As more and more vendors introduce primary storage de-duplication, we need to consider the impact this has on data protection and recovery.
Primary storage de duplication does not make disk any less reliable but it does magnify the impact that a failure could have and if we align this with the ever larger spindles and also the ongoing trend to stripe data across many disks, along with automated storage tiering, we are increasing the impact of failure. We are putting more eggs into one basket.
If we consider the impact of the failure of a de-duped virtualised server environment, this could take out a large proportion of a data centre. Restoration of many thousands of images from tape will not be viable. If we have a truly dynamic environment with virtual machines being brought up on demand; we need some way of storing state and recovery of state.
But 'Bod, how likely is this to happen? RAID ranks, especially RAID-6 protected do not fail often but they will fail. Is it more or less likely for your data-centre to burn down than a RAID-6 protected rank to fail? Well only you can make the risk assessment for your business but if you have built redundant data-centres, you are already catering for a certain amount of unlikely catastrophic failure but what constitutes a catastrophic failure may be changing.
Replication will have to become standard practice for nearly all environments as will seamless automated storage failover. You may even consider local array replication depending on distances between data-centres and complexity of individual component failover.
Yet even with this overhead, reduplication will still make a lot of sense for many workloads; just make sure you do proper risk/impact assessments.
