Re: [rmcat] Question on draft-ietf-rmcat-eval-test-03 and bottleneck queue size

[Ingemar Johansson S <[e-mail]>]

The problem I see is that with the original settings in section 5.7 in
draft-ietf-rmcat-eval-test-03 you have a number of TCP connections (ranging
from 1 to 10, depending on the randomized settings) that continuously
slowstart into a quite low bitrate bottleneck (only 2Mbps). What you get is
a very bursty traffic pattern that will hit a delay sensitive congestion
control quite badly with very low media bitrate as a result. 

Changing to an object size distribution between 30 and 50kbps gives bursty
transmission too but that is still manageable as the TCP flows do not
saturate the link.

 

/Ingemar

 

From: Zaheduzzaman Sarker 
Sent: den 29 juni 2016 00:15
To: Ingemar Johansson S <[e-mail]>; [e-mail]
Subject: RE: [rmcat] Question on draft-ietf-rmcat-eval-test-03 and
bottleneck queue size

 

Hi,

 

Not sure I understand the confusion.

 

If there is no other traffic on the path except the TCP traffic the path
capacity is supposed to filled up by the TCP traffic.

 

By the way, the short flows are supposed to me modeled according to
https://tools.ietf.org/html/draft-ietf-rmcat-eval-criteria-04#section-6.1

 

May be we should remove the test specific information section for the
eval-test draft.

 

 

BR

 

Zahed

 

From: Ingemar Johansson S 
Sent: den 28 juni 2016 22:48
To: Zaheduzzaman Sarker <[e-mail]
<mailto:[e-mail]> >; [e-mail]
<mailto:[e-mail]> 
Subject: RE: [rmcat] Question on draft-ietf-rmcat-eval-test-03 and
bottleneck queue size

 

OK, then I get it, thanks. 

Stumbled on another question, though. 

 

When I run TC 5.7 (many small TCP flows), it seems like the TCP pressure
becomes quite high. If I run completely without RMCAT flow the 2Mbps link
gets almost fully saturated 100% of the time. 

The settings I use are

Linux TCP with Cubic

10 FTP clients 

  100kB-1000kB file size (uniform distribution)

  Idle time between reads, 10s mean, exponentially distributed.

Have I misunderstood the description for TC5.7 ?

 

 

/Ingemar

 

From: Zaheduzzaman Sarker 
Sent: den 28 juni 2016 22:28
To: Ingemar Johansson S <[e-mail]
<mailto:[e-mail]> >; rmcat WG ([e-mail]
<mailto:[e-mail]> ) <[e-mail] <mailto:[e-mail]> >
Subject: RE: [rmcat] Question on draft-ietf-rmcat-eval-test-03 and
bottleneck queue size

 

Please see inline below.

 

BR

 

Zahed

 

From: rmcat [mailto:[e-mail]] On Behalf Of Ingemar Johansson S
Sent: den 28 juni 2016 18:13
To: rmcat WG ([e-mail] <mailto:[e-mail]> ) <[e-mail]
<mailto:[e-mail]> >
Subject: [rmcat] Question on draft-ietf-rmcat-eval-test-03 and bottleneck
queue size

 

Hi

 

The text in section 4.2 in draft-ietf-rmcat-eval-test-03 confuses me
somewhat

It mentions a bottleneck queue size of 300ms, given that the capacity
changes this means that the queue size in bytes of a drop tail queue should
change as well or ?

[Zaheduzzaman Sarker] No.

Or is the intention that one use the Reference bottleneck capacity to
compute a queue size 

With a ref bottleneck capacity of 1Mbps

[Zaheduzzaman Sarker] yes

and 300ms queue size that would then mean 

0.300*1e6 = 300000bits ?

 

/Ingemar

 

==================================

Ingemar Johansson  M.Sc. 

Master Researcher

 

Ericsson AB

Wireless Access Networks

Labratoriegränd 11

971 28, Luleå, Sweden

Phone [phone]

SMS/MMS [phone]

 <mailto:[e-mail]> [e-mail]

www.ericsson.com

 

    Real knowledge is to know the 

       extent of one's ignorance

                 Confucius

==================================