Space was getting tight last year in the Microsoft Corp. TerraServer data center: The Digital Equipment Corp. Alpha 8400 server that housed the TerraServer SQL database had grown to 20 feet long with all its upgrades, allowing no room to attach more equipment. Likewise, the TerraServer database was running out of internal headroom to handle its massive topological and imagery files, so Microsoft replaced the fat server with clustered Windows 2000 machines and swapped its direct-attached storage devices for a SAN (storage-area network).
TerraServer (terraserver.net) is the massive SQL database that houses the U.S. Geological Survey's (USGS') topographical maps and aerial photographs, which are used by a wide variety of people and organizations -- from real-estate agents checking on property, to guides mapping hiking trails, to the U.S. Department of Agriculture conducting its annual survey of land use in the states. This experiment-turned-public-service, which sprung from a joint research project with Microsoft, the USGS and Compaq Computer Corp., also serves as a hefty test bed for storing, accessing and culling metadata from terabytes of information.
By implementing the SAN, Microsoft was able to expand the TerraServer service. "We have more available CPU and disk space with the SAN, and it's easier to grow by adding another server to the configuration while it's live if we need to," says Tom Barclay, a researcher with Microsoft Research, which runs the TerraServer site. "When we had direct-attached storage, we had to physically drive to the site to move cables. With the SAN, there is no single path to the disks, so control of the application can transfer from one server to another" without anyone handling it on-site, he adds.
TerraServer has three database servers plus a hot standby. Among the specs: 18 TB of storage space in the Compaq StorageWorks disk arrays with 72-GB 10,000-RPM disk drives. The metadata gleaned from the databases is stored on smaller, 18-GB 15,000 RPM drives, and all the devices are Fibre Channel-attached with Brocade Communications Systems SAN switches. "There's no single point of failure in any of the storage arrays," says Will Monin, manager of Microsoft storage products development for Compaq, which supplies the servers and controllers for the SAN. TerraServer averages about 45,000 visitors daily, 900,000 custom-image page views (which require 5.6 million queries) and 10 GB of new imaging data added to it each day.
IT Department Info
Size of TerraServer IT staff: Three part-timers
Barclay's Average Workweek: 50 hours, 5 to 10 of which are dedicated to TerraServer
Biggest Challenge of TerraServer: Figuring out how to geo-spatially stitch together more than 250,000 images sized from 50 MB to 160 MB into a seamless, compressed mosaic of earth when the images arrive in random order over a four-year time span.
Latest Projects for TerraServer: Launching the TerraService Web service; adding celestial images to TerraServer.
Coolest Part of the Job: The people we've met and have had the opportunity to work with along the way. TerraServer is a successful joint research project that has brought together a diverse group of people and organizations.
REPORTS
Analyize In-Line NAC strategies and products.
ANALYTICS Plan and design your enterprise blade server deployments
InformationWeek U.S. IT Salary Survey 2008
Salaries for business technology professionals are falling. Here's what you need to know in order to make good hiring decisions and personal career choices. Purchase Today: $299