Refining storage backups

When petroleum giants Ultramar Diamond Shamrock and Valero Energy finalized their megamerger in December, the task of protecting two gargantuan data storage architectures became urgent.

Storage specialists at the merged company, the new Valero, knew they needed to craft a plan for a single, integrated back-up and restore system. The goal is to ensure data availability yet lower administrative costs.

Such integration will be no easy task. The data involved in operating the third-largest U.S. petroleum refining and marketing firm, a Fortune 50 company with annual revenue approaching $30 billion, is staggering.

The new Valero pumps 2 million barrels per day out of refineries and supplies gasoline to nearly 5,000 service stations nationwide. In so doing, it generates terabytes of data, which reside on hundreds of NetWare, Unix and Windows servers.

Advertisement:

Integrating backup is the No. 1 storage management priority for the merged storage team, says James Brents, technical specialist at Valero in San Antonio, Texas. "Our goal is to be able to restore the data first and foremost. Secondly, we have to continue meeting our back-up window as our data grows," he says.

The new Valero backs up an average of 12 terabytes of data nightly across a LAN or storage-area network (SAN). This is storage from 35 AIX and 40 other Unix boxes (running HP-UX or Solaris), 32 NetWare servers and 350 Windows NT or 2000 servers. The largest databases, those in the 100-plus gigabyte range, are backed up over a SAN.

Variety at Valero By using NAS and SAN technologies, the company avoids reliance on a single vendor. While the new Valero considers whether to standardize on a back-up solution, its overall storage environment will remain as diversified as ever Click here for more

Integrating the previously disparate back-up efforts promises easier maintenance and reduced costs. "We will be able to focus our administrators, and they will be able to share training and experiences," Brents says.

The new Valero currently uses IBM/Tivoli Storage Manager (TSM) and Hewlett-Packard's Storage Omniback for backups. TSM handles most backups at the Valero North campus and the NT backups at the Valero corporate offices, where the Omniback handles Unix backups.

"Part of our challenge is evaluating the technologies used by the two companies previously and determining whether it makes sense to integrate and how to do that, or whether it makes sense to standardize on one," says Mark Calkins, senior IT manager at the new Valero. "We are looking for synergies."

Brents won't disclose the cost of the back-up integration project. But he stresses the need for sound data retention policies for backups and archives when it comes to managing storage volumes as large as the new Valero's. His advice is to include back-up, archive and disaster-recovery costs into initial storage plans. In so doing, he says, "don't believe the users when they say they'll only need X amount of space. They always need two times, three times and more."

The storage team says the integration project should be complete in September.

Speed freaks

To meet its second goal of speeding back-up time, the team is investigating LAN-free backups, in which Valero's TSM clients back up data directly to tape drives on the company's SAN. Control information, such as storage policies and metadata, would continue flowing over the LAN to the TSM server.

Transferring data over the SAN rather than the LAN "will help avoid network bottlenecks and take a load off the LAN]," says Pablo Clifton, technical specialist at the new Valero. Today, the new Valero backs up its 1-terabyte SAP ERP database at a speed of 400G bytes per hour. In a recent test, Valero backed up that database in two-and-a-half hours. While the storage team isn't unhappy with that rate, it believes it can improve that time by at least 200G bytes per hour with LAN-free mode, Clifton says.

In Valero's pipeline The storage team plans to tackle these projects next: Disaster recovery The new Valero's storage team is conducting a business-impact analysis that it will use to craft a comprehensive disaster-recovery plan for the merged company. The plan will specify back-up and restore procedures on an application-by-application basis. Valero might back up certain applications to the SAN, while it moves others to the LAN. Click here for more

While Valero seeks out new back-up technologies, it also presses to achieve backups at the published speed of its tape back-up devices. Brents is particularly proud of that practice. "Reducing the back-up window is always the goal, but [you have to] keep in mind that a tape library also needs time to perform the daily administration processes, such as reclamation, off-site copies, expiration and the TSM database backup," he says. "The more time available for those processes to complete daily, the quicker the restores."

Valero tests its back-up and restore procedures often. It conducts major back-up tests four times per year, with project tests multiple times per month and hot data backups six times per week, Brents says.

Testing is a challenge, but absolutely necessary, he adds. "It's really tough to have a duplicate test environment when you have such a large operation."

Remember, Brents says, "you must manage your data, and not let your data manage you."

Contact Signature Series Editor Beth Schultz

Technology Insider: Storage virtualization
Virtualization is the buzzword in storage these days, but what does it really mean and how do you achieve it? In this report, we nail down what virtualization is and describe how the major vendors hope to get you there.
Network World, 03/11/02.

Storage newsletter
Get the latest news, information and analysis on storage and application management so you can make the best decisions.

SAN audio primer
In this 5-minute primer, you'll learn how a SAN works and its benefits over traditional storage systems.