Slowness Spikes

mperdue · Wed May 11, 2005 6:04 pm

If you observe this ping from our radio

63.169.54.1 64 byte ping: ttl=244 time=67 ms
63.169.54.1 64 byte ping: ttl=244 time=85 ms
63.169.54.1 64 byte ping: ttl=244 time=77 ms
63.169.54.1 64 byte ping: ttl=244 time=77 ms
63.169.54.1 64 byte ping: ttl=244 time=77 ms
63.169.54.1 64 byte ping: ttl=244 time=94 ms
63.169.54.1 64 byte ping: ttl=244 time=107 ms
63.169.54.1 64 byte ping: ttl=244 time=121 ms
63.169.54.1 64 byte ping: ttl=244 time=417 ms
63.169.54.1 64 byte ping: ttl=244 time=326 ms
63.169.54.1 64 byte ping: ttl=244 time=267 ms
63.169.54.1 64 byte ping: ttl=244 time=357 ms
63.169.54.1 64 byte ping: ttl=244 time=228 ms
63.169.54.1 64 byte ping: ttl=244 time=67 ms
63.169.54.1 64 byte ping: ttl=244 time=67 ms
63.169.54.1 64 byte ping: ttl=244 time=67 ms

It shows normal for about a couple of mins at 67ms then spikes up and slows the connection down. The antani's are in allignment and really don't feel there is any interfearance in the area.

Both sides are prisim cards on 802.11b.

Have any suggestions where I should look for this issue?

Thanks,
Michael

GJS · Wed May 11, 2005 7:31 pm

What is CPU useage like when you see the spike?

mperdue · Wed May 11, 2005 10:43 pm

Ok after some review of our recent antani change several days ago we discovered that one of our antani's was horizontal polorzation. We just got off the roof and fixed that. The ping rates are down to 8ms now. HOWeever.. very simular to the above information we still get spikes up to about 20ms every 20-25 seconds. then back down. These spikes last for about 6-8 pings.

This is a 5ghz link (previous signal strenth was -82) fixing the polorzation it is now -68.

Now I think there is a software/setting causing the current spike. But I get no timeouts now.

mperdue · Wed May 11, 2005 10:46 pm

69.18.108.1 64 byte ping: ttl=126 time=7 ms
69.18.108.1 64 byte ping: ttl=126 time=7 ms
69.18.108.1 64 byte ping: ttl=126 time=7 ms
69.18.108.1 64 byte ping: ttl=126 time=7 ms
69.18.108.1 64 byte ping: ttl=126 time=7 ms
69.18.108.1 64 byte ping: ttl=126 time=7 ms
69.18.108.1 64 byte ping: ttl=126 time=7 ms
69.18.108.1 64 byte ping: ttl=126 time=7 ms
69.18.108.1 64 byte ping: ttl=126 time=7 ms
69.18.108.1 64 byte ping: ttl=126 time=61 ms
69.18.108.1 64 byte ping: ttl=126 time=58 ms
69.18.108.1 64 byte ping: ttl=126 time=37 ms
69.18.108.1 64 byte ping: ttl=126 time=48 ms
69.18.108.1 64 byte ping: ttl=126 time=47 ms
69.18.108.1 64 byte ping: ttl=126 time=7 ms
69.18.108.1 64 byte ping: ttl=126 time=8 ms

Here is the new results.

randyloveless · Thu May 12, 2005 7:24 am

most likely this is when the radio is trying different modes .

mperdue · Fri May 13, 2005 1:00 am

After further testing of the orginal problem we found out the slow link was due to our backhaul link 802.11a was in the wrong polorzation.

But we still have this issue with 802.11b link on only one of the two hotspot/dedicated links.

I can ping from customer end unit to the ip address of the prisim card at 7ms constistant from five differnt sites. This unit works great.

On our second unit we have four sites and it too starts pinging at 7ms. And as you watch it it will jump from 7ms to about 280-300ms. It pings at about that range for about six to eight times, does one time out the returns to 7ms.

The prisim card is set for 1M rate, 100mw output, into a 1 watt amp. I will probibly go up and disable the amp tommorw and re-run my tests.

Do you have any other suggestion as to what else to check?

randyloveless · Fri May 13, 2005 11:12 am

2 things to look at amp and nic cards

we have had some issues with via itx board having slow ftp which is due to there on board nic card .

Randy

what type of M/B and what type of amp

wildbill442 · Fri May 13, 2005 11:28 am

have you tried doing an interference scan. I've noticed the Prism cards dont perform well when theres a lot of nearby interference. However this could be the case with any card, but I've yet to swap with an Atheros.

mp3turbo2 · Fri May 13, 2005 12:50 pm

people, why do you use amps? 1W amp with 100mW output power on card will give you ~250km reach for your clients! You don't need any amps on links up to 30km... damn.

You have different polarization on client antenna than transmitters one, you use 1W amps... Then, you're wondering why is the mikrotik so bad software.

It is not mikrotik, it is PEBKAC problem. Problem Exists Between Keyboard And Chair.

mperdue · Fri May 13, 2005 7:30 pm

I have found the issue. In planning to do another relay off this site I had setup the static ip routes to the next relay. That relay has not been installed. After removing the statatic route, and removing the routing on my main cisco router that pointed a dead subnet across the wireless link. I am getting a very stable connection of about 4ms across a 1.5 mile link.

I am still waiting for feedback from my voip customers as to if this is solved the spike issue and the voip problems. But the testing that I have done so far I think it is ok.

-------------------------------

As for the 1 watt amp, as the previous gentleman had complained about. It is defntly needed here in these mountains. Without the amp the 3.5 mile link that I have just dosn't work. With the amp I get a 1m solid connection running voip across it.

mip · Sat May 14, 2005 2:23 am

"It is not mikrotik, it is PEBKAC problem. Problem Exists Between Keyboard And Chair."
GigaLOL