Why You Shouldn't Implement Gigabit Networking

If there is one absolute truth of networking, it would seem to be this: faster is always better. Since the days when Novell Netware and Banyan Vines were synonymous with state-of-the-art networking and only NASA had 2400 bps dial-up, the conventional wisdom has held that network nirvana is only a megabit-per-second around the corner.

One hundred Mbps Ethernet has given way to Gigabit Ethernet as the demands of video, audio and other rich media traffic raised the bandwidth ante. With all of the exciting things that you can do over IP -- from IP television (IPTV) to voice over IP (VoIP) -- the rule that "faster is better" seems as absolute today as ever. But maybe it isn't. In this age of gigapipe IP, faster might not necessarily be better, says Info-Tech Research analyst Curtis Gittens. There is after all, a lot an organization can do to realize the benefits of higher bandwidth without the expense -- and they might not need it anyway.

Before shelling out for a super-fast infrastructure, it's worth keeping some additional truths about networking in mind:

Network upgrades can be expensive: "The thing to remember is that it can cost a lot of money to upgrade an enterprise network from, say 100 Mbps to 1 Gbps," Gittens says. "It isn't just a question of dropping a faster switch into the network, either. You have to be prepared to upgrade every device on the network to take full advantage of the speed boost."

One of the biggest reasons for cranking up network speeds is to accommodate the delivery of rich media like video and voice. However, Gittens says that, if you have no immediate plans for VoIP, or a need for video, then you probably don't need to be looking for a faster network."The average midsized company with normal workgroup software and e-mail won't see any payback from gigabit Ethernet," Gittens says. "Most of the traffic on most company networks is data, and not voice or video. By adding so much bandwidth, you'll just be adding unused capacity."

Indeed, Gittens says that upgrading to a faster network now in case you do VoIP some time in the future doesn't make any sense. It means investing in technology that will begin to obsolesce even before you have the opportunity to use its full capacity. If the recent history of network technology has demonstrated anything it is that prices will fall as feature sets blossom.

You can better use what you already have: The idea of "getting back to basics" is almost heretical in networking, but an efficient, slower network can provide much of the performance boost offered by the latest gigabit technologies. The truth is that gigabit Ethernet isn't actually any faster than 100 Mbps; after all, electrons don't move any faster across the network cables any faster. What is does, however, is allow more electrons to flow through the cables at the same time (but at the same speed).

Networks seem "slow" because files can be very big, and bottlenecks only allow smaller chunks of data through at once. However, unless you're doing something like real-time weather imaging, that could be a reason to improve network management, rather than investing in a whole new network infrastructure.

"Everyone wants everything faster," Gittens says. "You can't argue with that. But most companies don't manage their network traffic very well to start with. Good, old-fashioned network management can find the bottlenecks and determine whether it's hardware or software."

Prioritize and optimize: Depending on how deeply invested you already are, or plan to be, in technologies like VoIP or IP videoconferencing, there are still great efficiencies to be wrung from existing networks using virtual local area network (VLAN) technology and optimizing the network at the desktop level."VLANs can address many of the issues associated with VoIP and video," Gittens says. "It's not a difficult matter to split off a voice pipe and a data pipe, instead of upgrading the whole pipe."

Moreover, even though 100 Mbps Ethernet is an old, established standard, it's possible that the network interface cards (NICs) in workstations themselves aren't even optimized to take full advantage of that bandwidth. "People forget that the NIC is often a big part of the network bottleneck," Gittens says. "It's one of the places where organizations fail to look when they're trying to get the most out of their networks."

Quite aside from making sure that all your NICs are 100 Mbps -- it can be surprising how many legacy 10base-T cards are still in use -- Gittens points to new technologies like Level 5's EtherFabric cards as a possible key to more efficient networking. "I don't know how that will play out in the long run," he says, "but it could be important." Rather than going for raw speed, he points out, technologies like EtherFabric improve the processing efficiency of the cards themselves.

The bottom line is that, far from delivering a networking ne plus ultra, speed is often a bit oversold. Just as it would be overkill to drive to the convenience store in a Ferrari for a quart of milk, a super-fast gigabit network is, in many cases, completely unnecessary for many organizations, despite what Gittens sees as vendor-driven hype.

"The average mid-sized company won't be installing a videoconferencing server," he says. "Most networks are still strictly or mainly data, and data can run quite comfortably on a 100 Mbps network, despite what the hype says."With that in mind, the gigabit upgrade is often both unnecessary and a drain of funds that can be used more efficiently and economically elsewhere. "If a company has a lot of money burning a hole in its pockets, then there's nothing stopping it from spending as much as it needs to, to get as fast a network as possible," Gittens says. "Most companies aren't in that position, though."

Indeed, sometimes faster is not always better.