Ever since information lifecycle management (ILM) became all the rage, storage administrators have yearned to apply better data management policies to their storage architectures. But easier said than done. "Infastructures may have started with a data-centric business, process-focused view, but it rarely stayed that way over the years," said Jon Toigo of consultancy Toigo Partners International during his presentation Thursday at Storage Decisions.
Progress toward a "purpose-built IT infrastructure" will also be gradual. "Rip-and-replace is unrealistic in the current economy," Toigo said. Instead, he advised storage administrators to take advantage of a natural storage replacement cycle to phase in more storage products designed to support specific applications.
"It's about what we're legally obligated to do," said Maxwell Garrison, a systems management specialist for public schools in Edmonton, Alberta. Garrison is setting up a disaster recovery plan using the replication functionality built into his Dell EqualLogic iSCSI SAN. "We have to pay for services and pay our own staff, and student records are our business." That means the main applications for his shop are those that handle financial management, payroll and student records. Email has become more critical recently, but he said it's more important to protect the main databases.
Once those determinations are made, it becomes easier to match them up with the right infrastructure, said consultant Bill Peldzus, GlassHouse Technologies, during a presentation on setting service level agreements for disaster recovery. He said that customers with application consistency groups -- sets of interdependent applications that must be kept in sync -- have favored array-based replication because that keeps data from multiple applications together within the SAN. In contrast, host- and appliance-based methods send separate streams of information relating to specific hosts or workflows.
Garrison's organization picked EqualLogic because "management of direct attached storage became too complex," he said. "Management tools are much better in the SAN world, and with iSCSI SANs, more affordable."
A senior storage specialist for a financial institution in Toronto, who requested anonymity, said that his company is tuning its replication methods to set a standardized, asynchronous approach. He added that application-level changes, such as snapshot frequency, replication type and parallelization between primary and secondary data centers, would bring the business processes into line with the standardized disaster recovery plan.
At the Ontario Lottery, the most critical applications are ticket sale systems and a lottery ticket database. Lower priorities include support for 22 gaming sites, and financial and HR records, while email "isn't very critical for us at all," said SAN analyst Kevin Lin Jian.
The Ontario Lottery has also identified FCoIP as the technology it wants to deploy to meet more aggressive SLAs for the critical applications. "If we lose one site, critical applications have to be back up in 15 minutes," Jian said. The company has Hewlett-Packard EVA and IBM DS8000 Fibre Channel SANs in data centers 800 km apart. "And you need IP for distance," he added.