In doing so, it has cut its on-site hardware footprint from 24U to 4U, slashed equipment and licensing costs, and reduced data restore times from hours or days to minutes.
The move also gives Cranfield peace of mind in disaster recovery by gaining the ability to run all operations from any location using virtual servers running in Azure, should the entire site become unavailable.
The refresh comes alongside one in which the university replaced its existing Pure Storage flash storage arrays with 12 nodes of Nutanix hyper-converged infrastructure hardware.
The entire project is a drive towards simplifying Cranfield’s on-site physical infrastructure in a move that encompasses cloud as a site for storage (and compute in case of outages).
Cranfield is a leading research establishment in science, industry and technology, with 1,600 staff and 4,000 postgraduate students.
Its IT stack is based around Microsoft and Linux servers with Microsoft and Oracle-based applications. It is effectively 100% virtualised on VMware, with 400-600 virtual machines running at any one time.
Its existing backup infrastructure was based on Veeam backup software and Data Domain hardware, with replication to a third party-hosted Data Domain box.
That setup had reached end of life and was showing the signs, said head of IT infrastructure Edward Poll.
“Data Domain did what it was supposed to do, but it was time to refresh things and we wanted to reduce costs, management time and complexity, and increase performance,” he said.
“The major issue with Data Domain had become restores. It ingests well, but recovering was more problematic. It would be fine for one restore, but if we’d had to restore multiple – 50, 100 or 150 – servers, we would have struggled.”
Cranfield’s IT department had already started a journey towards cloud by using StorSimple appliances – with about 80TB on site and 0.5PB in the Azure cloud – and had discovered how cost-effective it can be.
“Azure was a good fit and we started by thinking we could use Veeam and Data Domain instances in the cloud, but it was suggested to us, ‘why not get rid of a layer of software?’, and we looked at using Rubrik appliances,” said Poll.
Rubrik is part of an emerging category of backup appliances that come as nodes that build into clusters in a similar way to hyper-converged infrastructure.
Rubrik’s software appliance can come on approved server hardware from Cisco, HPE or Dell with flash and spinning disk inside. Capacities for a minimum four-node cluster are in the 64TB-160TB range, depending on the hardware.
Customers can set policies to specify how long data should be retained as a backup and which can be accessed for production use from Rubrik hardware. Rubrik backup data is seen as an NFS file share before being sent to an in-house physical archive or the cloud.
Cranfield has deployed eight Rubrik R348S nodes with a total of about 80TB of storage on site, with flash and SAS spinning disk tiers of storage inside. Data is ingested, then copied off to the Azure cloud.
The key benefits for Poll’s team are the substantially better restore times, plus the ability to potentially restore virtual machines in the cloud, allowing staff to work from any location in the event of a disaster.
Rubrik’s CloudOn enables rapid recovery to allow for business continuity in the event of a disaster, said Poll. “If our on-prem site is down, we can quickly convert our archived VMs into cloud instances, and launch those apps on-demand in Azure,” he added.
“We don’t notice any difference in data ingest, but performance on restores is very much better.”
In cost terms, Cranfield had been spending £50,000 a year on off-site hosting. It now spends about £25,000 a year with Microsoft Azure.
Meanwhile, time spent managing backup is down from about half a day a week to five minutes a day.
In terms of physical space and equipment savings, Poll said the university had turned off 42U of storage and backup devices, of which backup servers and Data Domain comprised 24U.
“Overall, it has given us a simpler, faster and more reliable backup service,” he said. “It is more easily integrated with a department that is moving towards a DevOps model, and when it comes to data recovery, we are down to minutes rather than many hours.”
The storage and backup refresh – with the move towards hyper-converged infrastructure – forms part of a wider plan to rationalise IT by making use of contemporary devices’ formats with a smaller physical footprint, as well as the cloud.
Poll added: “The university masterplan is to knock down the IT department and to no longer have two large datacentres on site. Instead, there will be one datacentre, a ‘resiliency room’ for redundancy of network equipment, and the cloud.”