Washington Archives Management Deploys DataCore SANmelody

DataCore Software announced that its SANmelody software has been deployed at Washington Archives Management (WAM) to serve as a high-availability, virtualized, storage area network (SAN) that works in conjunction with virtual servers from Citrix to deliver the redundancy this customer was desperately seeking. Now WAM has a fault-tolerant storage virtualization infrastructure that provides automated failover and failback recovery in the event of a network or hardware calamity in order to maximize business uptime. WAM is a 50-person archiving service that stores, archives and manages paper records.

“A forward-thinking approach combined with technological advances in virtualization has meant that our firm could avoid past failures,” says Tom Radford, System Administrator, Washington Archives Management. “And by deploying redundancy in our network – as well as leveraging advances like automated failover and multi-pathing, XenMotion and – above all – a software-based SAN that delivers on the promise of virtualization, WAM has made it possible to eliminate costly, storage related downtime.”

WAM’s bread and butter is hardcopy storage. It also engages in document imaging and X-ray digitizing. WAM archives all of these various documents offsite as part of its service. For over two years, DataCore partner Moose Logic has managed the whole network for WAM. “However, when a hardware failure occurred we took that failure as an opportunity to upgrade and grow our infrastructure,” explains Radford. “DataCore partner Moose Logic was tapped to re-architect how our company addressed storage and did so by deploying a SAN solution from DataCore.”

A hardware failure leads
to a fresh start on the storage front
“Two ESX servers had WAM’s environment virtualized onto them, but the ESX servers had no shared storage,” notes Steve Parlee, Senior Systems Consultant, Moose Logic. “The two servers were running local, internal storage – and one had a direct-attached drive shelf. This was not a very convenient set-up because we could not, for example, VMotion the guests between servers in case of a failure.”

Moving the virtualized guests to the other server (in case of a hardware failure) was difficult because there was no shared storage – no SAN was in place, which is necessary to ‘motion’ the virtual guests. As fate would have it, the firm did experience one ESX server failure and therefore was forced to run on only one ESX server. Without redundancy in the network, if any other component had failed, Washington Archives Management would have been ‘out of business’ until it was fixed.

WAM considered two options. The firm could either go back to where it had been with the two VMware ESX servers and simply try to become more stable, or it could do some forward-thinking and put in place a SAN that would make the consequences of hardware failure a thing of the past. “We could have just replaced the ESX server and stayed on the path we were – risking future hardware failures that would yet again impact business operations,” says Parlee. “The choice Washington Archives Management made was to go with a shared storage approach and to use DataCore as the foundation of that whole strategy.”

Utilizing XenServer
and shared, network storage – a ‘no-brainer’
Moose Logic advised WAM to deploy DataCore because it allows the firm to replicate in real time their storage between two nodes. In addition, Moose Logic replaced the ESX servers with Citrix XenServers. With the XenServers, Moose Logic was able to utilize something called multi-pathing. “With muti-pathing, we have been able to take down half of WAM’s network for maintenance – without impacting day-to-day operations,” comments Parlee.

“Because of the shared storage and the fact that the storage is virtual, WAM now has the ability to ‘XenMotion’ the guests back and forth between two different pieces of hardware,” continues Parlee. “The other notable capability is high availability. If any of the Citrix XenServers were to crash, it will automatically restart on the other server. And what makes this possible is the back-end, shared storage.”

What accounts for the storage being ‘virtual’ lies in the fact that DataCore enables a system administrator to take a ‘volume’ (i.e. a piece of disk space for storage) from each of the two SANmelody nodes and mirror them. When these volumes are mirrored together, this creates a single ‘virtual volume’ which is what is then presented to the application server (e.g. XenServer, Windows Server, etc.). By doing this, the application servers only see the one instance of the ‘virtual volume,’ but in reality that virtual volume is tied to the mirrored volumes. It is this key feature – inherent to SANmelody – that enables the system to failover.

A virtualized IT environment – radically simple,
high availability storage virtualization solution
The redundant SANmelody SAN servers that are currently in operation on the back-end are mirrored to each other using synchronous replication. “It is all about the redundancy – the fault-tolerance for failover,” explains Parlee. “We can lose up to three components – two servers and a switch – and everything keeps running. You see, either one of the SANmelody servers can present storage. Only one is presenting it at any time, but if one goes down, without impacting the environment whatsoever, the system seamlessly switches over to SANmelody number two and just keeps right on running.”

Summarizes Radford: “The bigger picture here is that technology like this has previously been prohibitively expensive for a small to mid-size business customer. Sure, large companies have been deploying virtual servers and virtual SANs, but they have had to pay a huge price for doing so. DataCore’s SANmelody enables the small business to get this kind of virtualization environment built and to do so without paying for Fibre Channel or paying over $100,000 for the SAN alone, which is what has made having a SAN unattainable for a large number of customers – before DataCore paved the way.”