GlusterFS Case Study: RIS Belgium

Sometimes the solutions we put in place turn out even better than what we originally hoped.

That could sum of the experience of Belgian Internet Service Provider RIS Belgium,which turned to Gluster to solve the problem of distributed storage and ended up getting more benefit from the solution than they expected.

Initially RIS, a web site and infrastructure provider for customers in Belgium and the EU, was looking for a distributed storage solution that would be more economical and would be able to replaced more easily within RIS’ production environment.

“We wanted a cheap solution that could be replaced easily,” explained systems administrator Kenneth Dejonghe. “”f one server breaks down, we can drive to the datacenter, take the broken one out, and replace it with new hardware while we bring the broken one back to fix the issue.”

Dejonghe had a vested interest in this particular project; it was his first project assigned to him when he started with the 15-year old organization in Dec. 2014.

According to Dejonghe, RIS also needed a system that would provide them with good data distribution over the servers, which could be backed up easily and also easy to replace a node. It was at this point Dejonghe and his colleagues discovered Gluster.

Gluster was attractive to RIS because they run a cloud storage service similar to Dropbox and Google Drive for their customers and need to ensure customer data was getting backed up. Gluster not only would provide this type of distributed flexibility, but it also will enable RIS to scale as needed to meet customer needs.

Right now, RIS is using Gluster atop CentOS 6.6 on three nodes, with two volumes on each node, with two disks for the cloud service customers and a four disks for their own backups. Their Gluster instance is currently managing 72 Tb of raw storage, Dejonghe added.

Dejonghe was impressed by the easy of deployment of Gluster. “I think I set up the whole cluster in about a day,” he said. Dejonghe had the set up running on his test bench for just two weeks before he was satisfied that Gluster could handle their production loads. The only glitch Dejonghe experienced was a known issue with CentOS, where RIS’ RealTech NICs could not handle data distribution properly. This necessitated a shift to Intel-based NICs.

Gluster’s capabilities in data distribution over nodes, replication, load balancing, and self-healing has served RIS well so far. And, the change brought an unexpected surprise: RIS was able to use hardware that was more commodity based–hardware that used a lot less power than their previous systems. This would reduce the hit on RIS’ power bill by 20 percent every month.

Moving forward, RIS is examining using Gluster for slow storage. They are also looking forward to Gluster 3.7’s volume tier feature for volume sharing.

Dejonghe is more than satisfied with his Gluster experience, and it has put RIS in a much strong position to serve its customers and scale to their needs.