A third session that I went to dealt with Carnegie Mellon's 802.11n wireless roll-out. The folks at Carnegie Mellon started their roll-out in Fall 2007. They selected Xirrus and Aruba Networks as the respective vendors (I found it interesting that they decided to go with two different types of APs). One of the problems this presents is having two different management interfaces, though the folks at Carnegie Mellon don't find this to be as big a problem as gathering aggregate statistics. On top of this, they've implemented one flat subnet for their wireless network - which they admitted is not the best design. In addition, they do manual site surveys using AirMagnet and aim for -70dBm RSSI (received signal strength indicator - a measurement of the power present in a wireless signal). They're not big fans of the Aruba approach to wireless surveying - i.e., predictive surveying done by inputting information into the appropriate tool which then spits out its predictions as to the RF environment given the inputs. One of the problems they ran into ith site surveying was the fact that many tools out there - at least in their incarnation at the time they were doing the surveying - did not recognize 802.11n networks/adapters.
They recommend using the 5 GHz frequency wherever and whenever possible for 802.11n (and I agree with them on this point). The main reasons for this are that there is less interference from devices that operate on the 2.4 GHz frequency (e.g., microwaves, cordless phones, etc.) and that the use of channel bonding to increase throughput gives you only one non-overlapping channel (there are three if you forego channel bonding) whereas there are 12 non-overlapping channels (twenty-four if not using channel bonding) when using channel-bonding in the 5 GHz spectrum.
On the wired side of the wireless network, they run two Cat5e cables to each location (mostly for future expansion, it appears) and use 802.3af Power over Ethernet technologies. In addition, they use a centralized Aruba Networks wireless controller and have a 10G network core. The folks at Carnegie Mellon recommend having gigabit wired links for 802.11n access points due to the throughput that 802.11n offers.
Finally, they currently are using MAC-based authentication (which realistically is no real type of authentication given how trivial it is to spoof a MAC address). They have plans to look into 802.1x authentication in the next phase of the project. Other than this, their network is pretty open in that they're not using encryption or enforcing authentication (both of which we currently offer, though we only enforce mandatory authentication via a captive portal). One interesting finding that they revealed was that 36% of the students in the dorms using 802.11n - 58% of the dorm residents use 802.11g wireless adapters.
Overall, I found this session to be pretty interesting. There are a couple of things that the folks at Carnegie Mellon are doing that we probably should consider - e.g., gigabit connections to the access points, use of the 5 GHz spectrum where possible, use of AirMagnet, etc. There are also a couple things that I wouldn't do - namely, having one flat subnet for all wireless clients, using MAC-based authentication (although that's better than having no authentication/encryption), and going with APs from two different vendors. Although we are using wireless APs from Cisco and Aruba Networks, we are currently phasing out (slowly) the Cisco access poins while increasing our deployment of the Aruba equipment.
No comments:
Post a Comment