[c-nsp] Serial T1/E1 Card

Oliver Boehmer (oboehmer) oboehmer at cisco.com
Tue Jan 9 05:18:38 EST 2007


William <mailto:willay at gmail.com> wrote on Tuesday, January 09, 2007
10:19 AM:

> Hi Oliver,
> 
> The app that uses the link is indeed TCP, when we see the line "max
> out" at 1.5mbit their application freezes/whatever. We spoke to the
> end users and the timestamps where the line tops out at 1.5 matches to
> when they see issues with their application.

Ok.. Still: do you see any output drops on the interface? Which
direction is maxing out? Your input or your output? If you see max 1.5
Mbit on input, you need to check the other side..

> I'm looking into iperf now but if I require a daemon running on the
> other side of the link then I may be out of luck!

iperf requires a server, but maybe you have someone with a good Internet
connection who can start one for you.

	oli


> 
> 
> On 09/01/07, Oliver Boehmer (oboehmer) <oboehmer at cisco.com> wrote:
>> With SNMP, you usually query the input/output octet counters and do
>> the load calculation yourself (delta/interval), so the load-interval
>> doesn't play any role here. I don't think there is a "load" MIB OID
>> you can query directly.. 
>> 
>> But I don't think you'll see micro-burst decreasing the load
>> interval. I'd first try some load test application (like iperf) to
>> see if you really can't get more than 1.5 Mbit over the link.
>> Do you see any output drops on the interface? If you do, it could be
>> TCP's backoff behavior resulting in a less efficient BW use, try
>> enabling random-detect on the interface and see if this makes a
>> difference. It could if there was a larger number of TCP
>> applications using this link.. 
>> 
>>         oli
>> 
>> cisco-nsp-bounces at puck.nether.net <> wrote on Tuesday, January 09,
>> 2007 8:47 AM: 
>> 
>>> Um this is confusing for me as well, for some MIB's like frame relay
>>> i have studied that it will affect the results but not very sure ,
>>> some one else needs to pop-in here ?
>>> 
>>> On 1/9/07, William <willay at gmail.com> wrote:
>>>> 
>>>> Will it also change any of the statistics I'm getting via SNMP?
>>>> 
>>>> Cheers,
>>>> 
>>>> Will
>>>> 
>>>> On 09/01/07, Shakeel Ahmad <shakeelahmad at gmail.com> wrote:
>>>>> load-interval will only modify the samping time of interface
>>>>> statistics which you can see by show int s0/0 command.
>>>>> 
>>>>> gripping it down to 30 (the minimum) will enable you to see
>>>>> somewhat realtime traffic throughput in/out (30 seconds avg.
>>>>> statistics). 
>>>>> 
>>>>> 
>>>>> On 1/9/07, William <willay at gmail.com> wrote:
>>>>>> 
>>>>>> On 08/01/07, Alex <alex.arseniev at gmail.com> wrote:
>>>>>>> Hello there,
>>>>>>> Polling an interface every 2 secs probably would not give You
>>>>>>> meaningful data with "load-interval" at 5 mins (default).
>>>>>>> What is the latency across this link? With default TCP window
>>>>>>> size at 64 kbytes and sizeable latency there might be not
>>>>>>> enough data to fill the pipe... HTH Cheers
>>>>>>> Alex
>>>>>>> 
>>>>>> 
>>>>>> Hello Alex,
>>>>>> 
>>>>>> The latency is roughly 3-4ms across the link, when you talk about
>>>>>> the load-interval at 5mins are you talking about the mib I'm
>>>>>> querying? 
>>>>>> 
>>>>>> Cheers,
>>>>>> 
>>>>>> Will
>>>>>> _______________________________________________
>>>>>> cisco-nsp mailing list   cisco-nsp at puck.nether.net
>>>>>> https://puck.nether.net/mailman/listinfo/cisco-nsp
>>>>>> archive at http://puck.nether.net/pipermail/cisco-nsp/
>>>>>> 
>>>>> 
>>>>> 
>>>> 
>>> _______________________________________________
>>> cisco-nsp mailing list  cisco-nsp at puck.nether.net
>>> https://puck.nether.net/mailman/listinfo/cisco-nsp
>>> archive at http://puck.nether.net/pipermail/cisco-nsp/



More information about the cisco-nsp mailing list