NEEDHAM, Mass. -- It's a common stumbling block in the disaster recovery (DR) planning process: finding a way to get funding for a project that, in the best case scenario, won't ever be put to use. IT pros
Speaking at a DR Roadshow event yesterday, DR expert John William Toigo, managing partner for Toigo Productions, told attendees that the best way to demonstrate DR's value to the business is not to pitch it as a DR plan.
Toigo cited the example of a case in state government in Illinois; an IT administrator was trying to get funding for a DR facility, but his applications were denied or stalled in committee. Finally, according to Toigo's anecdote, he pitched the DR hot site as a "software testing facility" and was granted funding.
One attendee, Brian Durand, director of operations for Silverlink Communications Inc., which manages call center functions for healthcare customers, said he planned to pitch his secondary data center as a quality assurance (QA) facility.
"We're a software company," he said. "We need a QA environment, and we're already staging it out of our production facility. Putting it into a secondary data center will free up our production environment."
Currently, Durand said, Silverlink uses local replication between two Clariion CX500 SANs from EMC Corp.
Toigo pooh-poohed the approach taken by EMC and other major vendors, cautioning users to avoid "vendor lock-in" by purchasing multiple boxes for DR, as well as proprietary replication software to go between them.
"Stay clear of 1:1 replication -- it gets prohibitively expensive," Toigo said. "But most vendors will tell you it's the only possible solution."
Durand said he'd conducted his request for proposal processes for the storage and the replication pieces separately and was using XOoft's WANSyncHA software, rather than an EMC product.
Durand said he had picked WANSyncHA because it's application aware, does both replication and failover, and can be used either locally or remotely.
"We'll see how it works over a WAN link," he said. "But the fact that it's supposed to work over long distances as well was part of our decision to use it."
Durand has another CX500 at the secondary site. "We can't have cheaper disk in a QA and testing environment -- then if something fails, you don't know if it's the software you're trying to test or the disk," he said.
Many of those in attendance, however, said they hadn't gotten to the stage Durand was at. Joshua Maltby, systems administrator for Extraprise, which does database hosting for clients in the Boston area, said he was looking to implement replication (currently, he uses tape) and a secondary site but hadn't put his plans into action just yet.
"I need to figure out a replication product or plan I could put into place in stages," he said. "There's no way we can just go from tape right to synchronous replication."
Another user from a major East Coast Fortune 500 company who asked not to be named said he realized that tape was far from adequate after a series of mergers made his data grow exponentially. The customer said he had recently been testing IBM's Global Mirror asynchronous replication.
"We're well on our way," he said. "But it's not in production yet."
Among Toigo's further tips: create a "data map" of your business as a prerequisite for DR, but also show management it can be used for regulatory compliance requirements and business planning.
"Let's say a business manager wants to acquire a company but isn't sure about what that would mean for your IT infrastructure," Toigo said. "If you already have the information on what's in your infrastructure and how much it costs, then you're his go-to guy."
Currently, Toigo said, only about 50% of companies have any sort of DR plan, and only half of those companies test the DR plan regularly. In part, Toigo said, it was because IT administrators were afraid of a "failing" result.
"Don't be afraid of test results," he urged. "DR tests are more useful for rehearsal than getting a checkmark for an auditor. In the 'smoke of battle,' having done DR testing will help employees behave rationally in the face of a great irrationality."
Further Toigo talking points: The future of DR
Toigo said the trend coming down the pike that would change the way DR is done is a process called copy-on-write (COW). "This is the major value of virtualization," Toigo said. "And it'll turn all these back-end arrays into JBODs."
Simply put, copy-on-write allows the same data to be written to two targets at the same time. "You can already do this on tape, but not all virtualization vendors support it. Why should they when they make so much more money making you do replication inside the array?"
There are already some applications that have this function, including Oracle, Toigo pointed out. Oracle's Parallel Database product logs on as a user into another database to keep the application acquiesced. "Right now, it slows both databases to a crawl," he said. "But as the technology develops, it'll eventually become the norm."
Eventually, Toigo predicted that IP multicasting would also become de rigueur for DR. The process puts a "tailgate" on a disk drive, which assigns it an IP address. Data is sent from servers through an IP switch, which can read and write to several disks simultaneously.
"You don't need an HBA [host bus adapter], you don't need a RAID controller, you don't need a FC [Fibre Channel] switch," he said. Toigo cited the popularity late last year of Netgear.'s Storage Center "miniSAN" device, marketed as a consumer product (See also, Netgear launches dirt-cheap IP SAN, Sept. 14.)
"They couldn't keep these things on the shelf in the Christmas season," Toigo said -- and, he claimed, plans are already afoot to bring IP multicasting to an enterprise scale.
"Very soon, a major technical university in the Boston area is going to announce a multipetabyte version of this kind of system," Toigo said. "This scares the hell out of the industry."