A company doesn't have to be susceptible to cataclysmic disasters such as hurricanes, earthquakes, floods or terrorist...
attacks to require a good disaster recovery plan. For Salt Lake City, UT-based currency exchange firm Interbank FX, any outage of a mission-critical trade server could cost thousands of dollars a minute.
Interbank FX IT infrastructure manager Paxton Powers points out his company's trading volume is over $40 billion a month, which comes out to about $4 billion each business day. "An outage to our trading server would cost $100,000 an hour," he said.
So last year Interbank set up a primary data center in New York to replace a collocation center it was using there. Now it replicates data from New York to a disaster recovery site in Salt Lake City, as well as replicating between headquarters and the Salt Lake data center.
To help cover costs of a new data center, Interbank FX revamped its data storage setup. Powers says his previous environment was overkill, with an EMC Symmetrix DX-4 serving as the main system. The firm swapped out the Symmetrix for Hitachi Data Systems (HDS) AMS 2300 and 2500 midrange boxes for the two data centers.
"Symmetrix is a great box, but it's super expensive to maintain," Powers said. "It cost $50,000 a year for power. The AMS gives us the performance we need, and gives us space."
The next step was to find the best method of remote replication for his data. Powers said he first looked at SAN-based replication, such as EMC RecoverPoint or HDS TrueCopy, but didn't like their pricing structure.
"They charge per terabyte," he said. "Our data storage requirements keep going up. We're creating a lot of data and we have to keep it. The costs to replicate would keep going up."
HDS also resells InMage Systems Scout server-based replication software for DR, and Powers took a look at that. "With InMage, you load an agent on 10 servers and protect all the data on those servers," he said.
Interbank FX first used Scout to replicate data off the DMX-4 to the AMS boxes, and kept it to replicate between data centers. "We replicate between data servers, protecting our trade application server first and then migrating between production servers," Powers said.
He says Scout is used to protect about 5 TB to 7 TB of data. Interbank FX hasn't had any major outages, Powers says, but has run test cases without any production loss. The company has a recovery point objective (RPO) of less than two minutes for its trade server.
"It's the fastest way we have of recovering data, faster than restoring servers," he said. "We shut a server down, bring up another and we're trading again."
Powers says he's considering speeding the process by using solid state for the server where Scout captures real-time changes in data, called the CX. "When it's doing compression, you really work that CX box," Powers said. "Maybe I'll put a Fusion-io card in that box."