The state of the enterprise cloud and prepping for AWS re:Invent 2013
A comprehensive collection of articles, videos and more, hand-picked by our editors
Cloud disaster recovery and backup options have become more common, and some users say they provide a higher level of protection than traditional solutions -- at lower cost.
Disaster recovery (DR) means different things to different people. One IT pro's definition of DR might be simple file backup, while another might be referring to full standby server farms ready to take over production duties at a moment's notice.
At its most basic level, disaster recovery means storing backup data off-site, which increasingly means the public cloud.
For Prellwitz Chilinski Associates, Inc. (PCA), an architecture firm in Cambridge, Mass., nightly backups are the cornerstone of its disaster recovery plan. The data being backed up are large files architects generate using tools from vendors such as Autodesk Inc. and Adobe Systems Inc.
For many years, PCA replicated its data between a pair of storage appliances provided by a local integrator -- one on-site, the other off-site.
"It was a great solution -- we had the local backups for quick access, plus the safety net of off-site backup," said Dan Carp, systems administrator for the firm.
Then, about two years ago, Carp began exploring whether cloud backup could be more cost-effective. Using the replicated data appliances, PCA paid about $14,000 annually to protect 700 GB of data. After a few fits and starts with various cloud data protection providers, it settled on Zetta.net, which stores almost twice the data it had on the data appliances, for about half as much money.
"I like to think of this as a life insurance policy -- I hope I never to have to use it," Carp said. But if that day comes to pass, users will access their files by copying them back to their systems, or by accessing them over the WAN at the Zetta.net data center.
Cloud backup for faster recovery times
Cloud backup is relatively simple, but recovering entire systems from the cloud leaves a lot to be desired, said Rick Vanover, product strategy specialist at backup and replication vendor Veeam Software.
"Transferring a 200 GB file back down to your data center is not fun -- especially since you probably need it right now," he said.
Veeam Backup Cloud Edition software already supports backup to 19 different cloud object stores, including Amazon Simple Storage Service (S3), Amazon Glacier and Windows Azure, and Vanover said it is an increasingly popular backup target. However, upcoming versions of its software will also support a feature called Backup copy job with WAN acceleration, which allows administrators to pre-allocate known-good backup images on public cloud compute infrastructure, he said, for faster recovery times.
I like to think of this as a life insurance policy -- I hope I never to have to use it.
Dan Carp, systems administrator at PCA
"You've effectively already put your backup infrastructure in the cloud," he said.
Indeed, backing up systems to the cloud is only half the battle, users say.
"Okay, so you've got your backup replicated to an offsite storage facility. Now what?" said Bryan Bond, a senior systems administrator with eMeter, a software development firm in Foster City, Calif.
The company recently began using the Recovery as a Service option from Bluelock, a cloud services provider based in Indianapolis. The service integrates replication software from Zerto Inc. to minimize the bandwidth and time necessary to copy virtual machine (VM) replicas to its environment. If and when disaster strikes, it spins those VMs up on the BlueLock Infrastructure as a Service (IaaS) simply by switching its DNS entries, Bond said. And because Bluelock is based on VMware Inc.'s vCloud, it can do so without having to convert its VMware-based VM images.
Bond also explored "pilot light DR," in which a minimal version of an environment is left running in the cloud, ready to scale up and take over in the event of a disaster, but he ultimately decided that was overkill.
"As it is, the time it takes for us to return to operations is so short," he said. "Pilot light would have required us to re-architect our data to be aware of a cluster partner," using fancy networking techniques such as the Border Gateway Reservation Protocol (BGRP). "It's much easier to make a DNS change than it is to make the network complicated."
Weighing cloud DR options
IT professionals exploring cloud disaster recovery options have their work cut out for them, said John Morency, an analyst with Gartner Inc., a research firm based in Stamford, Conn.
"When I first began tracking this space five years ago, there were about five or six providers," he said. At last count, he had identified over 120 in the U.S. and offshore.
The viability of cloud-based DR services will depend greatly on the level of heterogeneity and support you need, he added.
Compared with traditional DR services from the likes of IBM and SunGard Availability Services, cloud-based DR services give companies "the freedom to test as frequently as you want," Morency said. But for large enterprises with non-virtualized systems, "most cloud providers will not be able to support those configurations."
About the Author:
Alex Barrett is editor in chief of Modern Infrastructure. Write to her at firstname.lastname@example.org.