For the longest time (and even now) most virtualisation deployments occur on scale-up storage platforms. However in the past few months several vendors have announced support for virtualisation platforms such as vSphere/ESX (as a starting point) and RHEV on their scale-out file and object based platforms. Is this trend here to stay? What is driving this trend? Let's explore.
I am a huge proponent of scale-out file-based storage....distributed file systems in my opinion have the potential to to dominate the future of storage one way or the other. The most attractive value proposition is the ability to deploy storage using commodity hardware.
However today, they suffer when it comes to multi-protocol support and they also suffer when it comes to supporting multiple and diverse workloads - one such example is server virtualisation. Many businesses see this as an impediment to deploying distributed file systems as the storage platform to host application environments, especially those that are virtualised.
But things may be changing rapidly.Several vendors are adding block support in addition to object interfaces and many others are adding support for virtualisation interfaces. When Isilon launched OneFS7, they made sure to highlight how OneFS7 now supports VMware VASA and VAAI. Just yesterday, Red Hat announced support for RHEV on Gluster. I expect similar efforts from Oracle, Citrix, Microsoft and others.
Meanwhile companies like Scale computing have leveraged IBM's GPFS file system to support virtualised workloads and there are companies like Compuverde that have developed an object based filesystem and optimised it for virtualised workloads. And then there are the likes of Panzura that support virtualisation workloads on a cloud platform.
Virtual images can then be distributed globally using the cloud. Finally VMware is also working on a distributed file system of its own that will create a storage environment using internal disks.
For now all these vendors are treading cautiously. Some of them are quick to point out that businesses should only deploy their test/dev or tier 2/3 environments on their scale-out platform.
Others (especially startups) are going full out in proclaiming that their platforms are absolutely well suited for virtual environments. There is also the fact that many scale-out platforms lack robust data management services that are essential to maintain operational and disaster recovery service levels.
Needless to say businesses are taking a wait and see approach. Many of them are not ready to move their existing environments to these scale-out NAS platforms - especially the ones that use iSCSI or Fibre Channel (note this also signals a general lack of confidence in the robustness of block interfaces in many scale-out platforms).
But this trend is going to continue...and I firmly believe that one day in the not too distant future scale-out platforms will become a viable platform for server and desktop virtualisation.
I know I have not discussed several companies and solutions that support my case for scale-out as a platform for virtualisation - especially the startups...please do let me know if you are such a company and I'd be happy to update this blog.
Posted by Ashish Nadkarni, IDC