Hyper-V CSV Cache competing with VMware CBRC?

Yesterday I got intrigued by a tweet from Andre Leibovici linking to this article stating that Citrix employees (sales) are leaving Citrix for VMware, in the article he explains why they’re swapping Citrix for VMware. Andre states ‘ The main reason for the vendor move according to these new VMware employees was the vSphere stack’ and goes on about features in the vSphere stack that are only leveraged by Horizon View.

Although one could argue this is a limitation as vSphere apparently isn’t that transparent for the top layers, I won’t be going into that discussion. What did triggered me was the following feature:

 

CBRC

Content Based Read Cache is a solution that leverages host based RAM as cache location for read IO’s. These IO’s won’t reach the underlying storage system and can contribute to the scalability of the environment.

As Andre states CBRC has been introduced to help with IO storms that have a read-intensive load such as OS (re)boot and AV scans. I’ve been looking for some benchmarks and found the following blogpost View Storage Acceleration (CBRC) by VMSkills. I also found this graph from Dwayne Lessner on the numbers when CBRC is enabled:

CBRC

CSV Cache

While I was searching for information I remembered Jarian Gibson talking about CSV Cache in Hyper-V.

CSV Cache is a feature that allows you to allocate system memory (RAM) as a write-through cache.  The CSV Cache provides caching of read-only unbuffered I/O.  This can improve performance for applications such as Hyper-V, which conducts unbuffered I/O when accessing a VHD file.  Unbuffered I/O’s are operations which are not cached by the Windows Cache Manager.  What CSV Block Cache delivers is caching which can boost the performance of read requests, with write-through for no caching of write requests.

From this blogpost: “CSV Cache will deliver the most value in scenarios where VMs are used primarily for read requests, and are less write intensive.  Scenarios such as Pooled VDI VMs or also for reducing VM boot storms”.

I’ve converted the data from this blogpost into a graph to display the results found by Yusuf Ozturk:

One of the big plusses of CSV Cache is that you can run Hyper-V with VHDs over SMB3 on scale out file clusters as you can read here. There’s a presentation from File Server PM Claus Joergensen that shows the impact of using the CSV cache in a VDI environment:

CSV Cache

CSV Cache

You can see the full video for this TechEd presentation at: http://channel9.msdn.com/Events/TechEd/NorthAmerica/2012/WSV410.

Comparing them

So comparing those two techniques they’re both used to use the internal RAM of a host to cache read IO’s to the used storage system and thus reducing the amount of reads to the storage and optimizing the usage.

We can see both solutions provide added value as they enhance the performance of the systems by using the internal RAM from the host running the hypervisor.

While CBRC is supported for Horizon View only, CSV Cache is supported for XenDesktop so when looking at the two technologies it’s save to say that the combination of Hyper-V with Citrix XenDesktop 7 could be just as appealing as vSphere with Horizon View as they’re both leveraging their own cache solution to optimize performance.

One of the comments I got while getting this reviewed was why there’s no mention of XenServer’s IntelliCache. I didn’t want to mention IC as it’s disk-based instead of RAM-based so although you can make a comparison it’s a different solution all in all.

The following two tabs change content below.

Kees Baggerman

Kees Baggerman is a Staff Solutions Architect for End User Computing at Nutanix. Kees has driven numerous Microsoft and Citrix, and RES infrastructures functional/technical designs, migrations, implementations engagements over the years.

7 comments

  1. Hi Kees,

    Good to see that both Microsoft and VMware are starting to mitigate the most common problem when virtualising interactive desktop systems: IOPS. What’s interesting is that while VMware CRBC is only available for VMWare View, not Citrix XenDesktop. Hyper-V CSV Cache is supported for XenDesktop but VMware View isn’t supported on Hyper-V making them mutually exclusive.

    Maybe I’m stating the obvious here, but there’s a software alternative that’s not only supported on both VMware AND Hyper-V (allowing XenDesktop customers to run on the VMware hypervisor) but also optimises write IOPS. Sure nonpersistent machines have less write IOPS then persistent machines, but Windows simply keeps writing (temporary and useless) blocks of data that are a waste of your expensive SAN! That same software solution can leverage both RAM and local disks (SSD / FusionIO) – Like XenServer Intellicache – making it more versatile.

    It will be interesting to see if this is a reason for Citrix XenDesktop customers to take (another) look at Hyper-V, or the software solution.

    Cheers,
    Ingmar

  2. Hi Kees
    Good comparison. As Server 2012 R2 introduces automatic storage tiering, I believe the Microsoft solution is the most complete one from a single vendors perspective. You can now also leverage up to 80% of your SOFS nodes RAM for CSV Cache and use SSD based and clustered storage spaces. VMware IMHO has too much focus on VDI in their enhanced feature stack. Traditional high peak server workloads have demands for those improvments too.

    Cheers
    Michael

  3. I could be mistaking, but isn’t VMware CRBC the result of the Virsto acquisition?
    http://blogs.vmware.com/vsphere/2013/06/introduction-to-vmware-virsto.html

  4. k.baggerman says:

    VMware released version 5 (including CBRC) of VMware vSphere On August 27, 2012, the acquisition of Virsto was on Feb. 11, 2013.

  5. gilwoodcs says:

    To follow up Ingmar’s point there.. Atlantis Computing have a software solution that works with ESX, Hyper-V and Xenserver to improve not only the read IOPS, but the write IOPS as well.

    Read performance very useful,say, in a boot situation. But, how quickly the VM boots is going to impact the user for a point in time. Having write optimisation will ensure a fast, consistent experience for their entire session. I wouldn’t consider Atlantis ILIO only as a replacement to CRBC – use them together and you’ve a very powerful storage optimisation solution and enhancement for your VDI environment.

    If you’re off to E2E in Rome – a virtualisation conference I can thoroughly recommend – by all means drop into our session 😉

  6. […] When Infinio supports Hyper-V, it looks like I’ll need to compare us to CSV Cache as well. […]

Leave a Reply