The college – which has
1,000 students and 120 staff – replaced the EMC NS480 Celerra storage array when it reached end of life and proved costly and poorly configured for the institution’s move to a virtualised environment.
Mark Lane, systems manager at St Richard’s, said the EMC storage area network (SAN) was over-engineered for the college. “In some ways it was very quick, with 200TB capacity in 42U – of which 20TB was flash. It was a datacentre product with trading-floor type capability, but it was end of life and not well configured.”
“We don’t know if it was the configuration of the original system or to do with the skillset that had been available. But, for example, for Cifs [common internet file system] and NFS [network file system] file storage there were issues that slowed down log-in and resulted in data becoming corrupted. This resulted in poor user confidence and log-in times of more than two minutes, which is a lot in a 45-minute class time.”
Lane also tried to use the EMC array to support a roll-out of Citrix virtual desktops, but in terms of storage performance there was no advantage to using traditional fat clients in terms of access times, so the project was abandoned.
The EMC array was at end of life and Lane was offered continued support for five years at £56,000 per annum.
Instead, the St Richard’s IT team started to look at alternatives.
Lane’s team eventually deployed two Scale Computing HC3 hyper-converged clusters, with around 20TB each in four nodes. The Scale hardware comprises compute and 2.4TB Sata hard disk drive (HDD), plus Scale’s own hypervisor environment, in which St Richard’s has deployed 15 virtual machines with no licence fee, as would be the case with VMware, for example.
Hyper-converged products combine compute and storage in one box with virtualisation capability. They have emerged in recent years as competition to discrete server and storage products, with key suppliers including Nutanix, Scale Computing, Simplivity and VMware’s EVO:Rail.
St Richard’s also evaluated servers and storage from Hewlett-Packard (HP) and Dell, as well as converged infrastructure provider Nutanix.
“We found the HP and Dell resellers to be slow and not very reactive to our needs with a tendency to over-engineer systems. Nutanix offered a scalable, one-stop-shop type system, but it did say it was a bit big for a college environment of our size,” said Lane.
He said the key advantages of Scale’s product were that it was quickly delivered and deployed – within a week with data migrated from the EMC array – and it provided much reduced log-in times, from more than two minutes to less than 30 seconds.
The money saved was also a big benefit. Continued support on the EMC storage would have cost £56,000 per annum for five years, while the two Scale clusters came at an all-in cost of £51,000.
Was Lane concerned about lock-in with Scale running a proprietary hypervisor? “Well, we certainly had to ask how big is Scale and will it be there in five years? It was a small gamble, but we were reassured it was a big enough provider. We’ve been using it for two years and it’s solid and reliable with strong support.”