Home | Cloud | Violin Memory Delivers Converged Storage
Violin Memory Delivers Converged Storage

Violin Memory Delivers Converged Storage

3 Flares Twitter 0 Facebook 0 Google+ 3 StumbleUpon 0 Buffer 0 LinkedIn 0 3 Flares ×

As mentioned in this recent post, there’s a lot of activity in the market looking at combining storage and compute not just in the same rack (version 1.0) but within the same server infrastructure itself.  We can look to Nutanix as being the first to bring this to market in a commercial product and where they have led, others have followed.  Most recently we have seen Violin Memory Inc, well known for their all-flash array, release a solution that integrates Microsoft Windows Server 2012 R2 into their hardware platform.

In some respects it makes sense to use a storage array for virtualisation.  After all, the components are basically the same in many of today’s storage platforms.  As an example, there have been rumours that EMC would do something similar with VMAX or VNX, which is a story for another day.  The tricky part of running the hypervisor and storage together is in the detail.  Here’s some information on the Violin solution.

W2K12 R2 runs on both of the memory gateways within a 6000 series array.  Each instance has 16 cores (2x 8 core processor) plus 24GB of memory with the amount of dedicated resources configured at boot time.  For resilience there is a hidden heartbeat mechanism that operates between the two clustered W2K12 instances.  Licensing is included within the cost of the 6000 array, although I don’t have specific details on this.

Strengths

It’s a nice idea to run a hypervisor on the storage array, given that sufficient resources are available to make this workable.  In device that already offers high latency, removing the storage fabric too provides for extremely high speed operations.  Windows Server 2012 comes with a raft of new storage features including Storage Spaces, which are designed to work with JBOD devices and effectively deliver thin provisioning and de-duplication for any virtual machines deployed within the array itself (as these features don’t natively exist).  For highly intensive workloads, this solution seems like a good deal.

Weaknesses

Of course there has to be some downsides.  Let’s think these through.  This offering seems like a good for server and desktop virtualisation however I can’t help thinking that the resource ratios of compute/memory/storage might be biased too much in terms of having more IOPS and storage capacity than can be effectively used.  Certainly the memory capacity seems too low for any useful scaleable Hyper-V deployment.  Then there’s the question of scale-out.  Violin arrays don’t scale out today, so scaling this solution would mean connecting isolated islands of compute/storage together using Infiniband.  It would be very easy to have fragmented and unused resources in each deployment; not a cost effective solution.  So how would things scale?  Can I keep increasing memory and CPU, or am I bound to the fixed deployments documented?

The Architect’s View

For this solution to be credible, we need to see some good examples of workload running on the box; many VMs and desktops or a large-scale SQL or SharePoint deployment.  The issues of scaling need to be explored and explained; how could this solution become a scale-out architecture?  Much of this may be in the works and yet to see the light of day.  Will Violin support VMware?  I don’t see why not; we’re moving to a scenario where the hypervisor is almost free and is, in some cases.  Converged 1.0 was hardware in the same rack.  Converged 2.0 is compute/storage/network on the same physical platform.  It wouldn’t surprise me to see other vendors moving forward with these kinds of solutions, as the future of the dedicated storage array steadily declines.

Related Links

Comments are always welcome; please indicate if you work for a vendor as it’s only fair.  If you have any related links of interest, please feel free to add them as a comment for consideration.

Subscribe to the newsletter! – simply follow this link and enter your basic details (email addresses not shared with any other site).

Copyright (c) 2013 – Brookend Ltd, first published on http://architecting.it, do not reproduce without permission.

About Chris M Evans

Chris M Evans has worked in the technology industry since 1987, starting as a systems programmer on the IBM mainframe platform, while retaining an interest in storage. After working abroad, he co-founded an Internet-based music distribution company during the .com era, returning to consultancy in the new millennium. In 2009 Chris co-founded Langton Blue Ltd (www.langtonblue.com), a boutique consultancy firm focused on delivering business benefit through efficient technology deployments. Chris writes a popular blog at http://blog.architecting.it, attends many conferences and invitation-only events and can be found providing regular industry contributions through Twitter (@chrismevans) and other social media outlets.
  • http://blog.fosketts.net sfoskett

    One little niggle here, Chris: Microsoft’s built-in data deduplication does not support active virtual machines except in VDI instances. And dynamic VHDX isn’t really fully thin provisioning, either. Although Hyper-V is good, it’s not that good.

    • http://thestoragearchitect.com/ Chris M Evans

      Stephen

      Just to qualify this; when you say “support” do you mean it does or doesn’t technically work? I assume you mean it does work, but Microsoft won’t provide user support.

      If that’s the case (and bearing in mind you are with MSFT at the moment), why is this? Thoughts?

      • http://blog.fosketts.net sfoskett

        I don’t believe you can run CSV and dedupe on the same filesystem. And Microsoft will definitely go nuts and yell if you try to do it in production. They are very clear on this. It’s not “we will look the other way” it’s “DO NOT DO THIS!”

        Windows only has post-process dedupe (MUCH delayed post-process) so it doesn’t provide much value anyway. It’s provided by a Windows Server service, not integrated into the filesystem stack. I imagine Microsoft is very concerned that active filesystems will quickly exhaust available space and leave you sunk.

        So I definitely wouldn’t recommend using Windows dedupe under live Hyper-V hosts.

  • http://blog.fosketts.net sfoskett

    One important application of this Violin capability is as a Windows 2012 File Server. Microsoft and Violin demonstrated a crazy fast Windows file server embedded in Violin using SMB Direct, and that’s the real application of this. Since Windows Server 2012 has the only production/supported SMB Direct server, this is an excellent way for Violin to be first to market with such a solution.

3 Flares Twitter 0 Facebook 0 Google+ 3 StumbleUpon 0 Buffer 0 LinkedIn 0 3 Flares ×