In-Depth

Clear Skies Over Atlanta

A sophisticated weather system at The Weather Channel is managing a huge stream of data to be channeled from storage to TV, radio, Internet , PDAs and cell phones.

• By Drew Robb
• 09/01/2002

For those seeking more up-to-date information with greater accuracy, however, there's The Weather Channel (TWC). Landmark Communications debuted the 24-hour cable service in 1982; it now has 80 million subscribers in North and South America. The company also provides information to radio stations serving 250 markets. Its Web site (www.weather.com) provides 14 million monthly visitors up-to-the-minute information on 77,000 locations worldwide, while another 150 online services, including EarthLink and CompuServe, use its information. The site also offers customized telephonic and wireless services.

To provide the latest information, TWC receives a continuous stream of data from satellites and ground-based systems, tailors it according to the needs of each of its thousands of customers, and transmits it within minutes.

While routine "Should I carry an umbrella?" requests predominate, news services demand instantaneous updates. Rapidly changing climate systems—hurricanes, tornadoes, blizzards—cause unpredictable system spikes, just as TWC's services are most valuable.

For TWC, it wasn't a matter of having adequate storage capacity, but being able to manage it rapidly. As weather data has a very short shelf life, most interest lies in today's and tomorrow's figures, not yesterday's. However, historical data must also be maintained accurately and instantly accessible.

Concerns about Use
The company had centralized storage provided by two EMC Symmetrix units, one holding 1.5TB and a loaner storing an additional .5TB. Most servers also had locally attached storage consisting of Sun D1000 arrays with Sun A3000 controllers. Other equipment used local RAID arrays. All together these disks provided enough space, but that space wasn't where it was needed when it was needed, crippling TWC's ability to rapidly respond to changing customer needs.

"The primary storage problem was less one of capacity than of ease of utilization," explains Vicki Hamilton, vice president of shared services and IT operations. "As projects waxed and waned, the amount of storage allocated to a given server was often inappropriate, and a considerable amount of disk space was either wasted or had to be physically moved from server to server to accommodate changing requirements."

Uncompromising reliability was also essential to meet the Service Level Agreements for timely delivery of localized content for each of its customers.

TWC decided to replace its fractured storage system with a Storage Area Network (SAN).

Selection criteria included price/performance, expandability and ability to be upgraded, supportability, data protection, single points of failure and raw performance numbers (such as time to recover failed disks and performance impact during the process). The company evaluated systems from MTI, Sun, IBM, EMC, and others before choosing a Hitachi Data Systems (HDS) Freedom Storage Lightning 9900 Series Model 9960.

"The Hitachi is the hands-down winner, with twice the performance of the nearest competitor, no single points of failure, and layered redundancy, such as dual FC-AL [Fibre Channel-arbitrated loop—a high-speed serial bus interface standard] to the spindles," Hamilton explains.

The Weather Channel's Storage Area Network project included implementation of Hitachi Data Systems Freedom Storage Lightning 9900 Series Model 9960 and Brocade Silkworm 2800 switches in a mixed Novell/Microsoft environment

Implementation
HDS and TWC designed a three-tiered environment consisting of the 12TB of RAID-5 storage on the Lightning 9960 (it scales up to 37TB), four Brocade Communications Systems Inc. Silkworm 2800 switches and the servers. TWC kept Veritas NetBackup and Veritas Foundation Suite storage management software for its new SAN.

The Hitachi system was delivered in December and was online within a week of power on. Migrating all the servers from their local stores to the SAN is an ongoing process. Specific issues encountered centered on making the host bus adapter (HBA) drivers work correctly.

"We also sidestepped a lot of potential zoning headaches by dedicating specific Hitachi ports and Brocade switches for use by Windows and Novell operating systems," Hamilton notes, "effectively creating parallel SANs front-ending the array."

She feels that better site planning and facility-requirements definitions would have made for a smoother implementation. Hamilton recommends others have vendors demonstrate in its test labs that the equipment works with the specific configuration the client will be using.

To ensure there's no downtime, the 9960 sends status and alert information to the Hitachi Support Center through a dedicated phone line. HDS support staff either fix the problem remotely or contact TWC with what needs to be done. In this way, potential trouble is handled before it results in service interruptions.

Overall, Hamilton reports that TWC has achieved significant soft savings including scalability and reliability and effective utilization of storage management personnel. Organizational "churn" has also been reduced.

"Our ability to be responsive to customers has improved 10,000 percent," she relates. "Satisfying ad-hoc requests for the odd few-hundred gig can now be filled in less than an hour, as compared to the days it typically required to plan ‘from what server do I "borrow" disks, and when can I shut that server down to do it?' games."