Like many other storage enthusiasts, I also attended the live Storage Tech Field Day 5 (#SFD5) presentation by Satyam Vaghani (@SatyamVaghani) of PernixData (@PernixData).  In case you missed it, here is a link to recordings of the presentation, conveniently broken down into areas of focus.

The presentation was about the current version: PernixData FVP 1.5, how it works but more importantly for me, the upcoming features.  In case you don’t know about this great product, it’s hypervisor-level fault-tolerant clustered write acceleration, achieved by utilizing in-host flash devices (currently, with RAM support coming soon – more on that below).  It seamlessly compliments the existing storage to provide significantly lower latencies and higher IOPS.  For more information, see the FAQs.

Being a PernixPro, I already knew about its awesome capabilities as it “powers” my lab so I anxiously waited for the announcement of upcoming features and when they came, they didn’t disappoint. Four new features were announced:

  1. Storage protocol agnostic – due to added support for NFS
  2. Network compression – intelligent compression of data before transmission
  3. Topology awareness – via user-defined replica groups
  4. FVP Clustering using RAM or Distributed Fault Tolerant Memory

Frank Denneman (@FrankDenneman) has blogged about these features here and there’s also another great article from Duncan Epping (@DuncanYB) here so I am not going to repeat all that as I can’t possibly top what they’ve already said. Instead, I am going to say what I think about the fourth one in the list i.e. DFTM as I think that will be key to FVP’s expansion in the market.

I am a fan of everything FVP does and will do in the future as per the list above but DFTM is my favorite of the lot.  That is because, I think, DFTM is the feature that will make it very attractive to a couple of use cases, fueling its popularity even further.

There are a lot of companies out there who not too long ago, invested heavily into virtualization setups with as many hosts (rack/blade) and storage as they could afford and virtualized as much as they could. For one reason or another, some of those companies bought hosts with a “finger in the air” approach, which generally resulted in the maximum amount of memory the host could handle and today, such setups are running at very low memory utilization.  Then there is also a situation where such companies are just embarking on their VDI deployments and although the storage was bought a few years ago, they’re told it’s not good enough to satisfy the latency and I/O demands of virtual desktops. DFTM will allow such companies to use their existing resources without additional investment in terms of hardware.

So far, the requirement to buy/install hardware (even flash) for existing hosts, especially in case of blades, was seen as a barrier to FVP deployment. If my understanding of the technology is correct (bearing in mind I haven’t used it yet), having spare RAM in hosts should be enough.  We all know that FVP is all software, installs non-disruptively and datastores/VMs can be (un)protected non-disruptively as well.  That also makes it quite easy for companies to try out FVP, without any investment or hardware modification and see the difference for themselves.  They can also buy it once convinced afterwards.

I am excited about the new developments and can’t wait to get my hands on the new version as soon as I possibly can. If you haven’t looked at FVP yet, I’d suggest you do now!