High ping times with 11 CPE's on RB532...

jd6strings · Fri Jun 30, 2006 10:31 pm

We have a RB532 running version 2.9.23 configured as an AP Bridge and we have 11 CPE's (RB112's) configured as Station WDS. Ping times to client routers are sporadically jumping into the 300-400 ms range with occasional request time outs. The AP processor utilization is around 20%. The frequency utilization in our area is minimal and interference id not a problem. Any ideas why the ping times would be so bad? It has progressively gotten worse as we've added more CPE's to the AP.

maxfava · Mon Jul 03, 2006 12:46 pm

I'm encountering the identical issue with our RB532.
I'm using 1 Senao card and 1 CM9 card, both bridge with one ethernet interface.
During the long ping, it could append on CM9 as on Senao the signal is good for clients and ack is 56 us.
The only different that I have notice is the ack of the card, that is about 200us.
I have set dinamic ack with threshould of -80.
after 1 min that the clients ping 400 ms the ping return to 1 ms as normal work. Some time the long ping occours every 10 min sometime every 2 min.

Have you verify or found a solution to the issue long pings?

durim · Mon Jul 03, 2006 9:05 pm

We have a RB532 running version 2.9.23 configured as an AP Bridge and we have 11 CPE's (RB112's) configured as Station WDS. Ping times to client routers are sporadically jumping into the 300-400 ms range with occasional request time outs. The AP processor utilization is around 20%. The frequency utilization in our area is minimal and interference id not a problem. Any ideas why the ping times would be so bad? It has progressively gotten worse as we've added more CPE's to the AP.

I had the same problem with on of RB230 with about 8 CPE's conected and a load of 4-6 Mbps, the Nstream with Polling was a solutoin but it might be a problem with cpu load, depending on how much traffic do you have on your Ap but I sugest you give a try.

jd6strings · Tue Jul 04, 2006 2:00 am

Thanks for the reply.

We did in fact enable nstreme/polling and the ping times are still pretty ridiculous. Anywhere from 16ms to 249ms.

I added a firewall filter to see if P2P traffic was causing grief and 0 packets were caught by the filter so I know that can be taken out of the picture. CPU utilization is staying around 35% pretty consistently.

Our ack timeout for wlan1 stays at 25us. Shouldn't this be dynamically changing based upon the client that is the longest distance? I'm assuming this is now a timing issue...????

jd6strings · Tue Jul 04, 2006 3:23 am

Here is a screenshot of the sporadic ping times we're seeing with the above mentioned config. Just upgraded the rb532 to v2.9.27 to see if software was a potential problem candidate...issue still exists.

Any ideas?

http://www.skywerx.com/ping.jpg

jarosoup · Tue Jul 04, 2006 3:33 am

Your basic-rates should not be changed...set it back to 6Mb.

Have you used torch and looked at the firewall connections to see what's going on on your network? Sounds like you are either overloaded with traffic, or you've got signal (interference or hardware) problems. Other things to try are setting periodic-calibration=enabled and disable default-forwarding. Are you throttling your traffic?

jd6strings · Tue Jul 04, 2006 4:18 am

Network is clean. Small subnets routed through PC based MT and then on to Cisco 3640 edges. Torch on the wireless AP does not indicate anything out of the ordinary.

We have disabled default-forwarding and enabled periodic-calibration and still no change. I starting to suspect a potential defective CM9 card. Any other ideas appreciated....

jd6strings · Tue Jul 04, 2006 6:46 am

Yes...we are throttling in the sense that we are setting AP Tx limit and Client Tx limits in the ACL....

durim · Tue Jul 04, 2006 10:09 am

Can you post what signal strengh, do you have with clients. your ap load your wireless configuration also have you tried different frame policy it works different in different scenaraios.

Regard Durim

jd6strings · Tue Jul 04, 2006 4:54 pm

Here is the reg table:

# INTE... RADIO-NAME MAC-ADDRESS AP SIGNAL... TX-RATE
0 wlan1 X0 00:0B:6B:4D:77:A7 no -78dBm... 36Mbps
1 wlan1 X1 00:0B:6B:4D:77:E7 no -75dBm... 36Mbps
2 wlan1 X2 00:15:6D:10:22:68 no -73dBm... 24Mbps
3 wlan1 X3 00:0B:6B:4E:CF:0D no -78dBm... 36Mbps
4 wlan1 X4 00:15:6D:10:21:F3 no -68dBm... 36Mbps
5 wlan1 X5 00:0B:6B:4D:77:97 no -77dBm... 24Mbps
6 wlan1 X6 00:0B:6B:4D:77:3C no -76dBm... 24Mbps
7 wlan1 X7 00:0B:6B:36:0C:D4 no -75dBm... 36Mbps
8 wlan1 X8 00:0B:6B:4D:91:3D no -87dBm... 12Mbps
9 wlan1 X9 00:0B:6B:4D:77:55 no -76dBm... 36Mbps
10 wlan1 X11 00:0B:6B:4E:CF:C2 no -79dBm... 24Mbps
11 wlan1 X12 00:0B:6B:4D:95:B6 no -77dBm... 36Mbps
12 wlan1 X13 00:0B:6B:4E:D4:09 no -77dBm... 6Mbps

Currently Framer Policy is set to none with a Framer Limit being 3200. Whay would be a recommended policy???

I was reading a another post somewhere here on the forum in regards to bridging potentially causing high ping times. Could this also be the issue? I wouldn't think so with only 12 registered clients...

jonbrewer · Tue Jul 04, 2006 10:34 pm

Thanks for the reply.

We did in fact enable nstreme/polling and the ping times are still pretty ridiculous. Anywhere from 16ms to 249ms.

I added a firewall filter to see if P2P traffic was causing grief and 0 packets were caught by the filter so I know that can be taken out of the picture. CPU utilization is staying around 35% pretty consistently.

Our ack timeout for wlan1 stays at 25us. Shouldn't this be dynamically changing based upon the client that is the longest distance? I'm assuming this is now a timing issue...????

Until you get out a spectrum analyzer and prove otherwise, I would assume your high ping time is due to interference from another system.

jd6strings · Wed Jul 05, 2006 6:38 am

Frequency scan show 0 usage.

jober · Wed Jul 05, 2006 8:19 am

Here is the reg table:

# INTE... RADIO-NAME MAC-ADDRESS AP SIGNAL... TX-RATE
0 wlan1 X0 00:0B:6B:4D:77:A7 no -78dBm... 36Mbps
1 wlan1 X1 00:0B:6B:4D:77:E7 no -75dBm... 36Mbps
2 wlan1 X2 00:15:6D:10:22:68 no -73dBm... 24Mbps
3 wlan1 X3 00:0B:6B:4E:CF:0D no -78dBm... 36Mbps
4 wlan1 X4 00:15:6D:10:21:F3 no -68dBm... 36Mbps
5 wlan1 X5 00:0B:6B:4D:77:97 no -77dBm... 24Mbps
6 wlan1 X6 00:0B:6B:4D:77:3C no -76dBm... 24Mbps
7 wlan1 X7 00:0B:6B:36:0C:D4 no -75dBm... 36Mbps
8 wlan1 X8 00:0B:6B:4D:91:3D no -87dBm... 12Mbps
9 wlan1 X9 00:0B:6B:4D:77:55 no -76dBm... 36Mbps
10 wlan1 X11 00:0B:6B:4E:CF:C2 no -79dBm... 24Mbps
11 wlan1 X12 00:0B:6B:4D:95:B6 no -77dBm... 36Mbps
12 wlan1 X13 00:0B:6B:4E:D4:09 no -77dBm... 6Mbps

Currently Framer Policy is set to none with a Framer Limit being 3200. Whay would be a recommended policy???

I was reading a another post somewhere here on the forum in regards to bridging potentially causing high ping times. Could this also be the issue? I wouldn't think so with only 12 registered clients...

Have you tryed to uncheck the supported rates of 54,48,36,24 and 18?
I only say this because I had a similer problem but with some older B clients in the mix and I had to set the mobe to B only on one of the APs also.

PS, The speed rate saw a tip from Sten.

durim · Wed Jul 05, 2006 10:52 am

Here is the reg table:

# INTE... RADIO-NAME MAC-ADDRESS AP SIGNAL... TX-RATE
0 wlan1 X0 00:0B:6B:4D:77:A7 no -78dBm... 36Mbps
1 wlan1 X1 00:0B:6B:4D:77:E7 no -75dBm... 36Mbps
2 wlan1 X2 00:15:6D:10:22:68 no -73dBm... 24Mbps
3 wlan1 X3 00:0B:6B:4E:CF:0D no -78dBm... 36Mbps
4 wlan1 X4 00:15:6D:10:21:F3 no -68dBm... 36Mbps
5 wlan1 X5 00:0B:6B:4D:77:97 no -77dBm... 24Mbps
6 wlan1 X6 00:0B:6B:4D:77:3C no -76dBm... 24Mbps
7 wlan1 X7 00:0B:6B:36:0C:D4 no -75dBm... 36Mbps
8 wlan1 X8 00:0B:6B:4D:91:3D no -87dBm... 12Mbps
9 wlan1 X9 00:0B:6B:4D:77:55 no -76dBm... 36Mbps
10 wlan1 X11 00:0B:6B:4E:CF:C2 no -79dBm... 24Mbps
11 wlan1 X12 00:0B:6B:4D:95:B6 no -77dBm... 36Mbps
12 wlan1 X13 00:0B:6B:4E:D4:09 no -77dBm... 6Mbps

Currently Framer Policy is set to none with a Framer Limit being 3200. Whay would be a recommended policy???

I was reading a another post somewhere here on the forum in regards to bridging potentially causing high ping times. Could this also be the issue? I wouldn't think so with only 12 registered clients...

I use the frame policy = best fit with framer limit 3200 (2200) but I would suggest to try all four framer policy, as jober sad try lowering supported data rates to 18MBps or 12MBps

maxfava · Wed Jul 05, 2006 12:23 pm

Are you using Nstreme protocol also with standard WIFI client?

jd6strings · Wed Jul 05, 2006 2:34 pm

Are you using Nstreme protocol also with standard WIFI client?

No...simply PTMP subscribers connected to our network...

sten · Wed Jul 05, 2006 2:44 pm

i only heard this from someone else who tried to use wds to each customer;
dont. You add latency for each wds interface added.

maxfava · Wed Jul 05, 2006 2:55 pm

I use the frame policy = best fit with framer limit 3200 (2200) but I would suggest to try all four framer policy, as jober sad try lowering supported data rates to 18MBps or 12MBps

Currently Framer Policy is set to none with a Framer Limit being 3200. Whay would be a recommended policy???

No...simply PTMP subscribers connected to our network...

Framer policy is used only with Nstreme protocol.

rpingar · Wed Jul 05, 2006 4:21 pm

i only heard this from someone else who tried to use wds to each customer;
dont. You add latency for each wds interface added.

I use wds for each customers. And I have the same strange problem about long ping.
But I use PPPoE so to bridge my customers the manual says I have to use wds.
May you explain howto use pppoe for my customers (that are all MT based) without using wds?

Regards
Rosario

jd6strings · Wed Jul 05, 2006 6:27 pm

We changed our data rates and it did improve the ping times a bit. Still sporadic and I think that 30ms is still high for such a small network.

I guess routing is the next step BUT I already have MT PC based routers throughout the network and then Cisco edge routers....sooo adding another "router" into picture in my mind simply adds another unnecessary hop in the network.

The way that it is currently configured...the AP that is giving us grief is connected to the same switch that the MT PC based router is connected to and the clients use the PC based router as the gateway through the bridge that is configured in the AP.

IF I was to use "routing" in the AP, wouldn't I simply set up a default gateway to the PC based router and then in the CPE set up a default gateway to the AP's IP address OR would I have to let the AP "route" an entire subnet hence adding another hop?

Thanks for the replys!!!

jarosoup · Wed Jul 05, 2006 7:03 pm

I don't think bridging has anything to do with your problem. We use WDS for a lot of networks, and they typically work well if interference is low, we have good signal, we use good pigtails and antennas, and do bandwidth control to reduce flooding the APs. Here's a ping from a gateway to an AP 3 WDS hops away [GW] -- [AP1] <--> [AP2] <--> [AP3] <--> [AP4] where Gatway=GW and the AP it pings is AP4 - each connected to the other via WDS bridging:

25 packets transmitted, 25 packets received, 0% packet loss
round-trip min/avg/max = 5/6.7/15 ms

This is on 5GHz to Routeboards with 2 radios (the other is 2.4GHz).

Yes...we are throttling in the sense that we are setting AP Tx limit and Client Tx limits in the ACL

I wonder if this might be causing some of your problems? Have you tried removing these settings and setting up queues on your gateway instead. Seems like this would add some overhead to the AP, especially if it were to become busy. I've never set tx/rx limits on the AP itself, so I'm not sure how well this works. Might be something to try at least temporarily.

jd6strings · Wed Jul 05, 2006 7:39 pm

I don't think bridging has anything to do with your problem. We use WDS for a lot of networks, and they typically work well if interference is low, we have good signal, we use good pigtails and antennas, and do bandwidth control to reduce flooding the APs. Here's a ping from a gateway to an AP 3 WDS hops away [GW] -- [AP1] <--> [AP2] <--> [AP3] <--> [AP4] where Gatway=GW and the AP it pings is AP4 - each connected to the other via WDS bridging:
Code: Select all
25 packets transmitted, 25 packets received, 0% packet loss
round-trip min/avg/max = 5/6.7/15 ms
This is on 5GHz to Routeboards with 2 radios (the other is 2.4GHz).

Yes...we are throttling in the sense that we are setting AP Tx limit and Client Tx limits in the ACL
I wonder if this might be causing some of your problems? Have you tried removing these settings and setting up queues on your gateway instead. Seems like this would add some overhead to the AP, especially if it were to become busy. I've never set tx/rx limits on the AP itself, so I'm not sure how well this works. Might be something to try at least temporarily.

Removed the AP Tx and Client Tx limits from all associated CPE's and still horrible ping times.

jarosoup · Wed Jul 05, 2006 7:58 pm

Do you have all of your ack-times set to dynamic (they should be if your clients are at different distances)? I can't easily find the frequency you are running...is it 5GHz or 2.4GHz (a or g mode?)? Also, what antennas are you using?

jd6strings · Wed Jul 05, 2006 8:21 pm

Do you have all of your ack-times set to dynamic (they should be if your clients are at different distances)? I can't easily find the frequency you are running...is it 5GHz or 2.4GHz (a or g mode?)? Also, what antennas are you using?

Ack Timeout on the AP is set to dynamic. We are running in 5GHz with a single 120 degree 16 dbi sector.

jarosoup · Wed Jul 05, 2006 8:48 pm

Hmmm. Can you post the output from "/interface wireless reg print stat"? Maybe just include 2 or 3 of the problematic clients from this output as all of them will be too much. Also, I'm not sure if you've mentioned this, but have you tried other channels? Your ping output you posted shows packet sizes of 50. What does it look like with a size of 1500? Better, worse, or the same?

sten · Wed Jul 05, 2006 8:58 pm

Here is the reg table:

# INTE... RADIO-NAME MAC-ADDRESS AP SIGNAL... TX-RATE
0 wlan1 X0 00:0B:6B:4D:77:A7 no -78dBm... 36Mbps
1 wlan1 X1 00:0B:6B:4D:77:E7 no -75dBm... 36Mbps
2 wlan1 X2 00:15:6D:10:22:68 no -73dBm... 24Mbps
3 wlan1 X3 00:0B:6B:4E:CF:0D no -78dBm... 36Mbps
4 wlan1 X4 00:15:6D:10:21:F3 no -68dBm... 36Mbps
5 wlan1 X5 00:0B:6B:4D:77:97 no -77dBm... 24Mbps
6 wlan1 X6 00:0B:6B:4D:77:3C no -76dBm... 24Mbps
7 wlan1 X7 00:0B:6B:36:0C:D4 no -75dBm... 36Mbps
8 wlan1 X8 00:0B:6B:4D:91:3D no -87dBm... 12Mbps
9 wlan1 X9 00:0B:6B:4D:77:55 no -76dBm... 36Mbps
10 wlan1 X11 00:0B:6B:4E:CF:C2 no -79dBm... 24Mbps
11 wlan1 X12 00:0B:6B:4D:95:B6 no -77dBm... 36Mbps
12 wlan1 X13 00:0B:6B:4E:D4:09 no -77dBm... 6Mbps

Currently Framer Policy is set to none with a Framer Limit being 3200. Whay would be a recommended policy???

I was reading a another post somewhere here on the forum in regards to bridging potentially causing high ping times. Could this also be the issue? I wouldn't think so with only 12 registered clients...
Have you tryed to uncheck the supported rates of 54,48,36,24 and 18?
I only say this because I had a similer problem but with some older B clients in the mix and I had to set the mobe to B only on one of the APs also.

PS, The speed rate saw a tip from Sten.

Jober: What i tried to say was:

supported-rates-a/g=24Mbps basic-rates-a/g=24Mbps

jd6strings · Wed Jul 05, 2006 9:02 pm

Hmmm. Can you post the output from "/interface wireless reg print stat"? Maybe just include 2 or 3 of the problematic clients from this output as all of them will be too much. Also, I'm not sure if you've mentioned this, but have you tried other channels? Your ping output you posted shows packet sizes of 50. What does it look like with a size of 1500? Better, worse, or the same?

Here's output for the first 2 CPE's:

0 interface=wlan1 radio-name="X0" mac-address=00:0B:6B:36:0C:D4
ap=no wds=yes rx-rate=12Mbps tx-rate=12Mbps packets=46323,28394
bytes=28836890,3724565 frames=34669,22744 frame-bytes=28852884,3692360
hw-frames=52920,250228 hw-frame-bytes=32111051,14755386 uptime=1h29m46s
last-activity=30ms signal-strength=-75dBm@6Mbps signal-to-noise=31dB
strength-at-rates=-75dBm@6Mbps 30ms,-75dBm@9Mbps 15m45s740ms,-75dBm@12Mbps
50ms
tx-signal-strength=-78dBm tx-ccq=36% rx-ccq=55% p-throughput=11315
nstreme=yes framing-mode=exact-size framing-limit=2200
routeros-version="2.9.24" last-ip=10.10.0.149 802.1x-port-enabled=yes
authentication-type=none encryption=none compression=no

1 interface=wlan1 radio-name="X1" mac-address=00:0B:6B:4D:77:3C
ap=no wds=yes rx-rate=12Mbps tx-rate=12Mbps packets=86208,38129
bytes=22797932,2155944 frames=75494,38561 frame-bytes=22490196,1971946
hw-frames=109330,287525 hw-frame-bytes=27715509,14327438 uptime=1h29m38s
last-activity=30ms signal-strength=-76dBm@6Mbps signal-to-noise=30dB
strength-at-rates=-76dBm@6Mbps 30ms,-75dBm@9Mbps 15s520ms,-76dBm@12Mbps
30ms
tx-signal-strength=-77dBm tx-ccq=35% rx-ccq=54% p-throughput=10410
nstreme=yes framing-mode=exact-size framing-limit=2200
routeros-version="2.9.24" last-ip=10.10.0.34 802.1x-port-enabled=yes
authentication-type=none encryption=none compression=no

Ping times are sproadic even with a packet size of 1500...between 5ms and 80ms. Limiting the data rates at least brought them below 200 and 300ms. If I change channels, the clients SHOULD reassociate based on the scan list correct?

Thx for your input!

sten · Wed Jul 05, 2006 9:09 pm

How much traffic is on the link while you test?

Yes to question about searching according to scan-list.

jd6strings · Wed Jul 05, 2006 9:16 pm

How much traffic is on the link while you test?

Yes to question about searching according to scan-list.

Bridge interface is passing no more than 768 Kbps as I type this. Most customers are residential and most likely currently at work. Even now with the traffic minimal, here's ping output to one of the CPE's:

00:0B:6B:4D:77:E7 64 byte ping time=35 ms
00:0B:6B:4D:77:E7 64 byte ping time=32 ms
00:0B:6B:4D:77:E7 64 byte ping time=34 ms
00:0B:6B:4D:77:E7 64 byte ping time=21 ms
00:0B:6B:4D:77:E7 64 byte ping time=44 ms
00:0B:6B:4D:77:E7 64 byte ping time=26 ms
00:0B:6B:4D:77:E7 64 byte ping time=16 ms
00:0B:6B:4D:77:E7 64 byte ping time=11 ms
00:0B:6B:4D:77:E7 64 byte ping time=30 ms
00:0B:6B:4D:77:E7 64 byte ping time=42 ms
00:0B:6B:4D:77:E7 64 byte ping time=50 ms
00:0B:6B:4D:77:E7 64 byte ping time=18 ms
00:0B:6B:4D:77:E7 64 byte ping time=21 ms
00:0B:6B:4D:77:E7 64 byte ping time=63 ms
00:0B:6B:4D:77:E7 64 byte ping time=68 ms
00:0B:6B:4D:77:E7 64 byte ping time=61 ms
00:0B:6B:4D:77:E7 64 byte ping time=30 ms
00:0B:6B:4D:77:E7 64 byte ping time=49 ms
00:0B:6B:4D:77:E7 64 byte ping time=48 ms
00:0B:6B:4D:77:E7 64 byte ping time=48 ms
00:0B:6B:4D:77:E7 64 byte ping time=48 ms
00:0B:6B:4D:77:E7 64 byte ping time=15 ms
00:0B:6B:4D:77:E7 64 byte ping time=32 ms

?????? Weird huh?

jarosoup · Wed Jul 05, 2006 10:06 pm

From the 2 clients:
tx-ccq=36% rx-ccq=55%
tx-ccq=35% rx-ccq=54%

While not horrible, these should be a little higher for a solid connection. What type and gain are the antennas you are using on the CPE side? There seems to be a weakness on the CPE side as there is quite a difference between the client and the AP. If you look at the registration table on the CPE, is it similar to what you see on the AP? I'd try out a few different channels and see if your signals and the ccq values change at all.

As a side note and slightly off-topic, consider the values of output posted:

1 interface=wlan1 radio-name="X1"

uptime=1h29m38s

strength-at-rates=-76dBm@6Mbps 30ms,-75dBm@9Mbps 15s520ms,-76dBm@12Mbps 30ms

Could someone explain why the link uptime is 1h29m while all of the combined transmit rate times are around 15 minutes? I've noticed this before, and don't really understand these as it seems they should be almost the same. Perhaps they mean something different?

sten · Wed Jul 05, 2006 10:37 pm

i only heard this from someone else who tried to use wds to each customer;
dont. You add latency for each wds interface added.
I use wds for each customers. And I have the same strange problem about long ping.
But I use PPPoE so to bridge my customers the manual says I have to use wds.
May you explain howto use pppoe for my customers (that are all MT based) without using wds?

Regards
Rosario

perhaps you should let the cpe do the pppoe ?

sten · Wed Jul 05, 2006 10:40 pm

Could someone explain why the link uptime is 1h29m while all of the combined transmit rate times are around 15 minutes? I've noticed this before, and don't really understand these as it seems they should be almost the same. Perhaps they mean something different?

tx rate times are times since last use of that tx rate i guess.

jd6strings · Wed Jul 05, 2006 10:40 pm

From the 2 clients:
tx-ccq=36% rx-ccq=55%
tx-ccq=35% rx-ccq=54%
While not horrible, these should be a little higher for a solid connection. What type and gain are the antennas you are using on the CPE side? There seems to be a weakness on the CPE side as there is quite a difference between the client and the AP. If you look at the registration table on the CPE, is it similar to what you see on the AP? I'd try out a few different channels and see if your signals and the ccq values change at all.

As a side note and slightly off-topic, consider the values of output posted:
1 interface=wlan1 radio-name="X1"
uptime=1h29m38s

strength-at-rates=-76dBm@6Mbps 30ms,-75dBm@9Mbps 15s520ms,-76dBm@12Mbps 30ms
Could someone explain why the link uptime is 1h29m while all of the combined transmit rate times are around 15 minutes? I've noticed this before, and don't really understand these as it seems they should be almost the same. Perhaps they mean something different?

We're using Rootenna cases with 19dbi integrated for the CPE's. Here's output:

0 interface=wlan1 radio-name="Oakbrush AP" mac-address=00:0B:6B:37:B5:20
ap=yes wds=yes rx-rate=12Mbps tx-rate=6Mbps packets=229521,227521
bytes=66677985,95310269 frames=103946,119077 frame-bytes=67389829,96136425
hw-frames=124344,2279907 hw-frame-bytes=71899169,199457306 uptime=1h17m16s
last-activity=10ms signal-strength=-74dBm@6Mbps signal-to-noise=35dB
strength-at-rates=-74dBm@6Mbps 0s,-75dBm@9Mbps 1h17m11s340ms,-75dBm@12Mbps
10ms
tx-ccq=33% p-throughput=11315 nstreme=yes framing-mode=exact-size
framing-limit=2200 routeros-version="2.9.27" last-ip=69.39.23.58
802.1x-port-enabled=yes authentication-type=none encryption=none
compression=no

This began happening after we added approximately the 6th CPE to the AP.

jarosoup · Wed Jul 05, 2006 11:04 pm

tx rate times are times since last use of that tx rate i guess.

I think this is it...makes sense after looking at a few APs. Thanks

How close are these CPEs to each other? As in, are all these in the same neighborhood?

believewireless · Thu Jul 06, 2006 1:44 am

We are seeing the same problem. The AP is an RB532 with 16 connected CPE running Station WDS and NStreme with polling. Testing with an AP with only two connected clients looks fine. Based on the results shown below, signal quality isn't a problem. The AP is running at 333MHz w/SR5 and the client is an RB112 w/CM9.

25 packets transmitted, 25 packets received, 0% packet loss
round-trip min/avg/max = 3/9.5/30 ms

signal-strength: -66dBm
tx-signal-strength: -68dBm
noise-floor: -104dBm
signal-to-noise: 38dB
tx-ccq: 91%
rx-ccq: 100%
p-throughput: 36822
wds-link: yes
nstreme: yes
polling: yes
framing-mode: exact-size
framing-limit: 3200
framing-current-size: 2560

jd6strings · Thu Jul 06, 2006 2:32 am

tx rate times are times since last use of that tx rate i guess.
I think this is it...makes sense after looking at a few APs. Thanks

How close are these CPEs to each other? As in, are all these in the same neighborhood?

CPE's are spread out by at least 3 miles each direction....

boriss · Thu Jul 06, 2006 4:09 am

Go to the Ip ,firewall ,connections,and check connections there you can see
which SU make you trouble .Also Change TCP Established Timeout from 1d to 1 hour .Works for me after I make this changes .Ping come down From 400 ms to 1 to 3ms and speed from 300k to 6MB .

Best regards

jonbrewer · Thu Jul 06, 2006 4:11 am

Frequency scan show 0 usage. :roll:

I don't think Frequency Scan will show non-802.11 systems using the frequencies. Can someone from Mikrotik comment please?

You could be sharing the area with one or more of the following:

1. 2.4GHz AV sender
2. 2.4GHz DSSS Trango
3. 2.4GHz Motorola Canopy

JB

jd6strings · Thu Jul 06, 2006 4:57 am

Frequency scan show 0 usage.
I don't think Frequency Scan will show non-802.11 systems using the frequencies. Can someone from Mikrotik comment please?

You could be sharing the area with one or more of the following:

1. 2.4GHz AV sender
2. 2.4GHz DSSS Trango
3. 2.4GHz Motorola Canopy

JB

We're operating in 5GHz...

jd6strings · Thu Jul 06, 2006 6:49 am

So I'm determined to make this situation right and I think routing vs. bridging is ultimately the way to go. I've posted a diagram of my basic network config here:

http://www.skywerx.com/netdiagram.jpg

I guess I'm just not seeing how to add "routing" to the picture WITHOUT adding an entire hop in the network. What config changes in both the AP and the CPE would allow for me to get traffic inbound/outbound to the WWW? I assume that adding a default gateway to the AP of 10.10.0.1 AND then giving the AP an address of say 10.10.0.4 and THEN adding a default gateway to the CPE of 10.10.0.4 would do it??? Am I on the right track?

How do some of you others do it if you already have routers in the network that handle the traffic and you DON'T want to break out another subnet and put it DIRECTLY on the AP? How are you "routing" your traffic WITHOUT making the AP a dedicated "router" but rather pushing 0.0.0.0/0.0.0.0 to the correct gateway?

Please note that a MAJORITY of our network is Trango gear and the routing capabilities are NOT built into the AP/SU hence the reason for the network configuration. We've recently fell in love with the low cost MikroTik CPE BUT because of the numerous Trango radios (450+) we have deployed, re-subnetting to support "routing" in the MikroTik RB532 AP CAN't realistically happen...

Thank you!!!

maxfava · Thu Jul 06, 2006 11:37 am

I suggest the follow

RB532 Ether. 10.0.0.1/24 - default getway (your modem or server) --
RB532 Wireless 10.0.1.254/24

Client must in the network 10.0.1.X/24 with default gataway 10.0.1.254

Max

jd6strings · Thu Jul 06, 2006 3:22 pm

I suggest the follow

RB532 Ether. 10.0.0.1/24 - default getway (your modem or server) --
RB532 Wireless 10.0.1.254/24

Client must in the network 10.0.1.X/24 with default gataway 10.0.1.254

Max

Per the diagram, we use the network 10.10.0.0/22 for all of our clients....for management purposes. Is there any way to make use of that network on the client side to add "routing" to the picture instead of WDS? Like I said...for some reason I'm NOT seeing how to make it work....

I assume that the AP would need to be placed in bridge mode and the CPE in station mode to make this all happen???

maxfava · Thu Jul 06, 2006 5:57 pm

if you use bridge you will implement a large network. If you use route you will separate the network between the routers.

in WDS you will use bridge between the WDS links. Client will be in this bridge.
But you can route this network or client could be connected to an Virtual access point or assigned to a VLAN created inside the WDS bridge.

sten · Thu Jul 06, 2006 6:57 pm

... But you can route this network or client could be connected to an Virtual access point or assigned to a VLAN created inside the WDS bridge.

one virtual access point pr client? wouldn't that reduce performance dramatically?

Hammy · Fri Jul 07, 2006 12:09 am

Frequency scan show 0 usage. :roll:
I don't think Frequency Scan will show non-802.11 systems using the frequencies. Can someone from Mikrotik comment please?

You could be sharing the area with one or more of the following:

1. 2.4GHz AV sender
2. 2.4GHz DSSS Trango
3. 2.4GHz Motorola Canopy

JB

MT reports that the noise floor and thus SNR measurements are accurate and take into account ALL transmissions at that particular time... Alvarion, Canopy, Orthogon, etc.

jarosoup · Fri Jul 07, 2006 3:47 am

Did you ever try testing out other channels? What do ping tests and bandiwdth tests look like at 2AM? What does the registration table and the reg stat look like early in the morning?

I still think your problem is RF related. Based on your registration table, some of the link's signal strengths, and the ccq readings reported, I'm not sure if switching to routing is going to change anything. I've been wrong many times tho so you never know

Also, I don't think the wireless sniffer accounts for RF that isn't 802.11 or some variation (like proprietary wireless, cordless phones, Sat, etc)...can anyone confirm this?

maxfava · Fri Jul 07, 2006 2:56 pm

jd6strings,
have you tried to insert a firewall rule that drop invalid connection, accept related and valid? instead of accept all?

I tried this yesterday and at first time it works, but response will be tonight after 24 hours.

jd6strings · Fri Jul 07, 2006 3:20 pm

jd6strings,
have you tried to insert a firewall rule that drop invalid connection, accept related and valid? instead of accept all?

I tried this yesterday and at first time it works, but response will be tonight after 24 hours.

No I have not. Good idea. I added a filter for P2P but it filtered 0 packets in an 12 hour period.

What does your filter rule(s) look like?

Thanks!!

jd6strings · Fri Jul 07, 2006 3:25 pm

Did you ever try testing out other channels? What do ping tests and bandiwdth tests look like at 2AM? What does the registration table and the reg stat look like early in the morning?

I still think your problem is RF related. Based on your registration table, some of the link's signal strengths, and the ccq readings reported, I'm not sure if switching to routing is going to change anything. I've been wrong many times tho so you never know

Also, I don't think the wireless sniffer accounts for RF that isn't 802.11 or some variation (like proprietary wireless, cordless phones, Sat, etc)...can anyone confirm this?

Actually we changed channels and I could NOT get the clients to re-associate. I was under the impression that the clients would use the default scan-list to find the AP with the associated SSID if NO country was selected. That did NOT happen. Maybe I've misconfigured something??? How long does the re-association process normally take after changing channels? I waited approx 3 minutes and 0 registration....

maxfava · Fri Jul 07, 2006 3:56 pm

from the manual

/ip firewall filter
add chain=forward protocol=tcp connection-state=invalid \
action=drop comment="drop invalid connections"

add chain=forward connection-state=established action=accept \ comment="allow already established connections"

add chain=forward connection-state=related action=accept \
comment="allow related connections"

jd6strings
Also are you using pptp connection between you customers and server?

jd6strings · Fri Jul 07, 2006 4:28 pm

from the manual

/ip firewall filter
add chain=forward protocol=tcp connection-state=invalid \
action=drop comment="drop invalid connections"

add chain=forward connection-state=established action=accept \ comment="allow already established connections"

add chain=forward connection-state=related action=accept \
comment="allow related connections"

jd6strings
Also are you using pptp connection between you customers and server?

Thanks for the repy. I actually found it in the manual right after I posted...thanks again. We are NOT using point-to-point tunneling. Why do you ask?

maxfava · Fri Jul 07, 2006 5:30 pm

I ask becouse I use VPN for my customers and
I'm tinking to change packet size to not have fragmentation. I use change mms change but on the edge server and not on the AP, it could be a problem.

pinotage · Sun Jul 09, 2006 7:42 pm

jd6strings

Try turning off the nstreme on the AP and the clients, We have found this to overload the RB532 too much and cause high pings and all sorts of other troubles in a point to multipoint configuration.

Hope it helps you.

Regards
Albie

Hammy · Sun Jul 09, 2006 8:00 pm

With the 532 as the AP or as the client?

pinotage · Sun Jul 09, 2006 8:01 pm

As AP

Hammy · Sun Jul 09, 2006 8:08 pm

So there shouldn't be a horsepower issue with an Athlon AP and a RB500 client?

pinotage · Mon Jul 10, 2006 7:37 am

No, probably not.

jd6strings · Mon Jul 10, 2006 3:15 pm

jd6strings

Try turning off the nstreme on the AP and the clients, We have found this to overload the RB532 too much and cause high pings and all sorts of other troubles in a point to multipoint configuration.

Hope it helps you.

Regards
Albie

I'll give that a try. Thanks! By the way do you use WDS in your PTMP configuration?

maxfava · Mon Jul 10, 2006 4:00 pm

After infinite trials, I have found a different behaviour changing the channels.

jd6strings · Mon Jul 10, 2006 6:18 pm

After infinite trials, I have found a different behaviour changing the channels.

What do you mean by this? Do you mean re-association issues?

maxfava · Mon Jul 10, 2006 9:56 pm

What do you mean by this? Do you mean re-association issues?

I wrote about high ping issue.
Operating on the 2.4GHz, I notice that on channal 11 the ping response is high and for some minutes there is not ping. after that it come to 1ms. But I have not reassotiation problem, the clients still connected without disconnection.

On channel 1 the ping is more stable and high ping issue is sporadic.

But the issue still live.

pinotage · Wed Jul 12, 2006 9:32 am

jd6strings

I'll give that a try. Thanks! By the way do you use WDS in your PTMP configuration?

No, we don't use wds, increases load and decreases throughput.

mspencer · Thu Jul 20, 2006 5:47 am

I am strugling with the same issue only I route my entire network. My ping times run from 1 ms to 999 to 2000 to nothing. I replaced one of the CB3 bridges with an EZ Bridge and had the same problem. Finally replaced it with a RB112 and the complete disconnect went away but the ping times are still eratic.

maxfava · Fri Jul 21, 2006 2:24 pm

AT this moment I have solved the issue changing the channel. But the strange is that the channel previous selected was free.

I'm becoming crazy.

jd6strings · Fri Jul 21, 2006 7:55 pm

As I mentioned in a previous post, the frequency scan showed that the frequency we are using was clean so we DID NOT change channels. I broke out a few new subnets for the AP to handle and simply routed all traffic. Took WDS OUT of the picture and removed ALL bridges. Happy to say the ping times are NOW consistently arround 3-7 ms versus the 30-700 ms we were seeing.

The lesson I learned here is...

ROUTE YOUR TRAFFIC and USE WDS ONLY FOR YOUR POINT-TO-POINT LINKS.

maxfava · Mon Jul 24, 2006 3:18 pm

Thank you,
so at the end using router instead of bridge solve the issue. Pheraps I need to remap all my network (about 100 customers)

.

Could Mikrotik support analize the differents? Becouse I sent many support out but the issue was addressed to the signal of the clients.

Massimo

believewireless · Mon Jul 24, 2006 4:26 pm

So you would just set the CPE to "station" and the AP to "AP Bridge" and route the traffic between the two WLAN interfaces?

sten · Tue Jul 25, 2006 1:31 am

i only heard this from someone else who tried to use wds to each customer;
dont. You add latency for each wds interface added.
I use wds for each customers. And I have the same strange problem about long ping. But I use PPPoE so to bridge my customers the manual says I have to use wds. May you explain howto use pppoe for my customers (that are all MT based) without using wds?

Let your cpe do the pppoe connections.

jd6strings · Tue Jul 25, 2006 5:55 am

So you would just set the CPE to "station" and the AP to "AP Bridge" and route the traffic between the two WLAN interfaces?

Yep..and create a default gateway in the CPE to your AP. Then create a default gateway on your AP to the next hop in your network (i.e. intermediary router, edge router, etc.).

maxfava · Wed Sep 06, 2006 1:40 pm

I confirm that using the router is better than bridge.
The high ping phenomena occour less, but it still present and only in a card (CM9). The other card is CM) too. the channel used are 1 and 10.

jober · Wed Sep 06, 2006 5:34 pm

I have an AP with the same ping problem. The only fix I could find was to set the AP's sr2 card to B-only. I think my problem was a bad mix of client radios. (CB3's,Deliberant 1300's/2300's, rb112's and rb532's)
I have been replacing the older CPE's with RB532s that have SR5 radios to connect to the tower and a CM9 or SR2 for hotspot service at the customers location. In 6 months I should have the whole town covered with the hotspot service.

maxfava · Thu Sep 07, 2006 11:11 am

The strange behavior occur when I change the channels.

First of all my clients are mixed in standard B and G with router DWL2000ap or DWL900ap as client on the roof of their houses.
I have RB532 with 2 minipci, both CM9 serving as access point, one with 5 clients and the second with 20 clients connected.
When I set channel 1 on the first CM9 and Channel 11 for the second one the ping results high alternate between clients connected:

When I ping to two clients, one connected to 1° CM9 and the other connected to the 2° CM9, the ping results in time out to first client while the 2° is stable to 1 ms and after 5 sec the behavior is reverse the 2° client result in time out and the 1° good ping at 1 ms. After other 5 sec it reverse again.
I try to make the visualization

Client 1 (to the 1°CM9) ---------------- Client 2 (to the 2° CM9)
1ms -------------------------------------- Time out
1ms -------------------------------------- Time out
1ms -------------------------------------- Time out
1ms -------------------------------------- Time out
..... for 5 sec
Time out -------------------------------------- 1 ms
Time out -------------------------------------- 1 ms
Time out -------------------------------------- 1 ms
Time out -------------------------------------- 1 ms
Time out -------------------------------------- 1 ms
.. for 5 sec.

This behavior disappears when I set the same channel to both CM9.
for example both to channel 11 or channel 10.
But it is clear that both on the same channel is not possible for interferences.

I cannot believe that I'm the only one that has this issue.

Max

High ping times with 11 CPE's on RB532...

High ping times with 11 CPE's on RB532...

Re: High ping times with 11 CPE's on RB532...

High Ping

nstreme

Re: nstreme

Re: nstreme

Who is online