Re: [ippm] Testing TCP Throughput Capacity in Operator Networks

["Barry Constantine" <Barry.Constantine at jdsu.com>]

Henk and Ruediger,

Thank for your comments and I would like to clarify a few points
concerning the proposal for a draft related to TCP Throughput Capacity
test methodology and metrics.

When I first addressed the subject with Henk and Matt Zekauskas, I was
pointed to RFC 3148, draft-ietf-ippm-treno-btc-03, and
draft-ietf-ippm-btc-cap-00 as a starting point.  By the way, I found
these to be very interesting and useful papers.

The TCP Throughput Capacity draft would reference these papers but would
focus more at the network level metrics of TCP performance, not the
specific behavior of various TCP mechanisms and implementations (i.e.
CAC, False time-outs, Self-clocking, etc.).

In other words, given a large complex network with various QoS and
queuing techniques employed, the TCP throughput technique would allow a
network provider to quantify the ability of the entire network to carry
TCP traffic at specified SLA levels.  Varying window sizes and MSS sizes
is one component of this type of test. Another key component to this
test is the throughput achieved with varying background traffic (perhaps
UDP) and verifying proper prioritization between the TCP traffic and the
background traffic.  

It is very difficult for a network provider to extrapolate TCP layer
performance from packet loss studies and metrics.  RFC-2544 is widely
used today and makes it difficult (if not impossible) to adequately
characterize how TCP will perform.

Concerning the interest level based upon the responses, the sample size
is small, but I have received 3-4 who are interested and 3-4 with little
interest.  Referencing Henk's final remarks; does this constitute enough
interest from this working group to work on a draft?

Thanks,

Barry

 

Principal Member of Technical Staff

 

JDSU Communication Test (formerly Acterna)

Emerging Markets and Technology Research         

One Milestone Center Court                              

Germantown, MD 20876                                         

(W) 240-404-2227                                                

(C) 301-325-7069

-----Original Message-----
From: ippm-bounces at ietf.org [mailto:ippm-bounces at ietf.org] On Behalf Of
Henk Uijterwaal
Sent: Tuesday, October 13, 2009 6:18 AM
To: Ruediger.Geib at telekom.de
Cc: Barry Constantine; ippm at ietf.org
Subject: Re: [ippm] Testing TCP Throughput Capacity in Operator Networks

Hi Ruediger,

Personal opinion, not discussed with Matt or AD's yet

> I'm not sure, whether IPPM is chartered to specify transport layer 
> performance metrics. I've checked the PMOL charter and it clearly says

> PMOL works above transport layer. A fair conclusion would be, that
IPPM 
> is chartered to investigate transport layer performance.

I looked at it from a slightly different angle.  Network capacity has
always been a topic of interest for IPPM, so this work would be in our
charter.  PMOL deals with higher layers, besides that (AFAIK) the PMOL
WG is in some form of hybernation.

So, my take on this is that we should see if there is interest to work
on this topic _and_ have a first version of a draft, we can decide later
where exactly this works fits in the IETF.

Henk


------------------------------------------------------------------------
------
Henk Uijterwaal                           Email:
henk.uijterwaal(at)ripe.net
RIPE Network Coordination Centre          http://www.xs4all.nl/~henku
P.O.Box 10096          Singel 258         Phone: +31.20.5354414
1001 EB Amsterdam      1016 AB Amsterdam  Fax: +31.20.5354445
The Netherlands        The Netherlands    Mobile: +31.6.55861746
------------------------------------------------------------------------
------

Belgium: an unsolvable problem, discussed in endless meetings, with no
          hope for a solution, where everybody still lives happily.
_______________________________________________
ippm mailing list
ippm at ietf.org
https://www.ietf.org/mailman/listinfo/ippm