Reviewer: Elwyn Davies
Review result: Not Ready
I am the assigned Gen-ART reviewer for this draft. The General Area
Review Team (Gen-ART) reviews all IETF documents being processed
by the IESG for the IETF Chair. Please treat these comments just
like any other last call comments.
For more information, please see the FAQ at
<
https://trac.ietf.org/trac/gen/wiki/GenArtfaq>.
Document: draft-ietf-mpls-rfc6374-sfl-??
Reviewer: Elwyn Davies
Review Date: 2021-02-02
IETF LC End Date: 2021-01-26
IESG Telechat date: Not scheduled for a telechat
Summary: Not Ready. Apologies that this review is rather late, but I found this
document extremely hard to work with. There appear to be a number of areas where
the work is rather too much in progress rather than ready for publication as an RFC.
That was some old text from an early version that was missed.
I also found it very difficult, not just as someone who is not at all familiar with this
area of work, but at a basic technical level to work out what the protocol was going
to be able to achieve and whether a LSR would garner the information it appeared
to need to deliver what was clamed.
This is a document where you need to understand MPLS, coloured marking and packet delay characteristics, but I think anyone seeking to deploy this would already be familiar with that.
Part of this appeared to be due to multiple
names being used for the same thing and being used with other than their natural
meaning particulaly in sections 7.1 and 7,2.
Major Issues:
s7, What is being standardized?:
A number of methods are described. The expectation is that the MPLS
WG possibly with the assistance of the IPPM WG will select one or
maybe more than one of these methods for standardization.
I find this statement very confusing. This document is intended for
standards track, so if it goes ahead as is, the three methods are
standardised and implementors would be expected to provide support for
all of them unless there are to be words to indicate that not all need
to be supported. Is this the intention? Or is it that this document
should only support the methods chosen by the MPLS working group? In
the latter case, this document is definitely not ready for
standardization; I assume the unused method(s) would be removed in this
case. Otherwise the second sentence is speculative and should be removed.
I have changed this text in response to other comments received:
It now says
A number of methods are described. Each of these methods has different
characteristics and different processing demands on the packet forwader.
The choice of method will depend on the type of diagnostic that the operator seeks.
s7, Title, purpose and general method:
Note that I have very limited knowledge of this area of performance
measurement so there may be misunderstandings here. However, given that
caveat, I did not find the document very helpful in enlightening me and
a considerable amount of background reading was needed to try and
determine what was going on.
It is always difficult to get the balance right between a concise document for subject matter experts and a detailed description.
Firstly, I assume that this section covers the 'additional techniques'
mentioned in the Abstract
That term does not seem to be in the abstract.
and described as 'more sophisticated
measurements' in s1. [Perhaps common phraseology would be desirable
between the two cases.] I suggest a sentence to make this clear would
be desirable.
I am afraid I cannot see the conflict that you are concerned about.
Secondly, AFAICS these techniques are basically about measuring and
communicating delay jitter in various ways.
SB> No, Method 1 is measuring jitter. Method 2 is measuring delay as is Method 3
It might be helpful to
link what is being offered here with RFC 5481 and the discussion of
delay variation measurement in RFC 6374.
SB> I think we need to assume that the reader is familiar with RFC6374
Section 7.1 is, as I
understand it, covering IPDV measurement using (in general) normal
service packets rather than just specialised RFC 6374 packets and
working primarily on one LSR. I assume that the technique in s7.2 is
primarily a means for reporting measurements derived from s7.1 and/or
s7.4, but given that actual delays are mentioned rather than
inter-packet gaps, the
SB> All measurements take place on user service data using SFL to
SB> indicate different groups of packets. We are using RFC6374 to
SB> trigger measurements and collective results.
s7.1: After the first sentence, the first paragraph talks about delay.
Since the receiving LSR has no knowledge of the transmission time of
each individual packet, it is not possible for the LSR to calculate
actual delays without additional information - I take it that the
packets are not intended to be RFC 6374 Delay Measurement Packets as
these would require corresponding responses which would contravene the
query interval setting and there does not appear to be a way for the
LSR doing the measurements to be told the inter-packet transmission
interval. Should this be written in terms of inter-packet gaps rather
than delays throughout?
SB> 7.1 is measuring the inter packet gaps so as you say is measuring
The variation in the delay rather than the absolute delay. However this
Is made clear in the text.
Further, The first paragraph describes two
methods of operation without saying which one should be standardised or
AFAICS providing a selection flag to allow either to be used.
SB> We could do that but there is a need for the operator to configure
SB> other characteristics of the measurement, for example the size of the
SB> time increments that the buckets represent, so this would just be another
SB> such characteristic. The math in the analytics engine to convert one
SB> method into the other is trivial (the difference in the techniques is
SB> about collection hardware optimisation) so I don’t think we need to
SB> pick one,
It seems to me that an outline of how this facility might be used is
pretty much essential. Would I be right in thinking that to measure the
delay jitter between a source LSR (S) and destination LSR (D), the
operator decides to send a set of packets at equally spaced intervals
from S to D and decides on the interval and the number of packets. S
then issues a Query setting the query interval to a time greater than
that needed to send the set of packets and using the Bucket Jitter
Measurement Message to set the bucket delay intervals around the
sending interval according to the Operator's expectations of the
network. D then sets up to measure the inter-packet delays up until the
next Bucket Jitter Measurement message arrives after the elapse of the
query interval when D returns the profile of inter-packet delays.
Does the arrival of this second Bucket Jitter Measurement Message
trigger a further set of measurements? And if so, how is the sequence
ended?
SB> No, you send packets in one color then you change color and then
SB> send an Query message and the response refers to the set of packets
SB> before the colour changed.
SB> The hardware continuously makes the measurement and the
SB> measurement system collects the results when it wants a test result.
s9.1: This section is headed by an Editor's Note saying that the section
needs review which may alter the format of the TLV. It is thus
impossible to say if this section is ready.
SB> That is a note I had forgotten to remove
Minor Issues:
s7.2: As with s7.1, there seems to be some confusion bettween delay and
inter-packet gap.
Nits/editorial comments:
Abstract: The primary purpose of this document, as set out in s1, is to
extend RFC 6374 to cover general MPLS networks rather than primarily
MPLS-TP networks and in particular to add support for
multi-point-to-point LSPs. I think that it would be helpful for the
casual reader to make this somewhat clearer in the abstract. I suggest:
OLD:
This document describes a method of making RFC6374 performance
measurements on flows carried over an MPLS Label Switched path. This
allows loss and delay measurements to be made on multi-point to point
LSPs and allows the measurement of flows within an MPLS construct
using RFC6374.
NEW:
RFC 6374 describes methods of making loss and delay measurements on
Label Switched Paths (LSPs) primarily as used in MPLS TransportProfile (MPLS-TP)
networks. This document describes a method of making RFC6374 performance
measurements on flows carried over general MPLS LSPs. In particular, it extends
the technique to allow loss and delay measurements to be made on multi-point to point
LSPs and introduces some additional techniques to allow more sophisticated
measurements to be made in both MPLS-TP and general MPLS networks.
ENDS
SB> Thank you that is a good proposal
s1, bullet 4: Would it be helpful to refer to RFC 7190 with respect to
aggregation?
SB> Yes, I will add the ref.
s1, bullet 5: s/counter again/counter, again/
SB> Fixed
s3, last sentence: s/co-responding/corresponding/ [co-responding means
responding together rather than matching. Look up co-respondent in
cases of adultery in the divorce courts!]
SB> Thanks. Co-responding seems like a good term to get into a protocol description.
s3, last sentence: s/packet/packets/
SB> Fixed
s4, para 1: Expand TC: s/TC/Trafic Class (TC)/
SB> Fixed
s5, para 1: s/proxy data service packets Section 4./proxy data service
packets (see Section 4)./
SB> fixed
s5, para 2: s/This it is/Thus it is/
SB> Fixed
