View 4 – Reference Architectures

Well – it’s finally here!   We’ve been lucky enough to be playing with View 4 here internally for a while – in fact we use it extensively around the super labs as in my earlier post (and for analysts coming to EMC Analyst days this week, you’ll see me using it tomorrow morning…)   We have several hundred View users within EMC, and there is an agressive goal is to make it available to 100% of the employees by mid-year in 2010.

Congrats to Jocelyn and the whole View team – this is a big release, with a lot of additional features, and dramatically improved experience in broader use cases.

As always – loads of work behind the scenes…  You can read what EMC and Cisco have done to support View 4 in the PR here.   So – without further ado, further here are supporting reference architectures and validation work you can leverage…

VCE Vblock Type 1 View Reference Architecture

The VMware, EMC and Cisco teams have created a joint reference architecture on a Type 1 (mid-range) Vblock.   This Reference Architecture is targeted for customers with low to mid thousands of clients as their scaling target.   High note of the testing work: $750/client – all in.   Higher user density than we were able to get from View 3 on VI 3.5 in the past.  Since View 4 can work on vSphere 4, we get more density, and overall better client performance.   A lot of the Virtual Center scaling and responsiveness we struggled with in large-scale View 3/VI3.5 work is now also resolved.  

Click on the doc below to get the whitepaper which is hosted by VMware.   Thanks to the View team, the EMC and Cisco Solutions teams!

Looking for something a little smaller? 

Celerra NS-120 Validation Test Report

The EMC Celerra Solutions team have published an updated View 4 Validation Test Report (contains a lot of performance data and step-by-step information) targeted towards use cases for customers with hundreds of clients.   This leverages a ~250 user building block approach.   Click on the doc below to get the whitepaper which is hosted by VMware.

What to expect soon:

• Even more Vblock testing
• The RSA SecurBook on View – very cool doc that covers how to futher secure the platfom, as well as the virtual desktop itself.

Comments

[David Barker]

Hi Chad, would you mind a request?...

When the army's not busy, would it be possible to re-run the NS120 tests with SSD for the gold images & SATA for the user data? I'm guessing most of the I/O was against the linked clone bases - a little SSD might go a long way :-)

[Chad Sakac]

Unfortunately, in View 4, you still can't put the base replica and the linked clones on seperate datastores.

This will be solved soon - either via FAST v2 (block level automated tiering) or View 4.1 which will enable that use case.

[Aaron]

David was referring to separating the boot LUNs from the user data which could be housed on CIFS shares backed by SATA storage, not separating the linked clones from the replicas (solved by FAST and/or future View releases). Testing of boot LUN deployments on EFD (SSD) drives has been done. As of now we are seeing 4-8x the desktop density per "spindle" and roughly 50% better average disk response times as compared to 15k Fibre drives.

[David Barker]

Hi Aaron - thanks for the update. If you get a chance, could you forward on any test results you have?

fyi: We have several lab pc's (~1000) all running the same image that we think VDI may be good for. In tests, linked clones of our image cost about 1Gb each, so we are looking at ~1Tb to store everything, requiring ~3000 iops. SSD seems to be a perfect fit.

[Brian Gracely]

@Aaron (or anyone else on the team),

When you plan to include the details within the paper, similar to what was done on the V-MAX VDI document from a few months ago. When you talk about $750/desk, what does that include (mostly interested in the storage side...is it all EFDs, how many disks were required, etc..).

Also, should we expect that the NS-120 report will be similar to the storage component of Vblock 0?

[ron]

Hi Dave,

Id suggest looking at NetApp with PAM Cards. Seems an obvious fit here

cheers
ron

[twitter.com/stor2]

Chad: What kind of VIC/NIC's/CNAs were used in the VMWare View exercise/Vblock1?

[Mark Bowker]

An ESG research report from early 2009 indicated that approximately 21% of organizations are using a VDI solution with another 8% planning to do so. Among these early and planned VDI adopters, only 15% plan to deploy the technology to all or most of their employees. The vast majority (82%) of respondents will implement VDI on a much more selective basis, either restricting it to certain employee types now and for the foreseeable future, or implementing it on a limited basis now with expectations of rolling the technology out to a wider audience at a later date.

The reference architectures are great, but how are you helping customers with POCs (proof of concept)?

[Chad Sakac]

@ Ron - PAM is NetApp's response to the challenge, and it certainly helps. That said, PAMs are in effect a cache, not non-volatile storage, which means they assist with some of the IO density challenge. Not suggesting it's bad - more use of RAM in hosts, in storage as cache, or flash in the form of non-volatile disk - it's all goodness.

Conversely, EMC's approach is SATA coupled with SSD - using View Compooser and other composition techniques. Regardless of the way you go, leverage the learning we're doing with customers around the world to help you!

@stor2 - we were using Palo adapters in the Cisco UCS.

@ Mark - thanks for the comment. We're doing hundreds (thousands?) of PoCs as we speak for customers of every shape and size. The reference archictectures serve as a baseline for customers to start with predictable building block configurations. On the VCE Solutions Support Team (the joint selling/technical teams) we have "hyper specialists" focused on VDI use cases. Those folks get engaged and support customers through their PoC phases.