comments on draft-ietf-psamp-framework-09.txt

[Jennifer Rexford <jrex@research.att.com>]

Hi folks,

I just read through the latest draft of "A Framework for Packet
Selection and Reporting" (draft-ietf-psamp-framework-09.txt) and
think the document looks great.  I have just a few minor comments,
and one suggestion for a few sentences of additional text (on the
subject of congestion-aware transport).  More below...

-- Jen

- page 11, Section 4.1: bullet "*" is missing in "Non-Contingency".
Also, the term "non-contingency" seems a bit strange.  The text for
the bullet could add something about why this requirement is included
(e.g., "in order to limit the complexity of the selection logic").

- page 12, Section 4.3: in the bullet on "Congestion Avoidance," it
would be good to include a forward pointer to Section 8.2 where this
issue is discussed further.

- page 13-14: Under "content-dependent sampling" or "non-uniform
probabilistic sampling," it may be helpful to include "sampling
packets in proportion to their length" as an example.

- page 15: At the end of Section 5.2, I didn't understand the
sentence "However, if selection not based on routing state has
reduced down from line rate...."  I had trouble parsing it.

- page 21: Section 8.2 talks about the need for unreliable,
congestion-aware transport, and shifts in the last paragraph to
talking about IPFIX rather than PSAMP.  I wasn't sure what the the
IPFIX paragraph was trying to get at, and if/how it relates to the way
PSAMP should achieve congestion-aware transport.  Somewhere in the
document (either in 8.2, 8.5, or in 10.2), it would be worth
commenting on lighweight ways to achieve the goal of congestion-aware
transport.  Section 8.2 could end by saying something like
"Congestion-aware transport of PSAMP records could be achieved in
various ways.  For example, the PSAMP device could have an export rate
limit (as discussed in Section 8.5) that is configured by the
controller based on observations of the packet loss rate in delivering
PSAMP records.  This would ensure that the transport rate adapts in
the presence of congestion without requiring the PSAMP device to
receive acknowledgment packets or implement the adaptation algorithm
directly."  I think this is important for conveying how the requirement
of congestion-aware transport can be satisfied without introducing
a lot of complexity in a PSAMP device that much operate at high speed.

- page 32: My room number is A139.




--
to unsubscribe send a message to psamp-request@ops.ietf.org with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/psamp/>