Forum

  • Within the last few weeks, I have gotten several inquiries about setting up 802.11 wireless access services for thousands (1000-5000) of people in a conference sort of area (assuming 100% subscription rate, which I think is sort of unreasonable, but that's another story), and I have told them that based on what I know, the 802.11 protocol breaks at those numbers.

    Is there any 802.11-based solution that can handle this density? The only way I have seen people get around it (like at the Superbowl press areas with tons and tons of people) is to try to offload a significant number of users on Ruckus devices using cat5.

    Does anyone have any suggestions here? In these situations, I would just probably put in a ton of smaller access points and then turn the power WAY down and then plan some sort of non-overlapping channel plan with 802.11a and 802.11b/g. I have heard of other solutions (e.g. Proxim) having soft limits on numbers of associations one each AP so that they can, at least, guarantee good coverage with the few who are able to associate to that access point.

    Anyone have any other ways around this? Based on what I know, access points (fat or thin, regardless of the model) crap out at around...

    --about 250 MAC associations
    --about 50 client associations
    --about 25 hardcore user sessions

    Any and all advice on the topic is welcome (even if it is to just tell me I'm stupid for even considering talking to these customers!)

  • Also, I have heard very polarized reviews of Xirrus in handling high density networks well or being total garbage. (Online reviews are awesome, people I know with a lot of 802.11 experience tell me they're garbage, but I'm not sure if it's garbage all around, or just garbage for what they wanted to use it for)

    Anyone have any personal experience with this that they can share?

  • The protocol doesn't break at any user number, its all about the channel. Depending on the usage, 11a/g should be able to handle 15-20 clients per channel in a given area. Notice I didn't say 15-20 users per AP.

    First major question, will the users have 5 GHz (a/n) cards?

    As far as Xirrus goes I can't comment on their product, but the idea is sound if their product is implemented correctly which I can't attest to.

    However, let me say how I consider approaching this problem. Again, keep in mind, even if you have to implement this on 2.4 GHz, the goal of the solution is to minimize co-channel interference.

    That can be done with antennas. Now, I don't know this building at all so I'm speculating, but in a large conference hall, one way to accomplish this is with directional Yagi antennas pointed down. So, image a Yagi with a 40 degree beam width. Given 20 ft ceilings that would give you about a 15 ft radius of coverage on the ground which isn't much.

    A few things that have to be done with this implementation. First, you may have to further shield the outside of the antennas to prevent co-channel interference. Also, the power output on the AP must be so low that the signal is really low. Like mid -75 dBm at the floor. This may have to be accomplished with an added attenuator on the RF cable.

    If you can imagine a large conference areas with flashlights hanging from the ceiling, this is what the signal will look like. If you want to take the analogy further, imagine that there are three different colors of light, red, green, and blue. This is what the three channels would look like. We don't want the light from one flashlight to reflect to an AP with the same color.

    Anyway, enough about that. Any more information you can provide would be great. Thanks!

    GT

  • Meru is the best choice for density with over 256 users per AP...

    Cheers

  • 256 users of what per AP? Ping? Maybe... :)

    GT

  • Rog Escribi?3:

    Within the last few weeks, I have gotten several inquiries about setting up 802.11 wireless access services for thousands (1000-5000) of people in a conference sort of area (assuming 100% subscription rate, which I think is sort of unreasonable, but that's another story), and I have told them that based on what I know, the 802.11 protocol breaks at those numbers.

    Is there any 802.11-based solution that can handle this density? The only way I have seen people get around it (like at the Superbowl press areas with tons and tons of people) is to try to offload a significant number of users on Ruckus devices using cat5.

    Does anyone have any suggestions here? In these situations, I would just probably put in a ton of smaller access points and then turn the power WAY down and then plan some sort of non-overlapping channel plan with 802.11a and 802.11b/g. I have heard of other solutions (e.g. Proxim) having soft limits on numbers of associations one each AP so that they can, at least, guarantee good coverage with the few who are able to associate to that access point.

    Anyone have any other ways around this? Based on what I know, access points (fat or thin, regardless of the model) crap out at around...

    --about 250 MAC associations
    --about 50 client associations
    --about 25 hardcore user sessions

    Any and all advice on the topic is welcome (even if it is to just tell me I'm stupid for even considering talking to these customers!)


    Pico cell comes to mind but that wouldnt fit here due to the special wireless client requirements.

    Installs such as this is really where you need to shape the RF with different antennas, either how GHILL suggested or through your own testing. Co-Channel is going to be your biggest problem. The medium contention will be very high at these numbers.

    I didnt notice what type of applications were being used. I would assume just general surfing i suppose. Or is there an application requirement ?

  • GTHill Escribi?3:

    256 users of what per AP? Ping? Maybe... :)

    GT


    :D :D

  • Hehehehe. This one caught my eye.

    I'm going to have to agree with several of you. First, Meru, Ruckus, and Xirrus are the obviously correct choices here. They should be tested as much as possible before implementation, but if that's not possible, there are some pretty decent 3rd party whitepapers out there - like Meru's from Novarum.

    Gene is correct in that it's all about the channel: avoiding co-channel interference, use of directional antennas where possible (as with Xirrus or use of TxBF with Beam Steering in the case of Ruckus). Meru's L1 QoS (controlling who accesses the wireless medium when) is unmatched in the industry at the moment, but a little birdy told me that you might want to consider looking at Aerohive sometime soon.

    You have to either avoid co-channel interference across a multiple-AP environment through controlling uplink and downlink medium access or through focused beams to get that kind of client density - take your pick. Then you have to hope that the AP's code doesn't puke. :) That's where a pilot test comes in... One current advantage of Meru and Xirrus over Ruckus is 5 GHz, but this is only an advantage IF your audience's client devices can support 5 GHz. At the kind of numbers you're talking about, the 2.4 GHz band would completely saturate, and no solution would work in any reasonable capacity. 5 GHz would be sorely needed.

    One interesting feature that can be helpful in a scenario like this is Aruba's "band steering" which has the controller placing clients that are capable of 5 GHz onto those bands (whether they like it or not) and holding them there. I'm not sure IF (or how well) it will work in a public access scenario though. I'm not sure it's designed for that, and it's part of Aruba's ARM 2.0 technology suite, which is brand new, and unproven.

  • For what it's worth, here is what one big ball stadium does:

    Get a whole bunch of Cisco 1200 series APs, turn down the tx power to 5 mW (7dBm), and attach 4"x4" 9 dBi panel antennas.

    e.g.
    http://www.sparcotech.com/cgi-bin/commerce.cgi?preadd=action&key=ID2450

    The 16 dBm EIRP only covers an area that they're fairly sure will get about 25 users or so.

    This is costly, but it supposedly works quite well, they say.

  • I guess at this point, it's worth asking, "At which point do the client devices start interfering with each other?"

Page 1 of 3