The customer is DoeLegal, a US provider of e-discovery and data hosting services. Over 1,000 end-users log into to DoeLegal’s systems each day to use litigation management applications and access data in over 1bn files.
At the beginning of this year DoeLegal was approaching the capacity limits of its two-year old Celerra NS500, configured with 12TB of capacity.
It called in EMC to recommend a way to extend that capacity, but eventually installed 16TB of new storage comprising NAS heads or controllers from Reldata, front-ending disk arrays from Infortrend, at a total cost of around $150,000.
Among the reasons for rejecting EMC was the fact that although the NS500 is powered by dual controllers, the company could not attach any disk to the second controller – at least not with failover between the two nodes.
We could get to 16TB in an active-passive configuration. But that wasn’t even close to what we needed, said Chalkley Matlack senior network engineer for DoeLegal.
With end-user logging in from 14 countries in multiple time zones, DoeLegal wanted that failover protection and availability.
Unlike Reldata, EMC only provides failover between active-passive and not active-active pairs of controllers. Matlack said that this was a major reason why he estimated that staying with EMC would have cost 50% more than moving to Reldata.
Reldata is not the only NAS supplier supporting active-active failover. EMC’s biggest rival, and the other giant in the NAS market, NetApp also supports active-active failover.
EMC said there is no automatic failover between active-active Celerra controllers because the company wants to provide no-compromise availability, with not risk of a reduction of service after a failure, or in its words no unhappy surprises when the same single processor [controller] is doing replication, file serving and RAID rebuilds.
And EMC went on to say that suppliers offering failover between active-active pairs will in any case recommend no more than 50% load on either controller, in order to avoid those surprises.
Is that what NetApp recommends? NetApp said its advice is to make sure each node can handle post-failover workload, but said that this does not always limit load to 50% of each node’s capacity, because applications are not usually running at peak load.
Having active-active controllers allows the user to take advantage of the memory, CPU, bus capabilities and other resources of both controllers during normal operation, said NetApp’s Rich Jooss, SANiSAN Business Unit, NetApp.
Meanwhile RelData and Infortrend are happy to have won Doelegal’s business. DoeLegal is using RelData’s 9240 unified gateways with a total of 24TB of Infortrend RAID 6 storage. The RelData boxes are replicating data – including data stored on Doelegal’s Celerra – to a disaster recovery site.
Doelegal was working with integrator Selenetix, which said that although it prepared a five-step project plan, the limited cost of the job did not justify an ROI calculation. Was Selenetix concerned that Doelegal was buying from a start-up with less certain prospects than giant EMC.
There’s too much emphasis put on that. It shouldn’t be a deal-killer, but it keeps people from looking at valid start-up alternatives. That EMC box was delivered two years, so that’s already been amortized off the books, said Selenetix CEO JD Wilson.
