Re: [bwm-tools-tech] Configuration

[Ravi P. <ra...@sw...>]

Almost 5 hours and no problems so far, neither on RAM or CPU as well.

If it can saturate a gigabit card, then its really good in fact better than
commercial products I know. Fantastic Job Nigel and thanks to all who have
helped in developing this product.

What do you mean renicing bwm??

Regards
Ravi 

-----Original Message-----
From: Nigel Kukard [mailto:nk...@lb...] 
Sent: Tuesday, July 25, 2006 9:55 PM
To: ra...@sw...
Cc: bwm...@li...
Subject: Re: [bwm-tools-tech] Configuration


> Ok changed IPQ_BUFSIZE to 16384 and recompiled and it seems to be 
> working for the last 5 minutes without any errors. Will keep you posted.

Cool .... only drawback is bwm will now eat ram like its going out of
fashion ... not tooo bad though, I'll try address the problem in the next
development release.


> 
> Please advise if you want me to do anything else on the box.
> 
> CPU usage is also around 1-3% only.

Great ...

> 
> Thanks very much for your help.
> 
> What are the limitations of bwm? I will be playing around with the 
> graphing utilities and will get back to you in case of any problems.

Well ... I've tried it out on a switched Gbit network and managed to fully
saturate the network cards ... on a celeron D 2.8 it pulled about 30% CPU.

I would also recommend you maybe increase your ip_queue size to 32768 or
something, just incase something else munches CPU and bwm is unable to keep
up .... I'm also looking at renicing bwm to give it high priority aswell.


-Nigel


> 
> Thanks again.
> 
> Regards
> Ravi
> 
> -----Original Message-----
> From: Nigel Kukard [mailto:nk...@lb...]
> Sent: Tuesday, July 25, 2006 8:29 PM
> To: ra...@sw...
> Subject: Re: [bwm-tools-tech] Configuration
> 
> I wonder ... what would happen when a LARGE packet has been fragmented 
> by the sending side, then re-assembled on a Linux router? interesting
> 
> I'm thinking this could be happening, this way the packets you get are 
> 1500 bytes, but being re-assembled into something massive ... which 
> overflows the small buffer used for reading from ipq.
> 
> Changing IPQ_BUFSIZE will tell us if this is true
> 
> 
> Ravi Patwari wrote:
>> How do I check this. I have not made any specific change anywhere 
>> that I know of.
>>
>> -Ravi
>>
>> -----Original Message-----
>> From: Nigel Kukard [mailto:nk...@lb...]
>> Sent: Tuesday, July 25, 2006 8:23 PM
>> To: ra...@sw...
>> Subject: Re: [bwm-tools-tech] Configuration
>>
>> Is it possible your packets are larger than 1500 bytes?
>>
>> -Nigel
>>
>>
>> Ravi Patwari wrote:
>>> Hi Nigel,
>>>
>>> Please help me on this problem with bwmd as it does not seem to be 
>>> handling packets from the QUEUE correctly.
>>>
>>> By the way I am using CentOS, which is redhat enterprise linux. 
>>>
>>> Regards
>>> Ravi
>>>
>>> -----Original Message-----
>>> From: Nigel Kukard [mailto:nk...@lb...]
>>> Sent: Monday, July 24, 2006 3:05 PM
>>> To: ra...@sw...; Bandwidth Management Tools General & 
>>> Technical Discussions
>>> Subject: Re: [bwm-tools-tech] Configuration
>>>
>>>
>>>>  
>>>> On running bwmd, I immediately get the following:
>>>> =================================
>>>> BWM Daemon v0.2.3 - Copyright (c) 2003-2006 Linux Based Systems 
>>>> Design
>>>>  
>>>> BWMD: Loaded 20 flows and 18 queues
>>>> BWMD: Found 1 modules to load
>>>> Loading ip_queue...done
>>>> IPQ runner started...
>>>> Flow runner started...
>>>> Stat thread started...
>>>> Report runner started...
>>>> Failed to get packet from IPQ: Received message truncated
>>>> passer: Received message truncated
>>>>
>>>> =========================
>>>>   
>>> Can you try the latest development snapshot please.
>>>
>>> -Nigel
>>>
>>>
>>>
>>
>>
> 
> 
>