this testing. All the infrastructure machines used for this test (Active Directory, VMware vCenter, VMware View, VSI Launchers) are virtual machines on the B230 M2 blade in the environment.

Note: At the time of testing, the Pure Storage had not completed UCS certification testing.

Logical Server Environment

The tests involved two UCS B200 M3 blades, one with dual E5-2665 processors and the other with dual E5-2643 processors.  The 2643 is a 4-core high clock/burst speed processor, and the 2665 is an 8-core medium/high clock/burst speed processor.  Here are the specs for the CPU’s chosen:

Now, you may wonder, are either of these THE processor you would choose for VDI?  Not necessarily!

Keep in mind the goal we set out with – to expose the relative impacts of # cores, clock speed, memory speed, #vCPU’s etc.  What you’ll take away from the results, are guidance on which parameters matter for specific types of VDI deployments.  You can then safely look at a VDI-“workhorse” processor like the E5-2680 or E5-2690, and apply what our CSE’s have learned through this testing, to that class of CPU, and make your best selection there.

The tests were conducted using Login VSI’s Medium with Flash workload generator.  As we explore the test results in this series, you’ll see reference to “VSImax”, which defines the threshold past which the user experience will be unacceptable.  The VSImax threshold will appear on supporting graphs that show the performance curve under various test scenarios.  You can learn more about how this threshold is derived here.

So that’s the test environment.  Through this series – let’s call it VDI – the Questions You Didn’t Ask (But Really Should) – our CSE friends (Shawn, Doron, and Jason) will explore and expose the findings they’ve documented for us, dealing with a new “question” each time.  If you join us for this journey, it’ll be worth your while – you’ll come away with a better appreciation of the impact that some simple decisions in your data center compute configuration can make.

So are you ready for the journey – You’ll find the Questions (answered thus far) below:

Special Web Event – You’re Invited!

If you’re enjoying our series, be sure to join our free webcast, where Shawn, Doron and Jason will discuss all the (Missing) VDI Questions Live + take your Q&A.  Access the webcast here.

Featured Whitepaper Now Available!

Need a convenient whitepaper-ized version of the discussion?  Download it now, here.

Tags: , , , , ,