> Fantastic suggestions by Tyler, Mike, and Russell. I have very > little to > add. > > If your APs are proper *nix computers, you can do some balancing/ > shaping > at the AP. If all your APs are pushing the same SSID, clients are > going > to associate by SNR, which in a big concrete box, might not be what > you > want. You may end up with 400 users on one AP due to unlucky > reflections. You could use a watch-dog script on the APs to keep the > associated users below some threshold (assuming they will associate > with > the next best if you reject them). Using different SSIDs might be a > way > to get users to balance themselves.
You're probably giving end-users too much credit here, I think this would just cause confusion with most. Any proper AP will allow you to set limits to the amount of associated users you'd like to allow, setting the same SSID is the cleanest way to "load balance" without some smart centralized controller. If you set the limit to something sane (this depends highly on what your AP can handle), it shouldn't be a problem. > I fully agree that doing LAN-drops to each AP is the only way this > will > work. You don't want to manage interference and capacity issues in > your > backhaul on top of all the other headaches. > > My intuition is that 50 users per AP is pushing it. Tyler says 15-20 > active and 70-100 total. He has a lot more experience with this than I > do. I would never suggest this with a consumer grade AP, or on many "professional grade" APs for that matter. However, if you throw enough CPU, RAM and a decent radio card at the problem, you should easily be able to handle at least 100 idle users. > Any way you cut it, you're going to need at least 10 APs for 500 > simultaneous users. At that point, limiting yourself to the 3 > non-overlapping channels might not be optimal. If you can't avoid 5 > APs > in a room, for instance, then 1,3,6,9,11 is probably better than > 1,1,3,3,11. Especially as you might be able to put neighboring rooms > on > channels that are orthogonal to the closest APs (hope that makes > sense). This can be some-what mitigated with high-quality radios and signifigantly lowering your radio power as I suggested creating "pico cells", the addition of channels 3 and 9 can also be a good idea, but sometimes it can cause more problems. It depends on the client devices and behaviors. Since neither can be predicted, I'd try to stick to 1,6,11 if you can help it. You can experiment with ch 3 when the AP in question is physically located farthest away from any APs on 6 and 11, and with ch 9 when the AP in question is physically located farthest away from any AP on 1 or 6. > Give some thought to the relationship between floor-plan and temporal > use patterns. If everyone is going to be in the main hall during one > time period and distributed during others, maybe it makes sense to > turn > some APs off or on at certain times. > > It's worth noting that blogging is not a high activity use-case - it > involves lots of periods of 'think-time' between HTTP requests. If > your > users really are just blogging, you might be able to squeeze more > out of > less than you think. Right, the real problems start when said blogger gets bored of blogging and starts watching YouTube to pass the time.... > It's not 500, but Russell and I once unwired a conference of this > style > with probably 100 users using 3 Netgear WGTs on separate channels and > the same SSID, each with a LAN drop, and each at a separate corner of > the space. It worked very well. > > If these all seem like hacks, it's because they are. This scenario > is a > motivating case in ongoing research. The good ideas are still a couple > years from making it into COTSE products. > > Let us know what you do and how it turns out :). > > Tyler Booth wrote: >> Joe, steer clear of mesh networking when dealing with large volumes >> of >> people in a small enclosed area. A far better approach is to hard- >> wire >> each access point. >> Utilize grossly over-powered access-point boards (Mikrotik RB600, >> Alix3d2, etc) and spend some money on the radios, Ubiquity XR2 is a >> very >> high-quality 2.4ghz radio. >> Lower the radio power to ~15db which will put it on par with about >> 90% >> of the devices connecting to it and will cause less interference. >> Because conferences tend to be in buildings with thick cement >> walls, you >> have the advantage of a very low noise-floor with the exception of >> the >> noise that you create yourself as well as other conference attendees. >> This means that if you space out your access points with adjacent >> access >> points on opposite sides of the spectrum (use channels 1,6,11) you'll >> minimize the effect of interference. >> >> You'll need to install enough access points to not only take care of >> active users, but associated users as well (how many >> iphones/blackberries do you expect to attend). >> An XR2 card with either of the two CPU boards suggested above >> should be >> able to handle 70-100 associated users, but likely not more than >> 15-20 >> active users at a time. >> If you have the budget for it, add an XR5 (802.11a) card as well >> broadcasting the same SSIDs which will allow users with 802.11a >> adapters >> to connect when the nearest 802.11b/g ap is overloaded. >> >> You'll want to utilize bandwidth shaping on the upstream router to >> prevent any single user from spoiling the party for everyone else. >> It's >> also probably a good idea to block p2p traffic as a rule for your >> wireless network as it can quickly consume all of your >> available time-slots regardless of how much bandwidth it's actually >> consuming. >> >> >> stephouse networks <http://www.stephouse.net/>*Tyler Booth* // >> President >> ph. 503.548.2000 | fx. 503.548.2002 >> 921 SW Washington St, Suite 224 >> Portland OR 97205 >> >> >> On Feb 11, 2009, at 2:22 PM, Joe Christensen wrote: >> >>> >>> Hello, >>> >>> I have been researching event wifi and mesh networking (Meraki & >>> Open-Mesh). Does anyone have experience setting up wifi for a >>> densely >>> packed room full of technical conference attendees? Could a mesh >>> network survive in this situation? >>> >>> It seems like 802.11 is ill suited for 500 live bloggers in a >>> conference room. How is this accomplished? My basic understanding >>> is >>> that APs have 3 desired channels (1,6,&11). How does this scale? >>> >>> I have read of two high profile conferences with 1000+ technology >>> focused attendees that had horrible wifi issues. (Techcrunch50 & >>> LeWeb) LeWeb paid 100k for the event networking contract. >>> Techcrunch >>> replaced the wireless vendor after the 1st day failure. Are >>> conferences full of technical people starting to exceed the limits >>> of >>> 802.11? >>> >>> Any links or personal experiences are appreciated. >>> >>> Thank you, >>> >>> Joe Christensen >>> >>> >>> >>> >> >> >>> > > > > --~--~---------~--~----~------------~-------~--~----~ The Personal Telco Project - http://www.personaltelco.net/ Donate to PTP: http://www.personaltelco.net/donate Archives: http://news.gmane.org/gmane.network.wireless.portland.general/ Etiquette: http://www.personaltelco.net/index.cgi/MailingListEtiquette List information: http://lists.personaltelco.net To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] -~----------~----~----~----~------~----~------~--~---
