The University of Bristol has completely migrated to HPE Nimble hybrid flash storage from legacy EMC hardware as it migrates its two on-campus datacentres off-site to two colocation sites.
The move to Nimble has allowed it to save an entire full-time equivalent (FTE) resource in the IT department that was previously devoted to storage issues, while providing storage performance “grunt” for all current workloads.
The university has approximately 23,000 students and 6,000 employees. Its IT department runs two datacentres with around 1,300 Hyper-V virtual machines and hundreds of applications.
Before it deployed Nimble, it was an EMC customer with Clariion and Celerra block and file storage (both now long discontinued). These reached effective end of life and the university faced a forklift upgrade with a full migration and downtime.
Harvey Ditchfield, storage and databases team manager at the university, said: “We could have got more capacity, but we had come up to the limits of performance. The EMCs didn’t have the extra ‘grunt’ we needed and management was taking up an entire FTE with problems and performance issues.”
Ditchfield said his team wanted storage for which they could expand capacity, performance and cache independently. When a Cristie Data representative turned up with a Nimble array in his car boot and convinced the IT team to try it, they were won over.
“It outperformed our existing storage by 2x at least,” said Ditchfield.
The university’s first Nimble deployment was a pair of CS260 arrays with capacity of about 30TB.
Now it has around 300TB, which will be at each of the two colocation sites to which it is transitioning. There is a TS5000 and CS500 array at each site.
The university’s Nimble arrays have been hybrid flash from the start, with about 10% of capacity as SSD.
Read more on flash storage
- When it comes to choosing between hybrid flash and all-flash storage, the question is increasingly not how much flash is enough, but whether you still need any disk at all.
- NVMe flash offers blistering performance gains but so far the big five storage array makers have tended to opt for gradual implementations rather than radical new architectures.
“We probably don’t need that much IOPS, but being able to scale cache, performance and capacity independently represents value for money for us,” said Ditchfield. “We can deliver very high I/O and very low latency.”
He described the pool of storage created from flash and spinning disk as a “giant bucket” that doesn’t have to be managed by his staff.
A key benefit for Ditchfield is the InfoSight management platform. “We got an email telling us we would run out of flash cache in 30 days’ time,” he said. “With the old arrays, we wouldn’t know until three or four weeks after, never mind a month before.”
HPE bought Nimble for $1.2bn in early 2017. Nimble Storage was established in 2008 and first made its name with hybrid flash arrays that combined small amounts of Nand flash with spinning disk and used data reduction.
Its hybrid flash boxes scale from a few TB into the low 100s of TB, with functionality that allows customers to specify that certain workloads can be pinned to flash storage for guaranteed performance.