HP's Hybrid Approach to Disaster Recovery
When the need came to increase data storage capacity at United Water, officials wanted to make sure that all of their mission-critical data would continue to flow steadily. The company wasn't able to expand the life of its current systems, so it went looking for new models that would provide scalability.
The idea was to purchase a robust storage array for United Water's headquarters in Harrington Park, N.J., and a backup system at an offsite location. But cost was a factor. So after talking to a variety of vendors, including HP (NYSE: HPQ), IBM (NYSE: IBM) and EMC (NYSE: EMC), United Water opted for an innovative solution from HP called "DR Lite" that utilizes two different types of arrays, an HP StorageWorks XP and an HP StorageWorks EVA.
"Our preference was a high-end, high-performing server at our primary location, but we couldn't afford that in our secondary location," said Natraj Subramanian, IT Engineering Manager at United Water, the second largest water company in the U.S., serving seven million people in 26 states.
So in May 2008, he and other IT officials had the opportunity to talk with lab engineers at HP and decided to test the idea of a hybrid model. "We wanted proof of concept, and we were very comfortable and felt it was the right way to go," Subramanian said.
He said United Water is "one of the early pioneers" of this approach, noting that as recently as 2007, HP had told them that if they wanted to replicate data, they would need to purchase identical arrays in the same product family for their primary and secondary locations.
"But we didn't want to go in that direction," Subramanian said. "The EVA is a great system, but some of the critical applications we'd prefer to have on an XP because of performance and because of the scalability of the capacity, because it can grow up to a petabyte."
This type of model "is very unusual," observes Arun Taneja, founder and consulting analyst of The Taneja Group, a storage consultancy in Hopkinton, Mass. He said the large majority of replication has historically been done between like arrays at the primary and secondary locations.
Taneja said that by far, the number one request he hears from clients when it comes to replication is that they want to have the ability to purchase different storage systems, but most vendors require like products at both ends.
"The number one reason for that, which I completely agree with, is that the customer does not mind buying the expensive storage for the primary site, which is where the [mission-critical] application is running," Taneja said. "But the customer tells me that for the disaster recovery site, they don't necessarily want the same high-end storage."
Companies would rather purchase "lesser" storage systems for DR sites because they want to make the combination more cost-effective.
"They do not mind taking the risk of lesser performance if a disaster were to strike the primary site and things had to flip over to the recovery site,'' he said.
Big Cost Savings
In February, United Water began implementing a high-end, HP XP 2400 with 75 terabytes in Harrington Park, and for disaster recovery and backup of critical data, an EVA 6100 with 37 terabytes located about 10 miles away in Hackensack.
Two XP arrays would have run the company $2.2 million to $2.5 million. Instead, the hybrid model cost United Water $1.26 million.
Subramanian said getting two disparate arrays to talk to one another was a challenge. The high-end system has different programming and micro code in the software.
"That means going from one type of server to another creates a challenge communicating to another family of storage, and that was the initial problem," he said. "It will not understand the other unit. It won't do a technical handshake and it will see it as an alien."
In order to get the two heterogeneous systems to replicate, United Water is using HP's Business Copy software and the XP's external storage technology, Subramanian said.
United Water is still in the testing stages with the arrays. The initial formatting and configuration of both units is complete, but they haven't been connected. The company will soon migrate some of its data from the old system to the XP array. The second phase will be to move the rest of the other critical applications to the primary system by the end of June. Then officials will start replicating the data to the secondary location for completion by the end of the summer.
"It's a pretty interesting process," said Subramanian. "Everything is working the way we expected it to, and there are no kinds of issues other than the time and logistics of having everything ready."
The financial aspect was the main impetus for this approach, he said. Since United Water now uses HP exclusively, "everything is connected together. If there's a problem with a particular operating system, there's no finger pointing and they cannot get away from me."
EMC Gets into the Act
Nowcom, a maker of software for car dealerships, has experienced widespread growth in the past few years as more dealers use the company to store and manage their financial information. As a result, Nowcom's former recovery solution on a SQL Server 2005, which stores inventories and other dealer information, was no longer able to keep up with the growth.
Based in Los Angeles, Nowcom needed recovery software that would protect the business and its customers from the risk of data center failure in the event of an earthquake or other regional factors. Nowcom estimated that its customers would lose between $50,000 and $100,000 in combined sales if its database was offline for an hour.
The software maker is now running EMC's RecoverPoint, a network-based data protection and remote replication application, also in a hybrid model, from an EMC Symmetrix DMX-3 SAN in the primary center to an offsite recovery location 20 miles away, where it resides on a Clariion CX-Series SAN.
Taneja believes that within five years, the majority of disaster recovery sites will replicate from dissimilar storage systems.
"All companies are under pressure to control costs, and there is no reason for them to have expensive storage on their disaster recovery site because the majority of them can justify that if and when a disaster happens, they're willing to take a performance hit because the cost is huge," said Taneja. "With these dissimilar storage servers, you could save as much as $20 million, so we're talking real money here."
Follow Enterprise Storage Forum on Twitter