perfsonar-user - Re: [perfsonar-user] Regular testing and bwctld
Please Wait...
perfsonar-user@internet2.edu
Subject: perfSONAR User Q&A and Other Discussion
List archive
- From: Ty Bell <tyabell@grnoc.iu.edu>
- To: Aaron Brown <aaron@internet2.edu>
- Cc: perfsonar-user <perfsonar-user@internet2.edu>
- Subject: Re: [perfsonar-user] Regular testing and bwctld
- Date: Mon, 9 Mar 2015 12:58:17 -0400
Hi Aaron,
I think I have it sorted, I needed to set window_size
--Ty
> On Mar 9, 2015, at 11:07 AM, Aaron Brown
> <aaron@internet2.edu>
> wrote:
>
> Hey Ty,
>
> I think that in regular testing, both UDP throughput, and owamp tests store
> the packet loss rates in the MA, so that’d explain why you’re seeing the
> loss rate concurrent with the throughput test. Do you see a similar UDP
> throughput packet loss when you do the test by hand?
>
> Cheers,
> Aaron
>
>> On Mar 9, 2015, at 10:59 AM, Ty Bell
>> <tyabell@grnoc.iu.edu>
>> wrote:
>>
>> Hi Aaron,
>>
>> In Maddash, it shows the exact same time as the UDP test result. Hovering
>> the throughput line, it says 100Mbps, 15:40:51, dropping down to the loss
>> line, it says Loss 22.194% 15:40:51. Is the UDP test reporting loss or is
>> it the latency test?
>>
>> --Ty
>>
>>> On Mar 9, 2015, at 10:15 AM, Aaron Brown
>>> <aaron@internet2.edu>
>>> wrote:
>>>
>>> Hey Ty,
>>>
>>> What time does the loss spike occur? If the iperf3 processes aren’t
>>> running parallel with the owampd ones, I’d not expect huge spikes like
>>> that. There aren’t any powstream processes running are there? Beyond
>>> that, my only guess is some other traffic that happens to hit at roughly
>>> the same time (possibly another UDP 100M test that happens to share a
>>> path or something).
>>>
>>> Cheers,
>>> Aaron
>>>
>>>> On Mar 9, 2015, at 9:25 AM, Ty Bell
>>>> <tyabell@grnoc.iu.edu>
>>>> wrote:
>>>>
>>>> Hi All,
>>>>
>>>> I’ve been using regular testing with bwping (owamp) and bwctl (iperf3).
>>>> Every 6 hours I run a 20 second 100Mbps UDP test, and what I’m seeing in
>>>> the graphs are very large loss spikes when the UDP tests run. In the
>>>> logs, I see the iperf3 test accepted at 20:56:01, it says 62 seconds
>>>> until the results are available. I see 9 owamp processes exit (the final
>>>> at 20:56:38), at 20:57:03 the iperf3 test completes and then all of the
>>>> owamp connections re-establish. During the iperf test, if I look at the
>>>> process list, I can see iperf3 is the only thing running, no owamp or
>>>> owping.
>>>>
>>>> Any ideas what could be causing the loss? No errors or drops on the NICs.
>>>>
>>>>
>>>> Thanks,
>>>> Ty
>>>
>>
>
Attachment:
signature.asc
Description: Message signed with OpenPGP using GPGMail
- [perfsonar-user] Regular testing and bwctld, Ty Bell, 03/09/2015
- Re: [perfsonar-user] Regular testing and bwctld, Aaron Brown, 03/09/2015
- Re: [perfsonar-user] Regular testing and bwctld, Ty Bell, 03/09/2015
- Re: [perfsonar-user] Regular testing and bwctld, Aaron Brown, 03/09/2015
- Re: [perfsonar-user] Regular testing and bwctld, Ty Bell, 03/09/2015
- Re: [perfsonar-user] Regular testing and bwctld, Aaron Brown, 03/09/2015
- Re: [perfsonar-user] Regular testing and bwctld, Ty Bell, 03/09/2015
- Re: [perfsonar-user] Regular testing and bwctld, Aaron Brown, 03/09/2015
Archive powered by MHonArc 2.6.16.