I am currently installing a wireless network in a school using PoE and dlink access points. These are DWL-2100+ (APs) and DWL-P100 (PoEs). There are around 18 APs at the moment.
The layout is like this...
There are two buildings, East and West.
East is L-Shaped with the longest part having three floors.
West is three floors high for a majority of its coverage.
The problem i am having is with the coverage for these buildngs. East sems to be stable at the moment but i won't hold my breath...West on the other hand is a nightmare.
The APs seem fine but for unknown reasons just fail and cease to respond to anything even the web interface.Its not just the odd one however. There are around 6-7. A Dlink consultant came in to do a site survey and suggested around 11 to cover both buildings and also said that the PoEs would be fine for these APs. There are also no new firmwares for my APs from what i can gather from their website.
At the moment i am running basic WEP (will be upping security once it is up and running properly). All APs are set to same SSID and each floor has usually two to three APs as the corridors are long.
There can be up to around 11 members of Staff using the one AP at any one time maybe more..it depends on who connects to what AP. The channels are all set as far apart as possible to avoid confrontation. I have also tried to pull back the signal broadcast strength as to avoid further interference but they still overlap from floor to floor i.e. according to WZC top floor APs can still be seen.
All i need help on is to figure out why the APs keep failing as this creates large gaps in the wireless network because as i said before the corridors are long.
There's no easy answer to your question and the end result of a good site survey isn't "around 11" AP's to cover an area. Also WZC is not a good tool to base signal coverage on. Use netstumbler and look for a signal-to-noise ratio of 23 db or higher for your coverage area. Take into consideration the location of PC's and/or Laptops relative to the location of the AP's and objects that may affect the signal, like people, metal cabinets, etc.
It could be interference, could be those simulated PoE adapters, or it could be user load. How far are the Cat. 5 runs to those AP's?
If the AP's have Syslog or SNMP capabilities, I would start there and see what activity occurs prior to the lock-ups.
If the APs are still broadcasting an SSID when they lock-up, look for interference.
What do you do to the ap's to get them running again? Do you power cycle them, or do they just start working again on their own? Try disabling WEP and see if the problem still occurs. I have run into problems with some dlink ap's locking up using wpa. The ap will be fine until a client attempts to connect without using wpa... For some reason this locks up the ap, and I have to unplug it to get it to respond.
I do not think I said that. May have in the wrong context. I do get concerned with the fact that it is a subjective reading being very dependent on how the program reacts with the wireless network adapter that is being polled.
The problem you are seeing is that the signal portion is ovewhelming the noise aspect. Kind of like holding fish closer to the camera to make it bigger.
Also to measure noise or intereference would be better handled by a spectrum analyzer, but in this case we are just concerned about the ratio. The better the ratio is, the easier it is for the receiver to distinguish usefull digital information from the clutter.
Without spending thousands on software applications for a single install, netstumbler would be better than WZC and that's the only reason it was suggested. Signal strength alone is not a very good measurement of performance.
Most wireless adapters have software tools. If yours can show packet retries (i.e. Cisco Aironet 350 b-only card), another easy test is to start a large download from www.testmy.net, map various stationary points around the access point (don't walk around during the test), and look for a packet retry percentage of 5 to 7. This will also show you how easily noise or multi-pathing can affect bandwidth. It's just another simple test that's better than WZC.
Thanks for the responses. I am still trying to get my head around the whole wifi thing. It's just my Boss wishes me to roll out wireless coverage across the school.
11 APs were chosen as this was the amount the D-Link consultant suggested would be needed to cover the areas. When he did the "Site survey" he positioned an access point in one of the rooms and pinged the ap constantly until it started to lose packets the further away he got. Which he then noted down as the maximum distance that the ap could be seen / used, and so on.
This site survey no longer applies however as the positions that he chose were not sufficent enough to cover classrooms. For security reasons the APs had been placed under a suspended ceiling (in corridors) where there are water pipes and metal beams. Unfortunately i cannot avoid any of these as they run the whole length of the corridors.
WEP encryption? I thought that due to the overload that basic WEP (no WPA at the moment) could have that when multiple clients are authenticating at once (around the same time in the morning all clients would/should be logging on) that it could possibly cause the APs to fail. The only way to get them back up is to power cycle them. Which when in the second block is a pain.
I have one AP that has never had this problem. It server all three floors (just) and runs WEP and also powers over ethernet. The only thing that is different is the fact that it is the only one in the vicinity. The next nearest access point is out of range...which suggests no interference with other APs.
Have you tried disabling WEP just temporarily to see if the problem persists? Also, do the aps lock up once a day, week, month? When the aps lock up, do they all do it at the same time, or is it random?
The APs can lock up at random times some can lock up about an hour later. After power-cycling it will then stay up for a bit but not even to the extent of a whole day.
The syslog does exsist but i have no idea on how it works. I have downloaded something called Kiwi syslog service manager and set the IP address of my machine as the syslog server but nothing much seems to happen. Not entirely sure on how to set up my machine to receive syslog messages. No firewall is turned on so it cant be being blocked.
Any ideas? I would like to see some sort of log as when power-cycled the event log gets cleared.
The one thing that is wierd though is that the problem is always with the same APs. Sometimes a different one goes down but generally always the same ones.
I have not yet disabled WEP encryption. This will be my next step. The APS so far have been stable for a bit, longer than two days, so i will monitor them closely. If anyone has any comments on the syslog thing that would be great thanks.
i have sorted the syslog app out thanks. just haven;t had chance to update the sit. However only get a response from the AP situated in the same office...
on checking the logs of the access points that are working the mac addresses listed as being connected are constantly re-connecting every 3-4 minutes. this i think is something to do with wzc as i read somewhere that it keeps looking for a better deal as it were.
the problem is that the dlink card util isn't much better. half of the time it doesn't even detect the card is present. the other half the util doesn't even start.
are there any other alternatives?
It will need to be rolled out across 20-30 laptops...with not the best users in the world at the keyboard.
First, I am dismayed that Dlink and "big wireless project" are in the same sentence.
Having said that, I have seen something very similar in Cisco AP's a firmware revision or two ago.
In version JA1, the buffers would fill up and then choke the AP. I don't remember exactly why, but version JA2 fixed it so I probably stopped worrying about it. The only way to fix it was to pull the jumper out of the switch and reboot, since you could no longer get to it via telnet etc...
No one thing specifically killed them and it happened whenever it felt like it anywhere it wanted. Sometimes several AP's would die on the same switch, while others still worked.
A Cisco engineer found it when we sent him the text files of the complete status reports from the AP.
It was like "show run" and "show version" and a few other show somethings all in one command...that I can't remember.
The buffer overruns had maxed out and in order to upgrade the Firmware, we had to manually increase the buffer so the firmware had time to take or it turned into a brick.
I have no idea what your problem is but it conjured up some flashbacks....
How would zero length IPs affect the status of wireless access points?
I have read that attackers can send them to bring down firewalls but these are being generated by the wireless cards (client side) themselves according to the syslog events being generated by ther access points. I have just this minute had one go down while i was trying to configure the card.
I would assume that it would have the same consequences.
I have since seen more of these zero length IPs coming from the cards and the APs are still up...
But yesterday i had an AP that has been solid since implemented around october last year crash. This i thought was strange even though i have had this problem in other areas, this one was a first. I decide to ask the users what they were doing last night. One was printing a big document and it failed to print.
Is this the sort of thing that you was experiencing?
I know that a few Routers have / had this problem.
I also get failed attempts at getting an IP from the DHCP server as well even though the Wireless connection is there.