[Sis-SCPS-INTEREST] SCPS with jitter

[Feighery, Patrick D.]

Keith asked me to pull the traces and take a look a them.

When I looked at the traces, I noticed that each data packet processed
by NISTNET appears twice in the tcpdump while TCP ACKS only appeared
once.  This made it a little harder to get a handle of what was going
on.  The short answer is NistNet's jitter does not preserve packet
ordering.  Therefore you are getting a plethora of unnecessary
retransmissions.  Setting the BIF_SNACK_DELAY and also the BIF_MINRTO
to ensure RTOs do not occur unnecessary should do the trick.  Let me
know if this helps...

The bigger question is...  Is the jitter that NistNet produces mimic
the environment you are trying to emulate...

Hope this help. 

	Pat


>>-----Original Message-----
>>From: Scott, Keith L. 
>>Sent: Monday, January 23, 2006 10:43 AM
>>To: 'frankm at klasonline.com'
>>Cc: 'sis-scps-interest at mailman.ccsds.org'; Feighery, Patrick D.
>>Subject: Re: [Sis-SCPS-INTEREST] SCPS with jitter
>>
>>Great, thanks.  I'm on the road at the moment and will pull 
>>the dumps when I get a chance.
>>
>>  --keith
>>
>>
>>-----Original Message-----
>>From: Frank Murray <frankm at klasonline.com>
>>To: Scott, Keith L. <kscott at mitre.org>
>>CC: sis-scps-interest at mailman.ccsds.org 
>><sis-scps-interest at mailman.ccsds.org>; Feighery, Patrick D. 
>><feighery at mitre.org>
>>Sent: Mon Jan 23 10:38:57 2006
>>Subject: Re: [Sis-SCPS-INTEREST] SCPS with jitter
>>
>>Keith,
>>I think you're probably right about nistnet reordering the packets.
>>
>>I've uploaded 2 dumps to our website:
>>http://www.klasonline.com/download/Klasrouter/scps_logs/tcp2_krt.log
>>http://www.klasonline.com/download/Klasrouter/scps_logs/tcp2_nist.log
>>
>>tcps_krt.log is taken at the FTP server end of the connection in the 
>>non-SCPS portion while tcp2_nist.log is taken between the 2 SCPS 
>>gateways. The FTP client is at 192.168.0.41 and the FTP server is 
>>192.168.150.254. FYI the link between the SCPS gateways over 
>>nistnet is 
>>GRE'd to make it easier to mask all traffic with nistnet.
>>
>>Thanks for your prompt reply,
>>Frank
>>
>>Scott, Keith L. wrote:
>>
>>> I wonder if NISTNET delivers packets out of order when the 
>>jitter is 
>>> high enough?  That could be bad, but shouldn't be too bad with pure

>>> rate control.
>>>
>>> With pure rate control (no congestion control) there's no 
>>way to get 
>>> stuck in slow start.
>>>
>>> Can you tcpdump both ends of the inter-gateway interface?
>>>
>>>   --keith
>>>
>>>
>>> -----Original Message-----
>>> From: sis-scps-interest-bounces at mailman.ccsds.org 
>>> <sis-scps-interest-bounces at mailman.ccsds.org>
>>> To: sis-scps-interest at mailman.ccsds.org 
>>> <sis-scps-interest at mailman.ccsds.org>
>>> Sent: Mon Jan 23 05:15:45 2006
>>> Subject: [Sis-SCPS-INTEREST] SCPS with jitter
>>>
>>> All,
>>> I am currently testing 2 SCPS gateways in the configuration below:
>>>
>>> PC1->SCPS GW1---->Satellite simulator--->SCPS Gw2----->PC2
>>>
>>> I use nistnet as a sat sim with a round-trip time of 1s and 
>>a symmetric
>>> bandwidth of 512K. I'm using FTP to test performance and as it is a
>>> dedicated link I'm using pure rate control for BIF_CC. The 
>>results have
>>> been good and I get to wire-speed very quickly with this setup.
I've
>>> injected errors and the performance is still very good.
>>>
>>> Problems arise however when I simulate jitter on the sat 
>>sim. I've set
>>> the jitter to have a standard deviation of 100ms and on 2 in 3 ftp
>>> uploads I get wire speed as before but on the other I get average
>>> bandwidths of about 30 kbits/s. The speed never builds up so I'm
>>> guessing that we get stuck in slow start because an ack 
>>gets delayed at
>>> startup. Is this behaviour to be expected with pure rate control?
>>>
>>> I switched to Vegas but got similar results. I read in one 
>>of the FAQs
>>> that jitter *can* be interpreted as congestion so played 
>>around with the
>>> alpha/betta/gamma parameters without things improving but 
>>there was an
>>> amount of trial and error about this. Any advice on how to 
>>tune these
>>> parameters would be great.
>>>
>>> Regards,
>>>
>>> --
>>> Frank Murray
>>> KlasRouter Project Manager
>>> Klas Ltd
>>>
>>> -------------------------------------------
>>> mail to : frankm at klasonline.com
>>>
>>> Tel: International +353-1-662-4270
>>> Tel: Direct line +353-1-4361768
>>> Tel: VoIP 3123 at blueface.ie
>>>
>>>
>>>
>>> www.klasonline.com - 'making PC encryption easier'
>>>
>>>
>>> _______________________________________________
>>> Sis-SCPS-INTEREST mailing list
>>> Sis-SCPS-INTEREST at mailman.ccsds.org
>>> http://mailman.ccsds.org/cgi-bin/mailman/listinfo/sis-scps-interest
>>>
>>
>>-- 
>>Frank Murray
>>KlasRouter Project Manager
>>Klas Ltd
>>
>>-------------------------------------------
>>mail to : frankm at klasonline.com
>>
>>Tel: International +353-1-662-4270
>>Tel: Direct line +353-1-4361768
>>Tel: VoIP 3123 at blueface.ie
>>
>>
>>
>>www.klasonline.com - 'making PC encryption easier'
>>
>>