Network Design M ade Easy: Routing On A Chip

By Art Wittmann  We all know TCP/IP is here to stay. Less chatty than IPX, more efficient than AppleTalk and more scalable than NetBEUI, TCP/IP has been extended and built upon to the point where this rock-solid protocol fills almost every imaginable networking need. Microsoft, Novell and Apple all have devised ways to make their back-end services, as well as their end-user offerings, work with TCP/IP. The database folks appear to be quite pleased to run their shows over TCP/IP and, of course, everything Internet begins with TCP/IP. TCP/IP is winning over the hearts and minds of vendors and of network designers as well. And now all the deep thought that's go ne into TCP/IP has brought us one more minor miracle: cheap switches capable of moving data at tens of millions of packets per second. I recently met with an old friend; he's been around the high-speed networking part of the industry just a bit longer than I have and--clever guy that he is--he's now a muckety-muck at Extreme Networks. He briefly described the product his company is producing--I have to admit, I wasn't as familiar with it as I probably should have been, so I listened intently. He then described some new products that his company would be bringing out soon. His own excitement for his company's products was understandable. And once the plan was laid out for me, I became pretty excited, too. Served on a Chip Extreme Networks, like a number of other Gigabit Ethernet vendors, has shrunk TCP/IP routing down to silicon, and in the process it's created one of the multimillion-packet-per-second engines. Extreme Networks now plans to deploy it in products that range from backbone swit ches right up to high-density desktop switches--all at a price that would seem to sound the death knell for any architecture that doesn't include Layer 3 switching to the desktop. It also should be the final nail in the coffin for MPOA (Multiprotocol Over ATM), which is too complex and looks way too expensive. The other thing it does is raise serious doubt about the proprietary technologies that have been devised to speed TCP/IP within a campus network. Ipsilon (now NEC) and 3Com in particular have technologies that seem less viable to me now that affordable Layer 3 routing will be available from many sources. VLANs Forever This isn't the end of the story for VLANs. It still makes sense to take groups within your organization and put them into broadcast zones regardless of their position within the company. Layer 3 switching simply means that if you aren't in the same workgroup as the guy next to you, then the packets that travel between you don't necessarily have to schlepp all the way back to some backbone switch or router to find their way to the cubicle next door. Layer 3 switching provides incredible flexibility to the network designer--and companies like Extreme Networks price it so you can use it even at the far corners of your network. Extreme Networks also intends to offer a management interface that is consistent with its backbone switches through its desktop offerings, making the job of managing the network somewhat less daunting. If Cisco's IOS (Internetwork Operating System) is good enough for the millions of Cisco customers out there, then the folks at Extreme Networks figure it's a good enough strategy for them. I don't expect--or recommend--that any of you go out and buy a stem-to-stern Layer 3 switched network from Extreme Networks or any other vendor for that matter. Most of the companies doing Layer 3 switching are small and have no proven track record--and many will be bought by the behemoths of the industry. But when it comes time to dole out the cash for upgrades, it's a good idea to think about Layer 3 switching as a technology that can be used anywhere, and it's really nice if you don't have to think about all those proprietary TCP/IP quick fixes. You simply don't need them. Art Wittmann can be reached at awittmann@nwc.com.

Other Columnists Corporate View By Robert Moskowitz Net Results By Dave Molta