As a grey-haired veteran of early Ethernet, I always try to keep those comparisons in mind. And for those of you who may be newer to networking, you should be aware that Chuck was the go-to guy for technical information in the early days of Ethernet. Thanks for the early education, my friend.
Your points about current and future 802.11 implementations are insightful and I agree that we should all be pushing the vendors for more information about performance while also pushing them on functionality, reliability, and cost. At the same time, we need to be careful in testing the outer boundaries of network performance. While I respect Phil Belanger's work, it is really hard with wireless to know if you are measuring the right system attributes and as has been pointed out, the unique elements of the physical medium make things different in different physical locations. In the end, I think the best tests are the real-world tests that take place at .edu's every day. That's what makes these mailing lists so valuable. I've thought about different ways to accomplish these scale tests, talked to folks like Frank Bulk, Phil Belanger and Craig Mathias. Perhaps if an organization like Educause got involved, they could exert enough pressure to get the vendors to participate. But I doubt that would be in the best interest of Educause or the vendors. My own opinion is that all these systems have breaking points, it's just a matter of finding them and then deciding whether they really matter for your application use case. In the mean time, all you guys get to live with the ambiguity of supporting very large networks while I get to spend time in the classroom teaching my students how to dissect whitepapers and identify vendor misrepresentations. dm -----Original Message----- From: The EDUCAUSE Wireless Issues Constituent Group Listserv [mailto:[EMAIL PROTECTED] On Behalf Of Charles Spurgeon Sent: Tuesday, April 01, 2008 12:00 PM To: [email protected] Subject: Re: [WIRELESS-LAN] Aruba's SCA vs. MCA whitepaper [was: Open Wireless in Higher Ed] On Tue, Apr 01, 2008 at 08:25:06AM -0400, Dave Molta wrote: > I agree with Chuck about the need for better information about WLAN > scalability. It's an issue I've struggled with for many years but I'm not > optimistic about a resolution. I've discussed this issue extensively with > Phil Belanger, the author of the Novarum report, and I just can't see any > way to get all the vendors to agree to a single test plan and commit the > resources necessary for such ambitious tests. > > Although it's clear that Meru designed these tests to best reflect their > competitive advantage, I also believe Phil played the role of objective > analyst, which is not always the case with "pay for test" projects. Phil > pressured Meru to include test runs that were not part of their original > plan and he was very transparent in disclosing his role in these tests. > Still, it's highly likely that the results would have been different if > Cisco and Aruba had been directly involved. Personally, I still think Meru > would have performed better, but I don't think the differences would have > been as great. > > Although the issue of co-channel interference is an important one, I think > it may be reasonable to assert that its importance will be reduced with the > adoption of 5 GHz 802.11n. With over 20 non-overlapping channels, I believe > it will be possible to design high-density, micro-cellular WLANs that do not > suffer from performance degradation as a resulting of co-channel > interference. Over time, I believe 2.4 GHz will be thought of as a > best-effort legacy technology for most enterprises. I'd be curious how > others are viewing this. > Dave, I agree with you in principle about the need to move to 5 GHz -- but we have an installed base of thousands of APs (and tens of thousands of clients) on our campus that are running 802.11b/g, and I find the vendor refusal to test scale and load issues on 802.11b/g systems to be indefensible. The wireless system on our campus has become critically important to the organization. When do the wireless vendors plan to step up to the task and provide real-world large scale testing and useful guidance on these important issues? As the author of a book on the orginal cabled Ethernet system, I can remember the days when Ethernet was being tested for scale and load. In the early days of Ethernet the scale tests were difficult to arrange, since large groups of high performance computers sharing a network were not as common. These tests resulted in a better understanding of Ethernet behavior and limits, and provided network managers with useful information on how best to configure and deploy Ethernet systems. If the wireless vendors persist in refusing to provide scale testing and persist in refusing to provide the testing results and technical information needed to better understand and operate complex 2.4 GHz wireless systems, then what can we expect when we move to 5 GHz? There is some level of co-channel interference even in "non-overlapping" 5 GHz channels. Given the levels of performance collapse that were reported in the Novarum paper, and given the insight documented in that paper that "the interference range is greater than the communication range of 802.11 products" isn't it reasonable to ask that these issues be further investigated? To your point on moving to 5 GHz: we've been working on wireless upgrade plans, including an emphasis on new 5 GHz designs for the usual reasons: increased channel capacity and less interference from non-802.11 sources. Also, we're very interested in seeing what 802.11n can deliver. Like everyone else, we need the new capacity to deal with the massive increases in wireless usage. Some issues that we're encountering include client behavior when it comes to choosing which channel to connect to on an AP that is providing both 802.11b/g and 802.11a/n channels, and client support for the set of 5 GHz channels. While there are 20 (or more) channels defined in the 5 GHz spectrum, the client support for those channels is not what you might expect. I've been digging into the client channel support issue, including some client testing, and will be posting more info as soon as I can get it into shape. Thanks, -Charles Charles E. Spurgeon / UTnet UT Austin ITS / Networking [EMAIL PROTECTED] / 512.475.9265 ********** Participation and subscription information for this EDUCAUSE Constituent Group discussion list can be found at http://www.educause.edu/groups/. ********** Participation and subscription information for this EDUCAUSE Constituent Group discussion list can be found at http://www.educause.edu/groups/.
