Completing advanced disaster
This means his staff is managing a SAN manually, the better to expire snapshots on the secondary system to make room for new data. "That has been annoying, but we're struggling to get additional disk," said Owens, who's also trying to get a Hewlett-Packard Co. (HP) MSA 2000 approved for tier 2 data. "But even trying to slip low-end storage in was a no-go."
Seeking disaster recovery in the clouds
Cloud services are an alternative to expanding disaster recovery on the cheap. Jay Vlavianos, director of production operations and IT for Ecast, a company that maintains digital jukeboxes and touchscreen kiosks at nightclubs and sports stadiums, had originally deployed Amazon's S3 service to serve up music and other content.
To improve response times, Ecast switched to storing some 10 TB of persistent data in Amazon's cloud rather than expanding its internally managed secondary disaster recovery site. Ecast still has some data at a colocation site in Chicago, but its primary data center in California has about three times as much data. Bringing the two sites into sync would have been more expensive than using Amazon.
"The service is still relatively immature, but I've never seen anything so cheap," Vlavianos said of S3, which he estimates saved the company $130,000 in additional operational and equipment costs it would've taken to expand the disaster recovery site. "The most important thing is the operational costs. At a DR site if I had a problem, I'd have to handle it remotely and open a ticket with a vendor. If Amazon has a problem, they handle that part on their end and I don't notice it."
Some organizations have been able to get creative and repurpose existing resources for disaster recovery purposes.
Mike Coppach, voice and network engineer for Mister Money, a Ft. Collins, Colo.-based company with subsidiaries including pawnshops, payday loan guarantee and collection agencies, said he realized he needed networked storage to offer clustering failover for his SQL database server. He went looking for iSCSI storage in a Web search in February, and decided on StarWind. "They were the only one with a free demo available on their website," Coppach said.
Best practices for mirroring SQL servers dictate the use of synchronous replication between SQL databases to ensure transaction consistency. That's an expensive proposition. So Coppach created what he calls a "poor man's shared-nothing SQL server cluster" by attaching servers running the StarWind iSCSI SAN software with Double-Take host-based asynchronous data replication.
The setup leaves Mister Money without automatic failover but Coppach said that's acceptable. "Ten to 15 minutes of downtime is tolerable to set up the failover," he said. "Beyond that, the costs would be significant." Now Coppach is looking to do something similar for his Exchange server.
EvensonBest turns to VMware instead of a SAN
Instead of a SAN, furniture distributor EvensonBest virtualized its servers with VMware Inc. software and attached them to IBM Corp. servers running Windows Storage Server software. "We already have service contracts on IBM servers and we can put spare hardware on the shelf," director of IT Martin Silverman said. Instead of SAN-based replication, the servers send data to a remote location using CA Inc.'s XOsoft Replication.
While mirrored SANs would be ideal, Silverman said times call for more frugal measures, likening a SAN purchase for DR to "buying a new car with air conditioning – it might be more comfortable, but if I can't make the car payments, it doesn't matter."