Melbourne IT turns to data deduplication to backup virtual machines

Hosting provider Melbourne IT has achieved 25:1 deduplication ratios using Data Domain to backup virtual machines, but hopes those ratios will fall over time as it learns more about how dedupe works.

Listed hosting and domain name company Melbourne IT has implemented EMC’s Data Domain deduplication appliances and achieved deduplication ratios of 25:1, but hopes less impressive ratios are in its future.

“In our retail hosting business we create a lot of virtual servers,” says Guye Engel, the company’s Group CIO. Engel’s not kidding: the company has more than 350,000 clients, generally provisioned with up to 50GB of disk space which may or may not be used.Many of those clients use virtual machines, and their hosted systems are therefore quite homogenous, leading Melbourne IT to protect many copies of what is essentially the same data.

“Our previous backup technology meant we would backup the same thing over and over,” he says.

Growth also created a challenge, because new customers mean the company has to add more storage capacity. “As that primary storage grows you have multi copies on backup,” Engel says. “That creates challenges to scale storage and keep backups in synch.”

Already an EMC customer – CLARiiON CX4 machines are the heart of the company’s storage infrastructure – Engel wondered if data deduplication could help it to control the amount of data it was backing up.

Two Data Domain model DD8800 appliances later – the pair arrived in January 2010 – Engel is happy to report deduplication ratios of 25:1 “for month-end backups.”

“The DD8800 takes the copy and takes out the blank non-used space and that works really well. It also does well with things like duplicated operating systems.”

Not all data, however, has achieved 25:1 compression. Some has reached just the lesser ratio of 6:1, but Engel puts the wide variance down to the learning curve the company is on as it develops a more refined understanding of how to make deduplication work.

“The plan is to get the ratio at 15:1 or higher,” he says. “That will be a key ingredient to helping us scale and will help us to create a cost base the business can afford.”

In the meanwhile, the company is enjoying other benefits.

“We are getting faster recovery times,” Engel says. It’s down to minutes thanks to online copies. Coming back from tape takes more time.”

Tapes also used to consume a lot of IT time, but this has been greatly reduced now that backup goes to the Data Domain appliance.

“We used to have 600 tapes a month and think we have saved seven man days per month on tape handling,” Engel says. The company is also taking advantage of replication to copy data between DD8800s in different states, to further improve its resilience and boost disaster recovery capabilities.

“It also occupies less racks so we use less power,” Engel says. “The backup component [of our power consumption] is down about half. Any drop in power consumption is good because it means we can bring on more new business in the data centre.”

Read more on Data quality management and governance