The Hubbub Over Hubs

y the time NASA's space shuttle Columbia was gearing up to come home last November, Cmdr. Ken Bowersox and his crew had circled the earth 256 times-once every 90 minutes-and had supervised 14 microgravity experiments sponsored by public and private institutions. Research conducted in the reusable Spacelab section of the shuttle's cargo bay included everything from evaluating the growth of protein crystals to studying the nutrient consumption of plants in a weightless environment.

Throughout the mission, information about the payload experiments-including telemetry data, and digitized audio/video-was beamed in real time via satellite to the Operations Support Center at Marshall Space Flight Center in Huntsville, Ala. From there, the information was instantly routed to scientists and engineers at sponsoring institutions and space agencies around the world.

From the comfort of their own facilities, researchers were able to monitor payload experiments around the clock and send commands to the shuttle's crew orbiting 150 nautical miles above the earth. And they were able to support these communications with relatively inexpensive, commercial, off-the-shelf computer equipment that was linked to NASA's custom-built, high-end systems.

"This is probably the biggest example of telecommuting ever recorded," says Keith Cornett, chief of the mission system division at the Marshall Space Flight Center. "Without the right technology in place, all those scientists would have to travel to Huntsville to monitor the payload and communicate with the shuttle."

The centerpiece of the technology described by Cornett is a 2-foot-high rectangular box known as a hub. Hubs reside between PCs-known as clients-and servers, which are powerful computers that store and distribute data to networks. Hubs govern where audio, video and data-communications units-called packets-should go within a network or series of networks. They do this with the help of bridges and routers, which identify the most efficient paths for transmission on both local-area networks (LANs) and wide-area networks (WANs).

Hubs from companies such as Bay Networks, Cabletron, Cisco Systems and 3Com Corp. are capable of handling communications from multiple network protocols. At Marshall Space Flight Center, for instance, scientists monitoring payload experiments send commands via Ethernet, FDDI and TCP/IP networks, depending on the computer architectures of their institutions. The Cabletron hub at Marshall collects the data and directs it to the appropriate location.

"Otherwise, we'd have to rely on point-to-point routing where we would string dedicated wires to all the research facilities-a prohibitively expensive task," says Cornett. "The hub lets everyone be connected on the same network, and enables us to transmit to remote sites over lower-cost lines such as the Internet."

One of the main advantages of hubs is their ability to sort through data congestion and quickly move files to their destinations. Hubs used by the Army Corps of Engineers, for example, act as traffic cops directing the network movement of large files, such as computer-aided designs of waterway construction projects.

"Hubs help to speed our backbone architecture and make bandwidth available for monster graphical files," says Ken Shafer, automation and telecommunications project director at the Army Corps of Engineers division in Huntington, W.Va. "They are essential for multimedia and imaging applications."

Hubs range in price from $5,000 to $50,000, depending on the number of ports and the sophistication of the switching hardware. The more expensive models generally have built-in bridging and routing devices and feature forwarding rates as fast as 750,000 data packets per second.

Network hubs are being used in many applications. The Customs Service uses more than 200 hubs from Bay Networks-the merged company of SynOptics and Wellfleet Communications-in a network comprised of 30,000 users at 2,000 locations nationwide. The hubs connect dozens of Ethernet LANs that carry passport information from databases at airports and border checkpoints to various agencies.

The Army, meanwhile, is using hubs in its Bosnia field networks to quickly transfer unclassified logistics information to command centers without exposing troops to land mines and sniper fire. And the Navy has started using hubs on battleships to eliminate the "sneakernets" that exist when data is manually transferred-by a person carrying a disk-between administrative systems such as payroll and personnel.

"Hubs may sound boring but they're absolutely mission-critical," says NASA's Cornett, who advises network managers to be careful when selecting hubs. He recommends buying a chassis that is easy and inexpensive to upgrade with printed circuit cards. "Get a hub that can handle 20 cards, even if you only need two cards initially. It's amazing how much your requirements can grow once the network is up and running."

Cornett also suggests getting a hub that enables network managers to "hot-swap" cards-replace individual circuit cards without having to shut down the entire system. That ensures most applications will not be interrupted during a card swap. For the same reason, he endorses hubs that come with dual power supplies-just in case the main power source goes down.

"Above all, make sure you've got some good network-management software in place to monitor hub activity," says Cornett. "If a card fails and we don't spot the error immediately, we will lose data and the institutions sponsoring Spacelab experiments will refuse to pay their bills."

Network-management programs-from companies such as Cabletron, Computer Associates, Hewlett-Packard, IBM and Sun Microsystems-provide round the clock status reports on hubs and other network components. Overuse of products such as bridges, gateways, routers and repeaters can lead to traffic congestion, component failures and cable faults. Special software helps managers of LANs and WANs quickly diagnose problems so networks can be reconfigured before data loss occurs. Building in that type of reliability can cost agencies $5,000 to $25,000 per program, depending on the level of network complexity.

In recent years, some companies have started incorporating artificial-intelligence tools into their network-management programs. These enable the software to keep networks running efficiently by automatically rerouting data or isolating faulty devices. They also provide multilevel security on networks.

The need for network-management software increases with the number of users, technologies and applications. If used properly, the programs can save agencies time and money.

"With more than 180 billion pieces of mail to deliver every year, we can't afford computer glitches," says Dan McPhillips, telecommunications program manager at the Postal Service, which uses Cabletron's Spectrum software to monitor a network of 75,000 computer terminals, 500 LAN hubs and 400 routers scattered across the country. "Centralized management is the only way to go."