By now, you know the big attraction of Gigabit Ethernet: it packages information just like classic Ethernet but moves the bits 100 times faster.

Since Bob Metcalfe at Xerox invented Ethernet in the early 1970s, he has enhanced it several times. Standard Ethernet’s transmission speed has been increased twice in recent years: Fast Ethernet (100 Base-T) was standardized as IEEE 802.3u in 1994 and Gigabit Ethernet (1000 Base-T) was standardized as IEEE 802.3z in June 1998.

Gigabit Ethernet ran first on fiber optic cable, but the unshielded twisted pair (UTP) option was standardized by IEEE 802.3ab in June 1999.

Most likely, the new standard will first be used for server farms and backbones, since this much bandwidth is hard to justify for a single user! Nevertheless, vendors expect prices to plummet fast enough to make Gigabit Ethernet to the desktop economically viable soon. In fact, Jeff Shapiro, author of “Gigabit Ethernet on the Fast Track to the Desktop,” Network World, (June 12, 2000), predicted Gigabit will cost just slightly more than Fast Ethernet by the end of summer 2000.

For some time, production network managers have been installing combination Ethernet/Fast Ethernet adapters and hubs so they may begin installing 10/100/1000 autosensing adapters soon, just to preserve their options with a small additional investment.

However, others are more critical of this trend toward Gigabit to the desktop. For example, in Kevin Tolly’s June 12, 2000 Network World article, “Gigabit to the Desktop: Foundry Get Real,” questions whether the current generation of desktop PCs can keep up with the extreme Gigabit data rate, given their available cpu speed, I/O bus rates, and disk access speeds. The obvious question is: Do we need Gigabit to the desktop? Gigabit is extremely hard to justify when many high-end work group and departmental networks use 10/100 switching hubs that allow individual users to enjoy large bandwidth connections. Even many current servers cannot take advantage of Gbps transmission rates.

What components are needed to operate a Gigabit Ethernet network? Computer, adapters, media, and a hub are required. The adapters are available, but most state-of-the-art computers (clients and servers) can’t handle such speeds. The exceptions in the computer category are the routers, which have been optimized to be able to switch millions of packets per second.

What about the wiring? The current wiring options available with maximum distances for Gigabit Ethernet are: laser transmission over singlemode fiber cable =5km; laser transmission over multi-mode cable=550m; electrical transmission over STP or coax=25m; and electrical transmission over STP or coax=100m.

Remarkably, Gigabit Ethernet’s distance has been standardized at 100 meters for Category 5 UTP, just like Fast Ethernet and standard (10BASE-T) Ethernet. This evidences not only the standards committee’s hard work to find a way to support distances of “at least 25 meters,” but also the rapid development and evolution of digital signal processor (DSP) and application-specific integrated circuit (ASIC) technologies. The current Gigabit Ethernet standard requires use of four pairs, while the Fast Ethernet (100BASE-T) has been adapted to run on only two pairs.

The new UTP 1000BASE-T standard may also reduce the demand for new high-performance UTP cable specifications. Since it became clear that Gigabit Ethernet would be developed, efforts have been made to standardize the cable it would run on. To date, however, the EIA/TIA 568 standard has not been officially extended to support more than 100 MHz (Category 5) cabling. Because there is a perceived need for these high-performance cables, however, products in the field are advertised as Category 5+, Level 6, and so on. Although no standard has been issued yet, one has been drafted and is in the approval process. An important issue for wiring contractors is whether the cable vendor will replace the prestandard products if they do not meet the standard when issued. Unlike the adaptor and hub products, where this warranty is often made, it is less effective and less likely to be made by cable vendors because installed wire is much harder to replace in hubs and network interface cards (NICs).

Another development is the experimental use of Ethernet in the wide area network (WAN). While Ethernet has always been used for local area network (LAN) communication before, when a simple point-to-point WAN frame structure is needed, it can be pressed into service. This is the ultimate irony, given the vision a few years ago of a ubiquitous asynchronous transfer mode (ATM) in both LAN and WAN environments.

ATM to the desktop was scrapped in favor of the simpler and cheaper Fast Ethernet. Now Ethernet may also be replacing some ATM applications in the WAN. However, no quality of service mechanism has yet been defined for Ethernet, so ATM seems safe for now in the WAN.

What’s next? The IEEE standards committee for 10,000BASE-T (10 Gigabit Ethernet) has already been formed and is studying a draft. It expects to deliver a standard by 2002.

SHINN is principal of compuTutor Consulting in Denver. For more information, visit his Web site at