InformationWeek.com

Welcome Guest. | Log In| Register | Membership Benefits

October 18, 1999

Recovery Strategies Take Center Stage
Storage management tools are key to better data-restoration policies

By Oliver Rist

Related links:

Site Security Gets Tougher

And from our sister publications:

InternetWeek Lawson Taps Imation To Back Up Data Online

Computer Reseller News Building a Backbone
n today's business climate, even a few hours of downtime can cost companies millions of dollars. That's why it has become critical to have a fast, simple, and bullet-proof business-recovery strategy in place.

In Canada, proceeds from the lottery--a healthy $1.2 billion last year-- are used to subsidize health care and other welfare programs at both the state and national level. John Sidwell, a technology consultant to the Ontario Lottery Corp., says there's no room for risk-taking. "We're in the business of selling tickets," he says. "Our customers need to be able to play no matter what happens to our data center, and our responsibilities to Ontario and the other territories are just as important."

Sidwell's network is based on Compaq OpenVMS servers, which manage the ticket validation process for every lottery machine in Ontario. The data are fed into the company's validation database at the primary data center and then immediately mirrored via a 6-Mbps asynchronous transfer mode connection to a secondary data center 500 miles away. "We need to shadow our validation database in real time," says Sidwell, which puts his business-recovery needs at 0% downtime. That's a harsh reality, but companies are getting some relief as business-recovery solutions mature.

"In the past we had to use VMS remote clustering," Sidwell says. "That means we were effectively clustering servers in real time over a 10-mile distance to our nearest data center." By using a new recovery product, called RemoteShadow from Advanced Systems Concepts Inc., Sidwell has been able to do away with the complexity of remote clustering and has even been able to save money by closing one of three data centers--all while keeping recovery operations for the validation database nearly instantaneous. "Recovery and failover is mission-critical to us, and this represented the best solution available at the moment."

Failures often can bring down the entire business--especially for E-commerce ventures--so more network managers are considering an interim recovery layer closely related to their storage-management schemes. Even E-mail can cripple companies if it suddenly goes down and plans for recovering the message database aren't solid.

Often, customers can't rely on their messaging software provider to include robust recovery features in its base package. For example, says Glenn Dekhayser, director of technical services at Clearview Technologies, a systems integrator in Whippany, N.J., "Many folks have complained about Microsoft Exchange's inability to intelligently manage recovery of a corrupt message store database. When it happens, it's a nightmare."

Dekhayser's customers are demanding more reliable E-mail servicing, so ClearView is offering Exchange-optimized solutions based on Network Appliance's NetApp Filer products. These are network-attached storage products that use a proprietary filing system to manage data from multiple operating systems at high speed. For recovery applications, Dekhayser is impressed with NetApp Filer's management function because "it allows you to take a snapshot of the message store database and then backs up only altered blocks," he says. "That's good not only for performance but for recovery, because now you can retrieve only what you really need, instead of the whole database."

To hear Dekhayser put it, this problem can reach epic proportions in enterprise-sized networks. "Take a company like GE, which has large numbers of Exchange message stores distributed all over the country," he says. "If that died, that would be hundreds of gigabytes of data and would take days to restore entirely from linear tape."

This puts into glaring relief a serious problem many IT managers are only now beginning to face: how to consolidate their storage-management and business-recovery schemes with their companies' ever-growing appetites for more storage, driven by the ever-larger data requirements of messaging, E-commerce, and other apps. In a recent report, Forrester Research estimated that most enterprises will see upwards of 50% growth in storage in the next year, and that IT spending on storage will increase from 4% of IT budgets to more than 15% by 2003.

IBM's Tivoli division thinks it has a way to help companies deal with this issue. The next generation of IBM's established Adstar Distributed Storage Management software, called Tivoli Storage Manager, is being billed as an A-to-Z business-recovery and storage-management tool. Storage Manager will let users centrally manage data as well as spot storage failures anywhere in the enterprise on everything from a SCSI disk array to tape drives or even a multivendor storage area network, says Troy Pladsen, director of strategy and business development at Tivoli. As a bonus, the company has built in code to help tape products increase performance in recovery situations.

bar chart Pladsen says the software already has support for 340 hardware storage devices from all kinds of vendors. "And we're leveraging that to allow distributed access to all your hardware, starting with tape drives. Storage Manager will sit on top of almost any tape device driver, regardless of operating system platform, and allow you to distribute data across all of the tape drives. We're also ready to share tape drives in SAN configurations." By spring, it will add the Tivoli Removable Media Manager, which uses the Windows 2000 removable storage management API to take over all communication with devices that's normally handled by the application being backed up.

If it works, Storage Manager would solve a significant hurdle in the sharing of tape libraries. What has stopped sharing in the past is that these libraries use robotic arms that are essentially SCSI devices requiring attachment to a single host. Software that lets several platforms arbitrate mount requests from several hosts will allow better use of storage hardware and tape space, especially in SANs.

continued...page 2