Realworld Wireless speed and errors

[michaelhess]

I've got 4 1130ag's, three at one site and one at another.

I get just about 2Mbps over them. I've limited the AP's to only use G speeds per Cisco's docs, I've verified the ethernet port on the 3560's they plug into, they transfer at real 100Mbps Lan speeds, easily over 2Mbps Speakeasy test on my laptops over the wire is 8Mbps, over the AP's it's usually only 1.5Mbps.

I should get more than this with B! I'm using linksys wpc54gs ver 2 and Belking F5D7011 cards on around 20 laptops. Speed tests were taken on each individual AP with others powered off and no traffic to the two test machines (1 server and 1 laptop). The linksys card is slightly slower so I think it may be a driver/XP issue. I've downloaded all the latest drivers to no avail.

I get a LOT of errors in the ap stats screens:

Retries: 10408

Packets With One Retry: 6728

Packets With More Than One Retry: 1734

Duplicate Frames: 564

CRC Errors: 122

This is over 3 speakeasy tests and one 33MB file transfer test with the Belkin. There are no other wireless devices, phones, etc in the same band. Open air from about 10-15 feet from the AP's

Any ideas would be greatly appreciated.

[scottmac]

Load Netstumbler (www.netstumbler.com) on the laptop nad take a look at the RF environment you're woring in.

NS will give you an indication of the best channel (1, 6, or 11) is the cleanest ... environmental interference as well as other wireless systems.

What kind of signal strength, *AND* what kind of signal quality are you seeing? Of the two, signal quality is much more important.

The indications from the stats you posted suggest interference and / or poor signal quality.

Good Luck

Scott

[michaelhess]

That was my very first thought so I did use NS to take a look. I'm on channel 1, nothing else at all shows up. I also tried 11 and it was about 3-6 dB lower, no performance difference. Signals are:

Belkin:

SNR 68 (54 at chnl 11)

Noise -100

Signal -32 dBm (-46 at chnl 11)

Linksys:

SNR 47 (49-51 at chnl 11)

Noise -100

Signal -53 dBm (-47 at chnl 11)

[robert.wright]

a drop of 3db = half the strength

a drop of 6db = quarter the strength.

Do the following, login to the AP's CLI and issue the following command. (change to appropriate radio)

show int dot11radio 0 statistics

Copy the results to your fav text editor.

Then issue:

clear dot11 statistics dot11Radio 0

Now wait 10 minutes or so and issue the first command again... Post the results.

show int dot11radio 0 statistics

[michaelhess]

After:

DOT11 Statistics (Cumulative Total/Last 5 Seconds):

RECEIVER TRANSMITTER

Host Rx Bytes: 2097968 / 704 Host Tx Bytes: 2351077 /1079

Unicasts Rx: 7590 / 10 Unicasts Tx: 7006 / 10

Unicasts to host: 7590 / 10 Unicasts by host: 6905 / 10

Broadcasts Rx: 101 / 0 Broadcasts Tx: 13269 / 48

Beacons Rx: 0 / 0 Beacons Tx: 13033 / 48

Broadcasts to host: 101 / 0 Broadcasts by host: 235 / 0

Multicasts Rx: 0 / 0 Multicasts Tx: 1482 / 6

Multicasts to host: 0 / 0 Multicasts by host: 1478 / 6

Mgmt Packets Rx: 0 / 0 Mgmt Packets Tx: 101 / 0

RTS received: 0 / 0 RTS transmitted: 4 / 0

Duplicate frames: 7 / 0 CTS not received: 1 / 0

CRC errors: 701 / 0 Unicast Fragments Tx: 7003 / 10

WEP errors: 0 / 0 Retries: 1790 / 1

Buffer full: 0 / 0 Packets one retry: 770 / 1

Host buffer full: 0 / 0 Packets > 1 retry: 464 / 0

Header CRC errors: 0 / 0 Protocol defers: 178 / 0

Invalid header: 1 / 0 Energy detect defers: 454 / 0

Length invalid: 0 / 0 Jammer detected: 0 / 0

Incomplete fragments: 0 / 0 Packets aged: 0 / 0

Rx Concats: 0 / 0 Tx Concats: 0 / 0

Before:

DOT11 Statistics (Cumulative Total/Last 5 Seconds):

RECEIVER TRANSMITTER

Host Rx Bytes: 45681718 / 0 Host Tx Bytes: 590600614 / 315

Unicasts Rx: 295262 / 0 Unicasts Tx: 434310 / 0

Unicasts to host: 295262 / 0 Unicasts by host: 426141 / 0

Broadcasts Rx: 27845 / 0 Broadcasts Tx: 882486 / 49

Beacons Rx: 0 / 0 Beacons Tx: 875149 / 49

Broadcasts to host: 27845 / 0 Broadcasts by host: 7337 / 0

Multicasts Rx: 0 / 0 Multicasts Tx: 87408 / 6

Multicasts to host: 0 / 0 Multicasts by host: 87406 / 6

Mgmt Packets Rx: 0 / 0 Mgmt Packets Tx: 8169 / 0

RTS received: 0 / 0 RTS transmitted: 324 / 0

Duplicate frames: 2066 / 0 CTS not received: 292 / 0

CRC errors: 12252 / 0 Unicast Fragments Tx: 434262 / 0

WEP errors: 0 / 0 Retries: 64411 / 0

Buffer full: 0 / 0 Packets one retry: 26613 / 0

Host buffer full: 0 / 0 Packets > 1 retry: 18184 / 0

Header CRC errors: 0 / 0 Protocol defers: 21 / 0

Invalid header: 86 / 0 Energy detect defers: 6320 / 17

Length invalid: 0 / 0 Jammer detected: 0 / 0

Incomplete fragments: 0 / 0 Packets aged: 0 / 0

Rx Concats: 0 / 0 Tx Concats: 0 / 0

[robert.wright]

The only thing i really see out of the ordinary is your Energy detect defers. This is the number of times your AP detected radio transmissions and waited befoer sending data.

However this could have just been during higher activity on your wirelses network.

[john.preves]

Wow!

What kind of environment are you in? I'm guessing an airplane hanger with trees in it wrapped in chain link fence!

Seriously, the signals you have assuming a -100 dBm noise floor should be fine for voice and data in a clean environment.

I am assuming mulipath and that is what Scott was alluding to when he spoke of signal quality.

It is supposed to be a diversity antenna so that should be taken care of, but what is the power set to?

Please describe the environment. The walls floors, ceilings...office space, bank vault, Kryptonite factory, etc.

I would like to see the sh run and any results you got from a site survey.

Did it ever work or has it always been like this?

[michaelhess]

One location is a single story office building, not much around it except a lot of parking lots. On the outer edges of the building (outside) I pick up 3 other wireless networks with NS, but the signal is so low I can't even associate to them. The building is brick with typical wood studs in the inner walls. Nothing too weird. The other site (where the readings are from but almost the same as the other location) is a house in a residential district, no other ap's even when walking around the yard perimeter. House was built in the 1960's, again nothing strange about it.

The speeds were better when I first installed everything, they just suddenly dropped in December. Nothing has changed.

I have power output set all the way up. What do you mean "the sh run"?

[john.preves]

I'm sorry "sh run" is geek for 'show run' which is a command that you type in to the AP to show the configuration.

I would suggest turning the power down as much as possible, at least to test the throughput in a less "hot" environment.

However, if this doesn't show any marked improvement, you will have to test a new Known Good AP in this environment and test a Possibly Bad AP in a Known Clean Environment.

Something should reveal itself at that point; either the good one will work and you have crappy radios that need to be replaced or, the good one won't work in your current environment but the bad works fine outside of it.

Then you need to either get some serious spectrum analysis or move and call a priest.

[Rob Huffman]

Hi John,

It's always nice to see some humor along with your good ideas :) I was laughing pretty good on these last two posts. 5 points from this end! I think your troubleshooting ideas were also very good.

Take care,

Rob

[michaelhess]

Current configuration : 4552 bytes

version 12.3

no service pad

service timestamps debug datetime msec

service timestamps log datetime msec

service password-encryption

hostname -------

clock timezone UTC -7

clock summer-time UTC recurring

led display alternate

ip subnet-zero

ip name-server ---------

aaa new-model

aaa group server radius rad_eap

server --------- auth-port 1812 acct-port 1814

aaa group server radius rad_mac

aaa group server radius rad_acct

server --------- auth-port 1812 acct-port 1814

aaa group server radius rad_admin

server --------- auth-port 1812 acct-port 1814

cache expiry 1

cache authorization profile admin_cache

cache authentication profile admin_cache

aaa group server tacacs+ tac_admin

cache expiry 1

cache authorization profile admin_cache

cache authentication profile admin_cache

aaa group server radius rad_pmip

aaa group server radius dummy

aaa authentication login default local cache rad_admin group rad_admin

aaa authentication login eap_methods group rad_eap

aaa authentication login mac_methods local

aaa authorization exec default local cache rad_admin group rad_admin

aaa accounting network acct_methods start-stop group rad_acct

aaa cache profile admin_cache

all

aaa session-id common

dot11 ssid --------

authentication open

authentication key-management wpa

guest-mode

wpa-psk ascii 7 -----------------

dot11 aaa csid unformatted

dot11 network-map

bridge irb

interface Dot11Radio0

description 802.11G Radio

no ip address

no ip route-cache

encryption mode ciphers tkip

ssid ---------

speed basic-1.0 2.0 5.5 6.0 9.0 11.0 12.0 18.0 24.0 36.0 48.0 54.0

channel 2462

station-role root

no dot11 extension aironet

bridge-group 1

bridge-group 1 subscriber-loop-control

bridge-group 1 block-unknown-source

no bridge-group 1 source-learning

no bridge-group 1 unicast-flooding

bridge-group 1 spanning-disabled

interface Dot11Radio1

description 802.11A Radio

no ip address

no ip route-cache

encryption mode ciphers tkip

ssid ------------

speed basic-6.0 basic-9.0 basic-12.0 basic-18.0 basic-24.0 basic-36.0 basic-48.0 basic-54.0

station-role root

bridge-group 1

bridge-group 1 subscriber-loop-control

bridge-group 1 block-unknown-source

no bridge-group 1 source-learning

no bridge-group 1 unicast-flooding

bridge-group 1 spanning-disabled

interface FastEthernet0

description Ethernet Link

no ip address

no ip route-cache

duplex auto

speed auto

bridge-group 1

no bridge-group 1 source-learning

bridge-group 1 spanning-disabled

hold-queue 160 in

interface BVI1

ip address dhcp client-id FastEthernet0

no ip route-cache

ip http server

ip http authentication aaa

no ip http secure-server

ip http help-path http://www.cisco.com/warp/public/779/smbiz/prodconfig/help/eag

ip radius source-interface BVI1

snmp-server view dot11view ieee802dot11 included

snmp-server community public RO

snmp-server location --------------------

snmp-server contact Michael Hess

snmp-server chassis-id -----------

snmp-server enable traps snmp authentication linkdown linkup coldstart warmstart

snmp-server enable traps tty

snmp-server enable traps entity

snmp-server enable traps disassociate

snmp-server enable traps deauthenticate

snmp-server enable traps authenticate-fail

snmp-server enable traps dot11-qos

snmp-server enable traps switch-over

snmp-server enable traps rogue-ap

snmp-server enable traps wlan-wep

snmp-server enable traps config

snmp-server enable traps syslog

snmp-server enable traps cpu threshold

snmp-server enable traps aaa_server

snmp-server host ----------- public

radius-server attribute 32 include-in-access-req format %h

radius-server host ------------ auth-port 1812 acct-port 1814 key xxx

radius-server vsa send accounting

control-plane

bridge 1 route ip

wlccp wds aaa csid unformatted

line con 0

line vty 0 4

sntp broadcast client

end

[michaelhess]

I replaced my wpa-ent with wpa-personal to rule out radius issues. I also tried as completley open, no diff. I tried with various data rate settings as well, all on, all off, g only, etc. I removed the user/password section as I assume that won't matter. Thanks for all the great suggestions guys! I do have a couple bluetooth devices in the general area that wouldn't show on NS, no cordless phones in that band though. I've tried with all the bluetooth devices off and it didn't have any affect.

[john.preves]

just a couple of points...the G radio is set to 1.0 and up.

That should read speed basic-11.0 12.0 etc....

The channel for this AP is 11. If there are any microwave ovens around close to this AP, you will take a good hit when they turn it on.

Channel 11 is the closest one of the three to Channel 9 which is pretty close to the cenetr frequency of a standard microwave. It takes the worst hit.

I don't see any reason for guest mode to be in there, maybe I missed something.

There is no channel set for A.

I would at this point lower all my AP's down to like 10mW just to see what happens. I would do the same with all the client cards as well. If everything is at full power you are most likely causing your own multipath.

Ensure you are using three separate channels for G, (1,6,11)

I would then take a Card you know that Works Elsewhere and run some ping tests from the client to the AP and then to the Gateway and then to something on the network on the other side.

I would also get some Wireshark and sniff the air and see if anything catches your eye. Maybe one bad card is killing you..who knows.

If you have any (as in even one) B client the whole thing takes a dive. Since G wasn't around when they made B, the B clients cannot detect G traffic and the G traffic has to stop and make sure no B cards want to send traffic. You take a huge hit for that!

Not much else I can guess at. But at the very least rule out same channel interference by checking all your AP's, check your ping times with a continuous ping, make sure you are roaming properly and not sticking to one AP too long, No B clients - microwave locations,

wireshark the air and get a sniff and see what is dropping retransmitting etc...

Take the aluminum foil wall paper of the walls (haha) and remove the tinsel somebody draped on the AP for Christmas (Go ahead and giggle I have seen this in reality!)

And just for grins make sure there are no Draft -N products being used anywhere.

If you get past all that It would be Time To Spend Some Money and buy Cognio and perform a spectrum analysis and let it find any phones, x-10 cameras, who knows what else...

Good luck bro - I feel your pain..

[michaelhess]

The 1.0 rates where my last ditch effort to try everything, as was guest mode. I actually had the a radio off during my tests, just turned it back on.

I've killed all devices that emmit anything around the ap's. No microwaves. No B devices. Didn't know ethereal changed it's name, had me stumped there for a minute, thank you google, I'll try that tomorrow. I've tried with two known good cards and the ap plugged straight into a cleanly reset firewall with nothing else on it. I killed all speeds but 54 and still bad.

I will lower the power tomorrow and see what happens. If that doesn't fix it I may get Cisco to RMA one of them and see what happens. I may also take the ap, two laptops and a couple of cards to the middle of nowhere (Wyoming has a lot of that) with one of my larger ups' and see what happens. :)

No way am I going to get the board to approve buying Cognio or any of it's brethren, tried that last year with wildpackets, non-profit with donated Cisco stuff.

[robert.wright]

Netstumbler interoperability with certain cards cause netstumbler to report a noise of -100 and never changes.. This is not the true reading and should not be used to judge an environment.

Also should note the command 'show dot11 ass all'.. To get what the AP is seeing in terms of SNR and sig strength.