Global recruitment company Hudson has dumped NetApp for HP Lefthand storage as part of a datacentre consolidation and disaster recovery project that saw it cut energy use by 50% and double storage efficiency.
The project began as a move to a single European datacentre in London to support the roll-out of Hudson's homegrown CRM system. But, it soon became apparent that a secondary disaster recovery site was needed and the existing NetApp storage would have to be augmented or replaced, according to IT director Europe Bas Alblas.
"We had NetApp storage, and there were no problems with it, but once we had the single datacentre it was a single point of failure, so we needed a second site," he said. "We decided we wanted synchronous replication between sites and that we should be able to provide instant access to the most critical apps in case of an outage."
Alblas's team evaluated what NetApp could offer, but it proved too costly. The company opted for HP storage. It had an existing relationship with HP as a supplier of servers.
Hudson implemented two HP Lefthand P4800 unified storage nodes at each site. HP MDS600 disk shelves house 72 600GB SAS drives for each node, which support HP servers using iSCSI via 10GbE networking.
The datacentres operate in an active-active configuration, with HP storage supporting HP Proliant servers that run Citrix XenApp thin clients running on VMware.
Alblas said HP won out over NetApp on value-for-money terms.
"HP gave the best bang for the buck, with lower cost and more functionality," he said. "We have reduced storage rack space and power usage by 50%, with three times more storage capacity and storage performance doubled. We have also achieved 50% more users per Citrix server and reduced the Citrix virtual server count by 30%."
Another factor in supplier choice was that Hudson wanted an active-active configuration with synchronous replication between primary and secondary arrays. It opted to have the same storage hardware at each location to make that easier to achieve. It could possibly have re-used its existing NetApp storage hardware at the disaster recovery site to achieve the same ends, but that would have been a complex job, said Alblas.
What could HP improve in future versions of the P4000 series? Alblas said work needed to be done by HP and VMware to improve iSCSI communications.
"The main issue we have is the way VMware communicates with the SAN is not smart. VMware hosts at the main site sometimes communicate with the secondary site when they should be talking to the primary SAN, and this puts strain on the WAN link," he said.
"HP and VMware need to work to improve the iSCSI stack so that this doesn't happen with hot/hot sites like ours," added Alblas.
