help with large temporary network

Hi,

looking for some advice (long and complicated advice)

i need to set up a temporary network in a hotel. 802.11b is required to accomodate all the users.

There will 200-250 users, the hotel backbone is a mixture of 10baseT and

100baseT...I will try my hardest to get to 100baseT connections to the Access Points.

The users will be +generally spread over two floors, but the biggest challenge will be opening and closing when everyone will be in the same room.

I've been reading mixed messages about capacity and access points. From what I can gather, 20-25 users per access point can provide decent service. So I've been looking at putting in 12-15 APs spread around the hotel. (and moving some into the big room for opening/closing)

For the big days, it's a typical open hotel ballroom 67' deep and 72' wide with 250 high-bandwidth using laptops to connect to a files server onsite and doing various things on the internet.

My question boils down to how to provide the capacity (much more than coverage, as the space is not too big, and it's open). If 20-25 per AP is a rule, I would need 10 APs in the room, but would interference then start to cause a problem between the APs and remove the benefit of the extra capacity?

My thinking right now is to place 9 APs in the room. (they will be connected either right to a lan jack or through a switch in the room, depending on the set up) One in each corner, one in the middle of each wall, and one in the centre of the room (using a 1/6/11 channel plan). That way (my thinking goes) there will be roughly equal overlap between the APs and users will be roughly evenly distributed in proximity to the APs. I want to make sure the users are spread out over all the APs, I understand that windows clients will "lock on" to the strongest signal, but will they also shift to an AP with more available capacity even if the signal might be a little weaker? I think the APs I'm looking at have varialbe output power. Would it help or hurt if I turned down the power of the APs?

I had a second thought entirely....overlay a second SSID to try to get users balanced. Using the same layout above, but with every second AP using a different SSID.

Can someone tell me if I'm anywhere near the ballpark, and offer any advice with respect to capacity requirements/layouts/tips tricks etc and offer answers to my questions?

I also have a few generic questions.

Is there a generic way to filter out specific types of traffic (ie VoIP) at the AP level, or does it depend on the Access Point itself?

If I manually limit the bandwidth of the APs (to 2 or 5.5 Mbps), would that help capacity, or are the automatic adjustments of the devices sufficient?

I'm hoping to be given a block of IP addresses to give to my APs so I can manage them quickly, but if I don't and they are assigned IPs through DHCP, how can I (using WinXP) find them and log on quickly? (i intend to use a web interface to administer them, I don't have time to really learn the command line interface)

thanks for any advice in advance, even if it's simply pointing to another forum or a good book. The books I've read have been helpful, but not really on point for what I need.

thanks,

I had downloaded that as part of my research but hadn't read it yet. It helps, but I'd still like some feedback on whether or not my current thinking will lead to a network that will even remotely serve the users.

service. So

So you expect up to 250 simultaneous users? What is the bandwidth of the Internet connection? How large of files will they be downloading from the file server?

You would see alot of interference with 9 0r 10 APS in a 67 x 72 foot room unless you could adjust the cell size down so you would really need adjustable power and/or consider using directional patch antennas.

depending on the

thinking

Help......... in a room of that size

using a

The client adapter is going to try to connect to the strongest signal regardless of the SSID so this wont help in balancing.

If you have an AP that supports QoS and it is configurable for various traffic types you could assign VoIP to the lowest priority.

sufficient?

Not sure what your budget is, but it sounds like a wireless switch with thin APs is what you need, they self configure, are bandwidth controllable, auto power adjust to the least interference etc... but for a temporary situation where are you going to get all this equipment for a temporary situation?

through DHCP,

You really need static IPs or it could be a nightmare

Thanks Airhead,

I was afraid of that (the interference in the room) . The access points I'm looking at right now do have power control, and i've been trying to find out exactly what the available power levels are so I can try to figure out what the smallest footprint would be....

I have to plan for 250 users for the opening and closing days. it will most likely be less, but a good 90% will have laptops and all will be running. It will be a T3 (45 Mbps) into the building with a 3 meg DSL backup.

...I located the power control numbers. The APs can be set to 12.5% (10 dBm), 25% (12 dBm), 50% (15 dBm) and 100% (17 dBm). They also allow the transmit rate to be fixed at 11 Mbps (or any of the other levels) - that's supposed to limit the footprint as well, right?

Budget is one of my problems.... I would have to rent a more sophisticated set-up (like a switch and thin APs)...but it's considerably cheaper to buy a simpler setup like I'm contemplating than rent one for two weeks (same thing with projectors).

I've got a couple more questions so I can confirm I'm thinking about things the right way. The nominal speed of an access point (i.e. 11 Mbps) is the throughput PER CHANNEL, so that all the users connected to the AP share that total amount bandwidth. If that's correct, then:

a) plugging a single access point into a 10baseT lan connection shouldn't hurt the performance of the AP because the throughputs are about the same, but if I plugged more than one AP into a segment running at 10 Mbps, there would be an impact on performance of each AP because they would be sharing a smaller bandwidth connection.

b) the formula for calculating how much bandwidth each user would have (simultaneously) would look like ((actual AP throughput) x (number of APs)) / (number of users). *This requires the assumption that the users are evenly distributed between the APs.

*** users is known at 250 *** 9 access points *** 5 Mbps actual throughput per access point ==== 184.3 kbps per user (which is the same as the Internet connection)

Is this a reasonable way to calculate things? Is the estimated actual throughput reasonable considering the level of interference I will get? (I don't intend to use any security on the wireless link, and I've read using a short preamble can help - is that worth it?)

thanks again

You don't want to limit the data rate to 11Mbits/sec or slower. That will IMPROVE the range: |

Approximate Max Indoor Range

1 Mbps 350 ft. 2 Mbps 250 ft. 5.5 Mbps 180 ft. 6 Mbps 300 ft. 9 Mbps 250 ft. 11 Mbps 150 ft. 12 Mbps 200 ft. 18 Mbps 170 ft. 24 Mbps 140 ft. 36 Mbps 100 ft. 48 Mbps 95 ft. 54 Mbps 90 ft.

but also use more air time for the same amount of data moved. If a client wants to move XX Gigabloats of data, then the less air time he uses means more air time for others to use.

Diminishing returns will also be a problem. You can spend 10 times as much, and only improve the performance slightly. There is a point where you create your own interference. Anything beyond that causes a decrease in performance, not an increase.

Wrong. 11Mbits/sec is the connection speed. There's lots of overhead involved. Dragging out the chart from an earlier posting:

Modulation Max Max Max Channels ------- | Link TCP UDP | | | | |

802.11b 3 CCK 11 5.9 7.1 802.11g (with 802.11b) 3 OFDM/CCK 54 14.4 19.5 802.11g only 3 OFDM 54 24.4 30.5 802.11g turbo 1 OFDM 108 42.9 54.8 802.11a 13 OFDM 54 24.4 30.5 802.11a turbo 6 OFDM 108 42.9 54.8

If you lock the connection rate at 11Mbits/sec, you'll effectively lock the maximum thruput to about 5.5Mbits/sec (or less). Try to think in terms of gigabloats downloaded instead of dividing bandwidth. How much data are these users expected to download?

I don't understand where you're going. If you plug 100 access points into a single 10barfT connection, the aggregate bandwidth will still be 10Mbits/sec. Ignoring interference and different connection speed issues, the bandwidth will divide equally among the access points. However, the usage pattern will determine how many gigabloats an individual user can download. Some lunatic running Bitorrent can monopolize all your available bandwidth. You will need some form of bandwidth management, traffic control, user exclusion, and clueless user admin. You cannot deploy such a large network and not expect someone to show up with a virus, worm, brand new XP laptop requiring

100MBytes of updates, or other idiocy. Two weeks ago, I had to temporarily block the windowsupdate site for a convention in progress because about 150 laptops simultaneously decided that this would be a great time to do updates. An acquaintance spends his spare time maintaining a rather large convention center network. By necessity, it's an open (non-encrypted) network. Keeping the neighbors out of the system has been a running battle as they seem to think it's their personal alternative to DSL or cable modems. Think in terms of administration and control.

No way. Unless this is Defcon revisited, you are not going to have

250 users simultaneously using the system to its limits. You may have 250 users associated and connected to the various access points, but you are not going to have all of them simultaneously downloading the latest Victoria's Secret fashion show during the event. If this is a convention, the usual litany from the podium is "kindly turn off your laptops and pay attention".

Incidentally, watch out for DHCP server problems. 253 is the maximum number of clients on a Class C (CIDR /24 or 255.255.255.0) network. If you really have over 250 laptops in the area, you're going to run out of DHCP addresses. Make sure your single DHCP server has a backup and is running a larger number of available IP's.

No. The first number is a bad guess. You may have 250 connections, but if they're not moving data, they're not eating bandwidth. A good rule of thumb for user loading on a single access point is: 100 casual users checking email and light surfing 10 business users 1 file sharing user This doesn't strictly apply to your situation, but the idea is the same. It's *WHAT* the users are doing that determines available bandwidth. One user doing something disgusting will eat all your bandwidth unless you do something about it (i.e. exclusions and bandwidth management).

No. Interference is on an exponential curve with a fairly difficult to calculate threshold. Place one access point in a room, on a given channel, and measure thruput. Add another access point, on the same channel, but with a different SSID, and again measure thruput with both going at the same time. There will only be about a 10-20% loss from airtime collisions. 802.11 is designed to work like that, so that no system can monopolize the public airspace. However, as the number of access points and clients increases in a given airspace, the collision rate increases exponentially. My guess(tm) is about 5 access points and clients running in the same airspace, downloading furiously, will cause the system to be next to useless because all the packets transmitted will be re-transmissions.

What you'll probably find is that most of your interference does not come from access points. These can be engineered and power controlled to produce the least amount of interference. It's the client radios that will be trashing each other. There's no tx power control, no synchronization, usually no flow control, and often some very aggressive retry algorithms. If you will have perhaps 10 clients on a single access point, all watching the same streaming media content, the thruput will be so bad that the clients will timeout.

Sorta. You'll gain nothing in thruput. The long preamble is only required with ancient 802.11 (1-2Mbits/sec) cards, which are essentially extinct. You can leave it on "auto" because the access point will automagically use a short preamble for everything other than 1-2 Mbit/sec associations, which will hopefully not happen.

Incidentally, my acquaintance told me about a convention network that was working at a tolerable level, but received a few connectivity complaints. So, they more than doubled the number of access points in the building. Of course, they interfered with each other and then everybody had connection problems. Biggah ain't always bettah. In my never humble opinion and very limited experience with such things, a small number of isolated access points, with carefully located sector antennas, is much better than an oversized tangle of overlapping access points with omni antennas.

Good luck.

But if I fix the data rate to 11 Mbps, will it go below that?

I had hoped to understand it conceptually before dealing with numbers

lots (see your point about administration)

I think that's what I was trying to say....if i plugged x access points capable of 97 Mbps into a 10barfT connection, they (the x) would have to share 10 Mbps (which might be a bottleneck), but there wouldn't be a narrowing of the bandwidth -at that point- if the x access points were plugged into a 100baseT connection.

i figured that (administration and watching for crazy people) would be a big part of it.......

crap.....i'm not going to have to worry about subnets, am I? - or worse, is that going to max out the whole dog gamned thing?

Well I guess it depends on your definitions of "business user" and "casual user". It's not a conference, it's a working meeting, and there will be somewhere up to 250 (closer to 200) heavy surfing/email checking clients in a single room the first and last day. They'll be more spread out (physically) the other days.

i figured that (administration and watching for crazy people) would be a big part of it.......

while I've got you, can you suggest some good tools for that?

when you say "airspace" are you talking about geography or spectrum? because even my thoughts of 9 access points only ever had three operating on the same channel (and 20 users per)....it's odd there's no power control built into the client hardware, although I guess that would just make the access points more expensive.

yeah, i know........

thanks, I'll need it

Probably not. It really depends on the access point and how the configuration is scribbled. If there is a setting such as "802.11b only", it will work from 1-11Mbits/sec. If there is an "802.11g only" setting, it will only use the more efficient OFDM modulation and go from 6Mbits/sec to 54Mbits/sec. Some firmware has a table of acceptable speeds, where you can prevent the access point from going below 5.5Mbits/sec. However the greatest majority simply give you the choice of picking exactly one speed.

That's fine and what I also like to do. However, eventually, you'll need to deal with reality as expressed in numbers. For me, numbers are everything.

I don't wanna play voice of doom (no matter how much fun it is). However, methinks it would be a good idea to give you a feeling of what might be involved. I forgot to mention that you'll probably have a dilemma dealing with client to client traffic. For the most part, you don't want any client to client traffic. That's the way worms and virus's propagate. Same with hackers attacking other users. However, there's *ALWAYS" some irate user demanding to be able to use the convention network to move data from one laptop to another. The right way it to tell them to setup a peer-to-peer network and not bother the convention network, but that doesn't fly with corporate types used to having IT do everything for them.

That's correct. Simply having more wireless bandwidth than backhaul doesn't improve anyone download speed. Now, allow me to screw things up a bit. Pretend you have two users. One connects with an 802.11b radio at 11Mbits/sec, the other with an 802.11g radio at 54Mbits/sec. You have a 10barfT backhaul and are therefore limited to about

6Mbits/sec (surprise, you're probably not full duplex on the access point). Quiz time:

1. Duz the bandwidth divide equally between the two users at perhaps

3Mbits/sec each?

1. Or duz the airtime divide equally between the two users allowing the faster user to download more megabytes than the 11Mbit/sec user.
2. Or duz the presence of 802.11b cause the 802.11g connection to slow down to the speed of the 802.11b connection resulting in primate equal bandwidth?

I hope you enjoy playing policeman. Part of the fun is answering the inevitable question: Where is all the traffic coming from and how do I kill it? Be prepared to use Ethereal or something similar to identify abusers and infected machines.

How should I know? You haven't defined the hardware. One reason to use a central switch is that all the admin is centrally controlled. That includes the DHCP server. There are no roaming issues. You can enable and disable access points on the fly without much effort. It's totally ideal from the management point of view. Want more DHCP scope? No problem, it's one setting. You won't have any subnets because it will be one big network. Don't worry about subnets. However, you should worry about running out of DHCP leases. The sloppy way to solve that is to limit the DHCP lease time and rapidly expire stale leases. About 30 mins is all that's needed. When someone turns off their laptop, their IP address is available for someone else in 30 minutes. That will help, but really solves another problem caused by excessively huge MAC address to port tables and ARP tables. If you expect 250 users, then plan for about twice that number of MAC addresses.

I've never attended one of those but I've helped set them up. The weak link is the number of 117VAC outlets. Everyone turns on their laptops, sleeps through the opening speech while playing Solitaire, and then looks for a way to recharge their laptops. Have some big power strips ready.

I'm not sure what they'll actually be doing. If they're looking at an online version of the (yawn) PowerPoint presentation, it doesn't take hardly any bandwidth. Same with email. What burns bandwidth is streaming content and massive downloads. Someone shows up with a file sharing program active and eats all your outgoing bandwidth. Someone downloads something with Bitorrent and eats all your incoming bandwidth. Again, it's not the number of users or connections, but what they're doing.

I almost forgot. There are an awful large number of VoIP users. I use Skype whenever possible. They're gonna want a low latency connection with will be impossible if you're dealing with large amounts of traffic. However, if VoIP is important, you might wanna look into QoS (Quality of Service) features in various bandwidth management schemes. (I'll be you though this was gonna be easy).

It really depends on the audience. I play wireless policeman for a few local coffee shops. I can go for weeks without any problems. Then, some chronic abuser drifts in and I have to deal with their crap. Usually, it's some spammer sitting in the back of a vehicle in the parking lot, using the wireless to spew his junk, and make a quick getaway if there's any sign of problems.

For what? Crazy people? A large baseball bat usually works.

If you mean wireless, Snort works well for intrusion and abuse detection.

've only setup Snort on one corporate system so I'm not the expert on using or setting it up. You can grab pre-configured rules and get running immediately.

Geography. However, it also applies to spectrum in terms of channel selection. Basically, airspace is any area that is covered by wireless that might be a localized source of co-channel interference.

I'll spare you the politics. Every technology adopted since 802.11 has required adaptive tx power control.

Also, 3Scum has a line of wireless switches: |

don't know anything about it but it looks interesting (and expensive).

Note: If it were easy, it would be no fun.

I should have mentioned, all the gear is "b" only, and according to the documentation, it does allow the transmit rate to be fixed rather than a cap/floor setting

Much better to hear it now than on the hotel floor....much, much better.....

Quiz time:

Thankfully I'm limited to b equipment , so i won't have to think of these things on the ground, but doesn't g equipment revert to b for everything if a b device is detected? In a mixed group of users with a mix of b and g devices, is there actually a benefit to using g if you only ever get b performance?

sorry, more exclamation than question......but thanks for the answer. I was was already going to have the lease times reduced. And I will be working with a network admin (a real one, presumably) who belongs to the wired network. My worries are mostly related to getting the wireless to work while not killing with (wired) network.

Power is taken care of, it will be at the tables. It is something that gets overlooked, and then as you said,m people start running around for a place to recharge.

Actually, that rasies a question I had about channel re-use. This again goes to the "big room" days. Using the 3 non overlapping channels with 9 APs, there are 3 APs on each of the channels. Using four channels that overlap a little bit, and 8 APs, there are only two APs on each channel, but the channels casue some interference. I guess the question is "which is worse". Now that I'm actually writing it down, it would seem to require an actual interference study to anser that. I guess I'm looking not for "which is worse" but which actually works better in the field (ususally).

I don't know anything about it but it looks interesting (and

Thanks again. As I said, it's better to think of these things now than on the floor of the meeting.