With flash solutions today much of the emphasis is on cost and performance, and that’s no surprise. Flash storage solutions are being offered at a much more affordable cost per useable Gigabyte, and we are seeing more and more virtualized environments demanding the high performance, low latency that flash delivers. Another characteristic that customers are looking for in flash-based solutions is simplicity: How easy is it to deploy and manage in my IT environment?  Easy set-up and rapid deployment is important, however, ongoing optimization and management also needs to be considered as a key attribute of simplicity.  Given the dynamics of cloud-enabled IT environments customers are looking for proven, integrated solutions that can easily accommodate change and evolve with their infrastructure and strategy.

 

I covered the importance of integration, and NetApp’s management enhancements for Managing Flash in a VMware Environment in an earlier blog (by the way, the answers to the Storage JUMBLE puzzle in that blog are posted below).  Integration lends itself to simplicity in that the effort to make NetApp storage work optimally with VMware has already been done for you – simple.  Our integration with VMware has always been a key differentiator for NetApp.  It is something that we have continued to build on and it is even more important today considering requirements to move data between private, public, and hybrid clouds.  Our integration with VMware helps to simplify that.

 

When choosing and implementing storage customers look to NetApp for guidance.  The reason they look to NetApp is not just because we are an industry leader (Gartner 2016 Magic Quadrant for Solid-State Arrays) and have the broadest flash portfolio.  They look to NetApp because we also have expertise in deploying and managing complete solutions (i.e., including tools and Best Practices) in VMware virtualized environments. Back in 2010 we deployed a then unprecedented 50,000 seat VMware View Virtual Desktop Infrastructure so we could validate some of our own best practices for large VDI deployments.  Last year for NetApp’s own internal virtual desktop project we deployed over 10,000 VMware Horizon seats across five NetApp data centers around the globe – in 12 days.  In between those two projects NetApp, along with our Partners, have helped thousands of customers deploy storage in VMware virtualized infrastructures.  You cannot do that and not gain a very credible level of expertise pertaining to VMware virtualization.  So yeah, NetApp’s expertise goes way beyond just the storage.  And when we build a NetApp Reference Architecture for VMware environments we roll that expertise in.  We look at all the things that we went through during implementations.  We factor in Murphy’s Law.  We even consider the unconsiderable.  Only then will we design, test, and validate a storage system and call it a reference architecture.

 

Earlier this week we published a new NetApp Verified Architecture (NVA) deployment guide: FlexPod Datacenter with NetApp All Flash FAS and VMware Horizon NVA Development (companion to this NVA Design Guide). For more information on the NVA check out this interesting blog for VMware’s perspective on how this FlexPod solution is impacting desktop transformation.  We also released a couple of new Reference Architectures that incorporate the latest and greatest NetApp and VMware technology for End-User Computing: NetApp All Flash FAS Solution for Nonpersistentn Desktops with VMware Horizon View, and NetApp All Flash FAS Solution for Persistent Desktops with VMware Horizon View.

 

When we develop a reference architecture we’re providing a blue print for solutions that our customers will use to help execute on their IT strategy.   That’s why these new publications don’t just lay out the infrastructure components, they provide insight into the rational behind the storage design decisions for the architecture and document the real world test scenarios used to validate them.  We realize that users do different things, at different times of the day, on different days of the week.  Monday is different than Tuesday, and Tuesday is different than Thursday…which is why we simulate a day in the life of a desktop and use test scenarios that span multiple days across a full week or more, testing both persistent and non-persistent desktops. Here is an overview of that testing with a summary of key findings for each category.

 

Persistent Desktops:

In this VMware Horizon 7 reference architecture 1500 persistent Windows 10 desktops were provisioned on an AFF8040.  This AFF8040 was configured with 1 shelf of 800GB SSDs (DS2246) and running the latest ONTAP 9 software which included inline deduplication, inline compression, inline compaction, and Advance Drive Partitioning v2.

 

The testing included the following criteria:

  • Provisioning 1,500 VMware Horizon View full-clone desktops by using VMware vSphere vStorage APIs for Array Integration (VAAI) cloning offload to high-performing, space-efficient NetApp FlexClone® desktops
  • Boot storm test of 1,500 desktops (with and without storage node failover)
  • Monday morning login and steady-state workload with Login VSI (with and without storage node failover)
  • Tuesday morning login and steady-state workload with Login VSI (with and without storage node failover)
  • Patching 1,500 desktops

 

Test Time to Complete Peak IOPS Peak Throughput Average Storage Latency
Provision ~80 min 81,002 2693 MBps 1.45ms
Boot storm test ~12 min 117,752 3082 MBps 2.26ms
Boot storm test during failover ~14 min 75,772 1802 MBps 12.1ms
Login VSI: Monday morning login and workload 20.78 sec/VM 31,088 533 MBps 441us
Login VSI: Monday morning login and workload during failover 19.12 sec/VM 40,000 915 MBps 468us
Login VSI: Tuesday morning login and workload 19.27 sec/VM 21,874 490 MBps 446us
Login VSI: Tuesday morning login and workload during failover 19.36 sec/VM 30,653 1054 MBps 503us
Patching desktops with ~900MB of patches (3-sec delay) ~110 min 42,172 1402 MBps 810us

 

Key findings for persistent desktops:

  • The NetApp All Flash FAS solution was able to very easily meet all IOPS requirements of the 1,500-user workload (boot, login, steady state, logout, patch storms) at an ultra-low latency of approximately 1ms, delivering an excellent end-user experience. The storage configuration can easily support up to 3,000 users.
  • During all login and workload scenarios, the Login VSI VSImax was not reached.
  • During boot storm testing, throttled booting was able to produce an excellent boot time of 10 minutes for all 1,500 VMs.
  • For all of the nonfailover tests, almost twice as many users could have been deployed with the same results. Only in the cases of the failed-over boot storm and initial login and workload did the CPU average over 50% of performance utilization.
  • Inline deduplication, inline compression, data compaction and FlexClone technology storage efficiency saved over 24TB of storage, which translates into a savings of 50:1, or 98%.

 

Nonpersistent Desktops

In this VMware Horizon 7 reference architecture 1500 nonpersistent Windows 10 desktops were provisioned on an AFF8040.  This AFF8040 was running the latest ONTAP 9 software and configured with 1 shelf of 800GB SSDs.

The testing included the following criteria: For persistent desktops, the following key findings were observed during the reference architecture testing:

  • Launch and boot storm tests of 1,500 desktops (with and without storage node failover) using VMware Horizon View linked clones
  • Login VSI login and steady-state workload tests using Login VSI 4.1.4.2 (with and without storage node failover)

 

Test Time to Complete Peak IOPS Peak Throughput Average Storage Latency
Provision 1500 VM Pool ~80 min 86621 2038 MB/s 457 us
Boot storm test (50 VMware Horizon View concurrent power-on operations) ~14 min 134722 3103 MB/s 373 us
Boot storm test during storage failover (50 VMware Horizon View concurrent power-on operations) ~13 min 109975 2376 MB/s 7.83 ms
Login VSI Monday morning login and workload 19.21 sec/VM 43528 808 MB/s 465 us
Login VSI Monday morning login and workload during failover 19.93 sec/VM 69700 1126 MB/s 802 us
Login VSI Tuesday morning login and workload 19.37 sec/VM 27159 620 MB/s 450 us
Login VSI Tuesday morning login and workload during failover 19.65 sec/VM 44250 1106 MB/s 707 us
Refresh existing VM Pool ~25 min 111789 2079 MB/s 723 us
Recompose VM existing VM Pool ~149 min 98404 2222 MB/s 465 us

 

Key findings for nonpersistent desktops:

  • The NetApp All Flash FAS solution was able to very easily meet all IOPS requirements of the 1,500-user workload (boot, login, steady state, logout) at an ultra-low latency of approximately 1ms, delivering excellent end-user experience. The storage configuration can easily support more users.
  • During all login and workload scenarios, the Login VSI VSImax was not reached.

 

Simplicity and Performance

NetApp All Flash FAS with ONTAP 9 is optimized for VMware Horizon 7 right out of the box.   SAN- or NAS-optimized systems ship with the ONTAP operating system preconfigured to for optimal flash performance, reducing initial setup. New fast provisioning workflow templates for VMware enable you to deploy storage in under 10 minutes from power on to serving data. VMware Horizon 7 delivers virtual or hosted desktops and applications through a single platform and provides end users with access to all of their virtual desktops, applications and online services through a single digital workspace in a few simple clicks.  These validated reference architectures deliver excellent performance and consistent low latencies for both persistent and non-persistent desktops so customers can deploy with confidence.

 

Answers to STORAGE JUMBLE: integrated, policy-based, nondisruptive, backup

Answer to CARTOON CAPTION: NetApp Plug-ins