The 10Gig switches we reviewed follow Ethernet’s established evolutionary pattern. Until now, Ethernet innovation has generally moved in factors of 10 and has typically borrowed specifications from existing technologies running at competing speeds. Beneath this framework, Ethernet innovation has followed a fairly regular course.But times they are a-changin’. Instead of welcoming Ethernet innovation with open arms and checkbooks, potential customers are looking for long-term returns on their hardware investments. Purchases are being scrutinized with an eye on the bottom line, which reflects on the capital expenditures required to feed Ethernet’s greedy bandwidth maw.This is further complicated by the schism that 10Gig seems to have created between enterprise customers and service providers. With 10Gig offerings weighing in at approximately $25,000 per switch port, every 10Gig customer is making his buying decisions much more carefully. That should mean we’ll see significant differences in products aimed at enterprises vs. service providers — even more so than with the Force 10 and Foundry switches reviewed here. This careful buying environment is almost hostile toward new technologies. Vendors that back the wrong horse will be punished in the marketplace. Which brings up the most basic question about Ethernet’s next step: How fast should it go? If we follow Ethernet’s standard path as dictated by the IEEE 802.3 working group — namely, 10 times the previous iteration — we quickly arrive at 100Gig.But considering the divide between enterprise users and service providers, Ethernet may require an additional iteration at 40Gig to keep pace with the WAN technologies typically employed by service providers, especially SONet, which tends to move forward in increments of four from OC-192 (10Gbps) to OC-768 (40Gbps).Another reason for 40Gig is that some of the development required to run at this speed has already been accomplished in silicon. Some companies, including Foundry and Mintera, are already well along the path. But roadblocks remain. Distance is crucial from the service-provider perspective. Implementing 10Gig beyond 50km is still a problem for many manufacturers, and researchers working on 40Gig have stated that even distances beyond 10km remain a challenge. To say nothing of required rewriting of basic switching algorithms and providing enough ASIC horsepower and buffer memory to allow for faster packet processing and table lookups.Of course, when those problems are solved for 40Gig, those same researchers feel they’ll be all but solved for 100Gig. In the end, we may see a simultaneous arrival of 100Gig products for the enterprise and 40Gig boxes for service providers. No need to fret now, however; we won’t be seeing anything for at least two years. Technology Industry