A random comment made on twitter a few days ago has been stuck in my mind and been going around and around. It’s finally emerged. Even as we start into 2011 we don’t really have scalable Storage Resource Management products for the Enterprise.
Sure, we have point products that can managed small numbers of arrays. All vendors produce those. They aren’t good; pick something like cross-vendor support – a feature that really doesn’t work and was made worse by the failed implementation of SMI-S by SNIA and you can see my point.
But what happens when we want to scale to hundreds of arrays and multi-petabyte deployments? From my experience what we see is end users falling back to the basic command line interfaces and abandoning the SRM products as the amount of overhead they add in terms of additional management, support and cost exceeds the benefit.
There is also another trend we’re seeing as we move to more comprehensive virtual environments; storage management is being added as a plugin to the virtualisation management platform. This means complex SRM tools aren’t required as the rate of change of the storage component in a unified architecture is much less and in some cases almost zero.
Does this really mean SRM tools have had an impact? I believe it does. Once environments reach a certain size, it becomes incredibly difficult to manage them effectively. How many Storage Admins have been asked the simple questions; “how much storage do we have” and “how much storage are we using/have free” and found themselves having to make excuses about how difficult that is to work out or to calculate across heterogeneous environments.
The virtualisation trend I’ve mentioned also poses another risk; the original premise of the SAN was to consolidate dispersed storage tied to servers. Consolidation made management more simple and reduced costs. However unified computing is being delivered as a package which doesn’t include consolidated storage (it includes storage arrays, but they’re not designed to be shared across multiple unified server/network deployments). As a consequence I believe we’ll see the rise of embedded storage blades and of SSD arrays to support unified virtual servers, with bulk “nearline” storage of data being the storage array’s only purpose.
Could this situation have been changed by better SRM tools? Well perhaps and perhaps not. It partially depends on what you include in the definition of SRM software. I would define SRM to encompass everything from reporting to management, where management means provisioning/deallocation and operation. One critical component of storage management is the ability to seamlessly move data around storage arrays as required. This is a feature only appearing today, some 5 years or more after it was really needed; server virtualisation removed the incumberances of the physical server from the OS/application – storage mobility (which should remove the incumberance of the storage array) is still in it’s infancy.
Hopefully 2011 will be a year where some of the data mobility features finally reach maturity and that can only be good for us all.