In a Minute: Software Defined Storage
As 2011 was a year of us talking about “Cloud”, closely followed by the “Big Data” wave of 2012 then 2013 is shaping up nicely as the year of the “Software-Defined” entity, where multiple technologies are being covered by the “SDx” banner. Let’s have a brief look at what this means for the world of storage.
In the world of data we are used to constants; Controllers that manage the configuration of the environment and the placement of data, disks grouped together using RAID to protect data and the presentation of this data to servers using fixed algorithms. In effect when we wrote data we knew where it was going and could control it’s behaviour, we could replicate it, compress it, de-duplicate it and provide it with the performance level it needed, and when it needed less performance, then we just move it somewhere else – all controlled within the storage array itself.
Software defined Storage changes this model; it can be thought of as a software layer, put in place to control to control any disks attached to it. The storage services we are used to (snapshots, replication, de-dup, thin provisioning etc) are then provided to the Operating System from this layer. This element of control software will be capable of sitting on commodity server hardware, in effect becoming an appliance initially at least, and will be able to control commodity disk storage.
This does not really constitute some of the features of storage virtualisation, where a control plane manages a number of storage resources, pooling them together into a single entity; rather it separates the management functionality removing the need for the storage controllers – the most expensive part of a data solution. Therefore one of the driving factors for the uptake of Software Defined Storage is an obvious reduction in cost, and the ability to provide data service regardless of the hardware you choose.
The challenge to this is that data should be regarded differently to other aspects of the environment; data is permanent, packets traversing network are not, and even the virtual server environment does not require any real form of permanence. Data must still exist, and exist in the same place whether power has been present or not. We are now starting to see a generation of storage devices, note I was careful not use the phrase arrays, which are looking more capable of offering a Software Defined storage service, through the abstraction of the data and controller layers.
So what does this all mean for storage in the datacentre?
My main observation is that physical storage arrays will be with us for a long time to come and are not going away. However the potential for disruption to this model is greater than ever before, the ability to use commodity type storage and create the environment you want is compelling. With the emerging ability of software to take commodity hardware, often from several vendors simultaneously and abstract the data layer then the challenge to the traditional large storage vendors becomes a real and present danger.
I believe the rate of change towards the software defined storage environment will ultimately be more rapid and see greater early adoption to the proven concepts of server virtualisation, it will cause disruption to many existing major vendors, but ultimately end-users will still require copious amounts of disk technology, so the major players will remain exactly that. Whilst some niche players may make it through the big boys will still dominate the playground.