Comment Although not officially announced, it looks like the acquisition of PernixData by Nutanix is a done deal.
Confirmation came from Frank Denneman, Dutch blogger and former tech evangelist for Pernix who has now left the company.
More ReadingEnterprise storage is a stagnant – and slightly smelly – pondNutanix has swallowed something. Cough once if you're OK, guysHyper-converged top-of-the-pops list is out: Guess who hasn't crushed it?Software-defined storage is glitzy, but E8 and Nimbus are still delivering hardwareEx-CTO: Yes, Nutanix is buying PernixData
So what can we make of this? What benefit is there in taking a caching technology and adding it to Nutanix's existing hyper-converged portfolio?
PernixData FVP and Architect
PernixData actually has two main products, FVP and Architect.
FVP is the VMware-only caching platform that accelerates I/O and takes the load off external storage arrays. In fact the benefits are two-fold, significantly reduced latency for I/O (with read I/O benefiting the most) and the increased throughput of serving read I/O directly from flash and/or DRAM cache. I’ve used FVP and it definitely works. The problem is though it’s obviously VMware-focused. Pernix didn’t move outside that boundary and produce anything for Hyper-V or the other lesser known virtualisation platforms.
Architect is PernixData’s analytics tool. It collects information from FVP deployments and can be used to analyse traffic and efficiency ratios of the platform. Architect produces some interesting statistics (and Frank Denneman has blogged on the output of those); however the Achilles Heel of the product is that it doesn’t have a holistic view of every part of the infrastructure.
There’s no integration with the backing store (the shared storage platform) or integration within the OS of the guests running on the hypervisor. This provides only limited capability and can lead to making assumptions on how well caching is performing.
Just be clear here – I’m not saying Architect isn’t useful. To go to the “next level”, however, array integration and awareness is needed.
PernixData and SDS vendors
The most obvious acquisition path for me was for Pernix to be acquired by an SDS (software-defined storage) vendor. Acquisition by a hardware vendor doesn’t really make sense as FVP (for example) is all about reducing the amount of external storage in use.
FVP is a great solution for improving performance at a potentially lower cost than buying more external hardware. A software vendor could be more appropriate as this would provide a more joined up solution that integrated both the hypervisor and external software-based storage.
PernixData and Nutanix
What about a tie-up with Nutanix? There are now two main Nutanix solutions: Prism and Acropolis. Prism is the management platform that runs across multiple hypervisors. Acropolis is Nutanix’s own hypervisor platform based on KVM, but modified and hardened for security.
This storage layer isn’t integrated into the kernel of ESXi as say Virtual SAN would be, and that may be an area where Nutanix is looking to deliver performance improvement. FVP would certainly do that. Additionally, Pernix hase the skills in writing VIBs, software plugins for the ESXi kernel. That may be another reason for acquisition. There’s also the requirement to bolster their own Acropolis platform built on KVM and the aim could be to provide more performance capabilities into this platform.
The Architect’s view
There are lots of places where PernixData’s IP could be used by Nutanix and it makes sense to acquire if the opportunity and price is right. At the same time, there’s nothing to stop Nutanix continuing to sell FVP and Architect into existing VMware environments because it provides valuable intelligence on how those systems could be migrated to Acropolis.
This could have been Nutanix’s aim all along: use FVP as a beachhead into new VMware accounts and move them over to Prism and then Acropolis. So clear synergy exists. I just hope the hard work of the PernixData team continues to live on. ®