Case study: Offis overhauls storage to boost IaaS cloud service

The cloud is rightly being recognised as a way to reduce costs, boost IT uptime, and cut out low-value technology and business processes. It's also becoming increasingly recognised as a way to dump ugly IT problems onto someone else.

That's all great when you're the IT department doing the dumping. But what happens when you're the one whose job it is to manage someone else's mess?

Just how do you maintain real-time disaster recovery while keeping storage costs low, all while managing 30 percent year-on-year data growth volumes and maintaining near 100 percent uptime during peak demand periods?

That's just the problem that Australian infrastructure-as-a-service (IaaS) and cloud operations provider Offis faced.

According to the company's executive director and founder, Craig Allen, the company was battling with RAID 5-based storage infrastructure, and needed to move to a mid-tier SAN-based storage environment to better allow the company to keep ahead of customers' cloud infrastructure needs.

"Prior to using SAN technologies, Offis was using dedicated IBM System X rack and blade servers. These servers were generally set up dedicated with RAID 1 and RAID 5 SCSI Ultra- 320 disk technology," he said.

"It was very hard to increase the size of disk partitions without disruption, and replace the existing drives... Disk failures would also impact the customer experience — when a disk needs replacing in a RAID 5 configuration, the performance can really be degraded."

The answer, Allen said, was to look for a storage solution that could meet three core requirements: Enabling a disaster recovery approach, which could restore services immediately and automatically if primary systems fail; maintaining a real-time mirror of data to ensure full service continuity during any outage; and minimising clients' data footprints to avoid blowouts to hardware costs.

The solution

Being an IBM-house and ex-IBM staffer, Offis' Allen said it meant that turning to Big Blue's IBM V7000 storage infrastructure was the answer. But not before testing out the market first.

Allen said the foremost consideration when looking for storage infrastructure wasn't big brand name recognition, but reliability.

"The reality is that customers don't care what sort of hardware it is running on — whether it is IBM, HP, Apple, or anything — the only thing they care about is that when they move their mouse and click on a button, something happens.

"That's all they should care about... These days, the CIO doesn't care what equipment [applications] are running on. It is no longer part of the argument. It is all about the service, the operations and HA [high availability]."

The results

According to Allen, Offis has benefited in its move to the V7000 platform in a number of ways.

Most notably, Allen said that the performance of Offis' storage has also received a major boost — at least four times the 4000 to 7000 IOPS (input/output operations per second) that the company was getting on its previous storage infrastructure. This in turn has led to significant savings on licensing costs for the company.

"Performance went through the roof," Allen said. "When you start experiencing 30,000 IOPs with a very low latency on a machine, it means the CPUs are doing less waiting for the data. The less wait time you have for the data, the more time you have to compute. The more compute time you have, the fewer licences you need on the system."

In experiencing 30 percent year-on-year data growth, the ability to access a real-time compression capability has meant that the company has been able to reduce the rate at which it needs to purchase additional storage as well as further licences.

"We are experiencing 40 to 50 percent compression, with only a 2 percent [system performance] overhead," Allen said of the compression tool.

"The reason we use real-time compression is that it is a cost benefit to us. The way licencing is done on SANs is by enclosures, and if I can reduce the number of enclosures I use through real-time compression, I can reduce my licensing costs."

In addition to compression, upgrading the V7000 to add more storage capacity is also significantly easier than on Offis' old storage infrastructure, Allen said.

"In a dedicated environment, once you have filled up a RAID 5 environment, it is a real tricky devil to move up to a bigger disk size," he said. "But in a SAN environment, you can do it non-disruptively and add storage to the customer's environment."

Allen said that in addition to benefiting from the V7000's intuitive graphical user interface (GUI), the company also benefited from the V7000's ability to take snapshots:

"The biggest gain is the ability to do snapshots," he said. "In the past, if there was a problem with a change or upgrade, then the system had to be restored from tape. Now it takes a couple of minutes to revert back to the original version.

"In the past, if you were applying patches or upgrading or doing anything to the box, it was pretty much do a big backup the night before, then cross your fingers.

"Now, you can snapshot it, get it all done, test it, are you happy with it? Then remove the snapshot. We can do all that in 15 minutes, where before, it was a 12- to 24-hour process depending on when the customer wanted to do it."

Allen said the V7000 has also improved disaster recovery through gaining a WAN replication capability.

"WAN replication means we can replicate the data to a different geographical location," he said. "While this was possible with the old gear, it was extremely complex and expensive to set up remote mirrors. The old storage also contained single points of failure — particularly with the RAID cards."

In being a long-standing IBM customer, Offis also benefits in the support it received.

"I would say that with any organisation, when you're a one-house shop, you build relationships with them ... understanding how the processes and the systems work, and once you have those relationships, they are worth a lot of value to our customers," he said.

"We know we can escalate issues extremely quickly. We know who to go and ask for solutions, and when the proverbial hits the fan, we know who to talk to, rather than calling up a 1300 number and getting a ticket number and waiting for a response."

By way of advice for organisations with similar storage challenges, Allen stresses the need to look for a solution with high IOPS, but also one with low latency.

"The car may be able to go at 200 miles an hour, but if it takes an hour to get up to that speed, then it is useless," he said.