SAN survives earthquake without data loss

Posted on August 01, 2001

RssImageAltText

SAN benefits include reduced administration and better backup, data sharing, and data protection.

BY BILL REICHEL

A storage area network (SAN) played a key role in helping the Washington State Department of Natural Resources (DNR) survive the Seattle earthquake earlier this year, without losing a single byte of data. The DNR is replacing most of its server-attached storage with a "SAN-in-a-box" because it simplifies data management by combining the capacity and performance of individual disk drives into a pool of shared storage.

The DNR manages more than five million acres of land and has 1,400 employees in 200 communities across the state of Washington.

The department is currently in the process of transitioning from its primary information systems of 13 file servers, each with its own local storage, running NetWare 4.11. These servers previously handled all of the data required to operate the DNR. The primary problem with this approach, however, was that the servers created a substantial maintenance workload. Each server and its associated storage had to be managed separately, creating a headache for our IT staff. Data could not be easily shared, moved, or protected, making it difficult to perform some types of analysis and there fore, making things more difficult for users. In addition, backup was a complicated process that generated a considerable amount of network traffic. Expanding storage was expensive because in most cases it was necessary to purchase another server to go along with the additional storage.

I was part of the management team that made the decision to consolidate to a new configuration consisting of six servers with all storage residing on a SAN that could be accessed by any of the servers. SANs place the storage on a high-speed Fibre Channel network, making it possible to centralize storage and connect multiple servers to the storage pool.

Why a SAN?

The primary reason for making this change was the desire to reduce the management workload and increase disaster protection. With a SAN, storage resides in a single location, making it much easier to move, share, and protect it from disasters.

For the most part, backup traffic is offloaded from the data network and does not impact network performance. We use Veritas software to back up through the individual servers, improving network performance because backup traffic is confined to a single IP subnet. All but two servers are located in a common computer room on the same IP subnet. Two of the servers slated to be replaced by the SAN are located in electrical closets on separate IP subnets. Storage can be readily increased at any time without the expense of adding servers.

Selecting the SAN

One of the most critical decisions was selecting the SAN. One important requirement was that the array(s) support both the NetWare and Windows NT operating systems since the DNR had long-term plans to transition to NT. A key reason for selecting Xiotech's Magnitude SAN was that it could support both operating systems in a single storage array, while many competitive platforms required separate arrays for each operating system.

The DNR installed six new 733MHz servers running NetWare version 5.1 with Novell Cluster Services at the same time that it purchased the disk array with 10 18GB Seagate drives. We installed QLogic Fibre Channel host bus adapter cards in each server to connect them to the SAN. The servers host the Microsoft Office 2000 suite, together with data used by our program development group.

Unlike many other storage systems, the Magnitude array employs storage virtualization by striping data across all available spindles, using all the storage capacity in the system. This approach makes it possible to perform nearly all storage management tasks quickly and easily from one central console. Users can copy, swap, and mirror data between virtual disks and perform LUN masking/mapping and clustering on Fibre Channel devices attached to the array. All spindles and actuators are used, thereby maximizing the random performance of the system. In addition to supporting NetWare and Windows NT/2000, the disk array supports Linux, Mac OS, AIX, HP-UX, IRIX, and Solaris simultaneously.

The earthquake hits

The SAN has already shown substantial performance improvements and workload reduction. Its ruggedness was put to the test by the earthquake that shook the Pacific Northwest in February 2001. The epicenter of the quake was 12 miles northeast of the DNR's headquarters in Olympia, about 30 miles southwest of Seattle. Some older buildings in the downtown Olympia area sustained heavy damage.

The DNR computer room has a series of shelves that hold the servers. The SAN array was positioned next to the computer rack that holds the six servers. Neither the SAN nor the rack was anchored; they were on rollers with stabilizing feet. The drives are oriented with their spindles perpendicular to the floor, while the shaking motion was horizontal to the floor. The power went off and the building went dark until a few minutes later when diesel generators kicked in. The servers and disk array are protected by a battery-powered uninterruptible power supply that kept them running while the power was out.

No errors or data lost

A few hours after the quake, DNR staff entered the building and manually shut down all the computers and the SAN array. All of the equipment was still running and functioning when they shut it down. An alert was flashing on the disk array but only to indicate that the unit had briefly switched to battery power.

The next day, I returned to the site and brought the systems back up, beginning with the servers and ending with the storage array. The SAN ran through its diagnostic checks without any errors.

The earthquake demonstrated the effectiveness of the disaster preparations made by the DNR and the ability of the SAN array to withstand a major earthquake. The array moved an inch or two, and the computer rack sitting next to it moved about four inches. A disk drive on one of the older servers failed to operate because a connector had come unfastened during the quake. But the SAN itself came though with no problems-and with no data loss.

Bill Reichel is on the network management staff at the Washington State Department of Natural Resources, in Olympia, WA.


BILL REICHEL
Washington State Department of Natural Resources
Click here to enlarge image

The earthquake demonstrated the effectiveness of the disaster preparations made by the DNR and the ability of the SAN array to withstand a major earthquake.

null


Comment and Contribute
(Maximum characters: 1200). You have
characters left.

InfoStor Article Categories:

SAN - Storage Area Network   Disk Arrays
NAS - Network Attached Storage   Storage Blogs
Storage Management   Archived Issues
Backup and Recovery   Data Storage Archives