Frankly, Ive been surprised and a little disappointed lately by the number of people who ask, when the subject of 10 Gigabit Ethernet comes up, “Who needs all that bandwidth, anyway?”
What surprises me most is which folks are asking. Often, they arent your typical corporate Luddites—the CXOs and bean counters who are paid to build stone walls around the bottom line, often at the expense of investment in network needs. Strangely, the question is frequently asked by IT people, the very folks youd assume would know better.
After all, bandwidth is like money. Even if you have enough today, you will eventually need more—lots more. Which means that even if you wouldnt know what to do with 10G bps of data pumping through your companys optic fiber today—or couldnt even handle a measly 2G-bps Fibre Channel—you had better be planning an adequate infrastructure now for the demand youll be facing in a few short tomorrows down the road. In which case, 10 Gigabit Ethernet looks like the answer to all our LAN-WAN-MAN-SAN prayers—if it can get over a few hurdles.
Ethernet has a deserved reputation for outlasting better-designed rivals (remember Token-Ring?), but its future at these speeds is going to depend on ancillary technologies over which Ethernet vendors have little or no control.
Take, for example, storage area networks, an area in which Fibre Channel has a huge head start. As an IP technology, 10G bps cant make much headway in the SAN market unless there is simultaneous rapid development and deployment of the iSCSI standard. While iSCSI is certainly showing great promise, its success cant be guaranteed.
Whats more, there exists no backbone application that requires 10G-bps speeds. But at least at the backbone level, the routers, switches and other hardware can keep pace. At the enterprise level, the fact is that the vast majority of servers today cant handle traffic at anywhere near the speed of Gigabit Ethernet, let alone 10G bps. That means its initial adoption in the enterprise market will be pretty much confined to network aggregation—often at the WAN level. It would offer no advantage for client/server applications, a fact that will probably slow the development of network interface cards and other hardware components.
Yet, chip companies are betting servers will eventually catch up—and that fiber may not even be needed for all enterprise networks. In fact, some, including Intel, are talking about running 10G bps over standard Category 5 twisted-pair cables more than 300 feet without a repeater. That would enable the well- endowed desktop computer, for example, to receive dozens of HDTV-quality videos simultaneously.
Hm, who needs all that bandwidth, anyway?
How fast will we move to 10G? Tell me at [email protected]