RE: question on draft-ietf-tewg-measure-04.txt

["Lai, Wai S (Waisum), ALABS" <wlai@att.com>]

Jing,

  Thanks for your comments on our draft.  We'll add appropriate text to

clarify the points you raised.

  As Richard described below, delay variation can be measured by either

1-point or 2-point methods, depending on the intent.  Y.1541 is already

referred to in our draft, but we will make an explicit reference to its

Appendix II in Section 10.1.  (As indicated in the 3rd paragraph in page

4, our intent is to avoid duplication of effort, by referring to other

standards documents as far as possible.)

  As defined in Section 4.2 (and also later in Section 9.1), throughput

and goodput are the same in our draft.  Also, as indicated in Note (5)

in Section 10.1, traffic sent by source nodes are indeed measured, as

you pointed out.

  Our draft's focus is currently mainly on identifying additional

measurements for TE as described by the TEWG charter.  (Non-goals are

described in the second paragraph in page 4.)  While measurement

precision/accuracy are important issues, the authors feel that such

issues should be dealt with in future drafts.  As an example of

precision/accuracy issues, see RFC 2679, section 3.7, "Errors &

Uncertainties".  There, the accuracy issues for one-way delay are

dealt with in great detail.  Each measure potentially has its own

unique precision/accuracy issues, and to deal with them in the

framework draft would make it quite voluminous.
However, I would like to hear if there is a general
interest in this area.

Thanks, Wai Sum

-----Original Message-----
From: Dr. Richard W. Tibbs [mailto:tewg@oakcitysolutions.com]
Sent: Monday, January 13, 2003 2:57 PM
To: jshen@cad.zju.edu.cn
Cc: te-wg@ops.ietf.org
Subject: Re: question on draft-ietf-tewg-measure-04.txt


Greetings, and thanks for your question. Apologies if you get this twice.

A one-point delay is sometimes desirable in several situations.
First, ITU Y.1540 states in appendix II:
<quote>
II.4 Guidance on Applying the Different Parameters
Guidance that serves the practical side of measurement is as follows:
- When synchronized clocks are not possible (or temporarily unavailable)
in measurement devices,
1. 1-point Packet Delay Variation (PDV) is a possible substitute for
1-way delay range/histogram, applicable for measurements on packet
streams with periodic sending times (once the reference arrival time is
appropriately set).
2. IPPM Inter-packet delay variation is applicable to all traffic flow
types.
</quote>
Second,
when flows must be regulated to ensure QoS, as for example in RSVP, a
similar reference arrival time exists based on the negotiated traffic
envelope parameters, e.g.,
1/peak_rate or 1/mean_rate, etc.
Our draft is intended as a framework document, and we wish to be quite
general in our definitions, so that subsequent drafts may focus on such
areas for TE measurement.


Hope this helps.

Yours truly,
Dr. Richard W. Tibbs.
Radford University
Dept. of Information Technology



Jing Shen wrote:

>Hi,
>
>
>in sec10.1, there is a "Delay Variation" measured at Interface,
>to my understanding IPDV is a type of measurement at e2e level or link
>level. Why does the draft propose that? and how to measure it?
>
>I'm not clear what's the relationship between "throughput" and "goodput"
>used in the draft. If they all refer to the amount  of traffic
>successfully
>delivered,  do we need to measure network traffic sent by source nodes?
>
>
>And, what about the measurement precision? I means should the accuracy
>of measurement be discussed?
>
>
>
>regards
>