Disaster recovery: Ad hoc and expensive |
 |
By Jo Maitland, News Director
15 Jun 2006 | SearchStorage.com |
|
|
ORLANDO, Fla. – Just two users in a room of about 300 people attending Gartner Inc.'s storage show this week raised their hands when asked: who performs change control testing as part of their disaster recovery (DR) plan?
"You two get best in class," said Donna Scott, vice president and analyst at Gartner. "The rest of you, see me after class," she joked.
Most users in the crowd said they test their disaster recovery plans once a year. In an electronic poll of the audience, 44% said DR was still a project within their company, 30% said it was a process, and only 6% said it was integrated with business objectives.
Asked what their DR strategy is for their most critical applications, 20% said they perform production load sharing, 22% said they perform production failover and recovery to development and test equipment, 23% said they perform production failover and recovery to an outsourced facility, and 30% of the audience said they perform production failover and recovery to a standby facility. A quarter of the crowd did not have a DR plan for less critical applications.
Scott went through the list of architectures for disaster recovery, including application-level replication, database log-based replication, storage controller-based replication (synchronous vs. asynchronous), host-based file and volume replication, and outlined the pros and cons of each approach. She also mentioned some emerging technologies, such as network-based replication, from companies like Topio Inc., Kashya Inc. (now owned by EMC Corp.), FalconStor Software Inc., DataCore Software Corp., StoreAge Technologies and IBM's SAN Volume Controller.
The majority of the audience said they were using host-based file and volume replication. Scott said she was "surprised" by the number of people doing application-level DR. "You have to be very systematic in your approach with that method.
Most companies interviewed after the session seemed to be struggling in one way or another with DR.
A spokesman for a medical equipment supplier in Minneapolis, who requested anonymity, said his firm hadn't had that "burning bridge" to nudge them into implementing DR yet. "We're still at step one, which is to build a true enterprise-class data center, then we can get a second site," he said. The firm is currently moving out of an old facility that's at 95% utilization for power and floor space.
Harry Quoc Hoang, engineering manager for Helio Solutions, a storage reseller, said that not many people are asking for DR "because of budget issues." For a lot of companies he talks to, DR is still "a luxury" they can't afford, he said.
Media General Inc., owners of the Tampa Tribune and WFLA TV station, would, in an ideal world, outsource its disaster recovery operation to a co-location facility, according to Mark Holt, a spokesman for the company's IT Group. "I like the Google idea of unlimited storage that you rent at a co-lo facility," he said. Today Media General has multiple levels of DR depending on the application. For example, Exchange has its own application-level recovery on disk, which the firm is spending "a lot of money on," while file servers use tape backup, according to Holt. "The calculation is done by whether the application is valuable enough for us to spend money on," he said.
Proximus Belgacom Mobile, a telecom provider in Brussels, is also looking to save money on its DR plan. The company uses Hitachi Data Systems Inc. storage and TrueCopy software to replicate between two data centers. "From a storage management perspective, controller-based replication is easy," said Wim Vanhoof, manager of storage systems at Proximus. "But it's expensive, and where it's not needed, we're looking to get off it."
BAE Systems is another company trying to establish a DR plan but is finding that security issues are a sticking point thus far. In replicating from site to site "we have issues over who can see the data," said John Volanski, hardware systems manager at BAE. "With cross replication [between sites], there are multiple levels of security, and I don't think just zoning is the answer." BAE uses IBM DS4500 storage and is looking at SAN Volume Controller to replicate between sites.
And then there are folks simply trying to get it to work. The World Bank Group uses operating system-level mirroring on IBM AIX, which "doesn't work as well as it should," according to a storage expert at the bank. If the company reboots a switch, it breaks the operating system-level mirroring on AIX. "This is not a niche technology, we're not asking for something major, it should be bulletproof," the user said. Gartner's Scott agreed and offered to elevate the problem with IBM on behalf of the user. "Vendors don't like it when we analysts know about this stuff -- they tend to respond faster," she said.
Scott offered some recommendations at the end of the session:
Make DR management processes that are integrated with business and IT processes and that can adapt to changing business requirements.
Infuse a continuous improvement culture.
Plan for DR and availability requirements during the design phase of new products and annually reassess for production systems.
Test, test and test more.
Use automation to reduce complexity and errors during failover and failback.
Select replication methods that match your Recovery Point Objective (RPO) and Recovery Time Objective (RTO). If a single infrastructure based offering is desirable, consider storage controller-based replication. If 24/7 continuous availability is required, consider application, transaction or database-level replication.
|
|
|
|
