RE: [Sipping] [PMOL] SIP End-to-End Performance Metrics

["Marian Delkinov" <[e-mail]>]

Daryl,

Regarding adding 401/407 to the numerator of SEER, my opinion is that it
depends on the idea behind SEER definition.
1) If SEER should complement SER, but exclude the called party
behaviour, then I support keeping SEER as it is. This case matches the
NER definition in PSTN. 
2) If SEER should represent the ability of the infrastructure to
establish sessions, then I would support adding 401/407 codes to the
numerator. (If you remember during our discussion over the last year, I
proposed adding all 4xx codes except 408 in the numerator of SEER, but
we then agreed that it would cover much larger scope than just excluding
the customer behaviour. Perhaps we have to define another metric to
cover those cases, if the WG supports that.)  

Best regards!
Mario.

-----Original Message-----
From: Daryl Malas [mailto:[e-mail]] 
Sent: Monday, 21 January, 2008 16:19
To: Marian Delkinov; Hadriel Kaplan
Cc: [e-mail]; [e-mail]
Subject: RE: [Sipping] [PMOL] SIP End-to-End Performance Metrics

Mario,

I agree.  This is the traditional difference between SER and SEER, much
like the difference between ASR and NER (ITU-T e.411).  Now, one thing
that might make sense is to add 401/407 to the numerator of the SEER
metric algorithm.

--Daryl

-----Original Message-----
From: Marian Delkinov [mailto:[e-mail]]
Sent: Monday, January 21, 2008 3:45 AM
To: Daryl Malas; Hadriel Kaplan
Cc: [e-mail]; [e-mail]
Subject: RE: [Sipping] [PMOL] SIP End-to-End Performance Metrics

Daryl and Hadriel,

I have some comments regarding SER and SEER:

> SER - need to specify that a 401/407 would also be subtracted in that 
> denominator, or better yet the Invite would not be double-counted 
> period.  Also, need to point out that "# of INVITE" is for 
> new-sessions only (ie, ones without to-tags) - not re-Invites. (I know

> it's obvious, but ya never know)

If we go back to the analogy with PSTN and the ASR defined by ITU, I
would suggest leaving SER as it is. Thus the metric would really be
useful for estimating the %revenue compared to the total volume of
invites, regardless of the reason for failure (as it is in PSTN). For
analysis of the different aspects of the infrastructure efficiency I
suggest using complementary to SER metrics, as SEER (already defined),
and why not defining some more.

SEER: Section 3.7 defines SEER, however the formula still uses SER=...
I suggest using "SEER" in the formula in order to avoid confusion with
SER.

Best regards!
Mario.
 

-----Original Message-----
From: Daryl Malas [mailto:[e-mail]]
Sent: Friday, 18 January, 2008 17:17
To: Hadriel Kaplan
Cc: [e-mail]; [e-mail]
Subject: Re: [Sipping] [PMOL] SIP End-to-End Performance Metrics

Hadriel,

Comments in-line...

On 1/17/08, Hadriel Kaplan <[e-mail]> wrote:
>
> Hey Daryl,
> Comments:
>
> Failed RRD - I'm a little confused by the wording of how we decide
what constitutes a failed registration.  Clearly a 401/407 challenge
round does not, but you state a second round would.  Why?  It's totally
possible it's not a failure.
> I'm also confused what the value/purpose of measuring this delay is.
> The section says measuring failed RRD is useful for detecting problems
with reaching the intended Registrar - that would be true if you counted
transaction timeouts, not request/response delays for auth failures, no?

[DM]
I agree this could create some false indications of failures.  I can
update the metric to only determine a failed condition based on a
timeout scenario.

Also, I'm going to put together a suggested list of metrics from this
draft.  The topic has come up several times on how many metrics should
be included.  Since, no one has suggested a list, I will throw one out.
>
> Failed SRD - need to specify a 401/407 is not a failed SRD (you say
any 4xx).
>
[DM]
I agree and will update the draft.

> Successful SDD - you say at the end "In these two examples, TB and TS
are the same even if the UAC/UAS receives the indicated messages instead
of sending them."  I'm pretty sure the delay in receiving a Bye and
sending the 200 ok for that is not really useful. :)  Or at least a very
different issue than that measured from sending a Bye and receiving a
200.
>
[DM]
I agree.  At one point, someone requested this to be exhaustive in all
conditions.  I can remove this to indicate relevance only in the
situation of initiating the Bye from either the UAC or UAS.

> Failed SDD - what do we count the time of a Bye to a 4xx (but not
401/407) or 5xx as?  Since those are the failure conditions for the
other cases, this seems a bit odd to no longer count it as one here.
[DM]
I'm not sure I understand this question.
>
> Failed SDT - this seems to be at odds with the definition of
successful SDT.  Successful SDT is stopped on sending the Bye, but this
one is stopped on timing out that sent Bye. (in other words, every call
will thus count as a successful SDT, and some will also count as a
failed one)  Also, what good is this failed SDT info?  You already have
Failed SDD.  It seems to me an SDT can only be of type "successful", and
the timer should be stopped at send/receipt of Bye.
[DM]
I agree.  I will update this to only be between the 200 and Bye.  It
should be the Bye regardless of whether or not it times out based on the
intiator.
>
> AHR - how does the UAC or UAS know this info?  I.e., how does the UAS
know what max-forwards the UAC started with, and how does the UAC know
what max-forwards the UAS received?
[DM]
The thought is the clients (or some other in-line monitoring device)
would have to capture this information at both ends and hold it in order
to correlate and make the determination.  This metric is a tricky one
and has come up much scrutiny.  It will work and could be very useful,
but will require some capture outside of SIP in all likelihood.
>
> SER - need to specify that a 401/407 would also be subtracted in that 
> denominator, or better yet the Invite would not be double-counted 
> period.  Also, need to point out that "# of INVITE" is for 
> new-sessions only (ie, ones without to-tags) - not re-Invites. (I know

> it's obvious, but ya never know)
[DM]
Are you saying 401/407 should be included with this sentence "...to the
total number of attempted INVITE requests less INVITE requests resulting
in a 3XX response..."?

>
> -hadriel
>
> > -----Original Message-----
> > From: Daryl Malas [mailto:[e-mail]]
> > Sent: Friday, December 14, 2007 5:42 PM
> > To: [e-mail]
> > Cc: [e-mail]
> > Subject: [Sipping] [PMOL] SIP End-to-End Performance Metrics
> >
> > At the conference, I introduced the SIP End-to-End Performance 
> > Metrics draft (draft-malas-performance-metrics-08) to the PMOL
working group.
> > Although this draft is referenced in the PMOL WG charter, I wanted 
> > to ask everyone to review the draft and provide feedback of whether 
> > or not you feel the current version is ready for the WG to accept as

> > a working group item.  A couple of questions came up, which I think 
> > should be answered regarding this consideration:
> >
> > Is the SIPPING WG content with the current set of metrics?
> >
> > Are there too many?
> >
> > Do these metrics capture the relevant concerns of performance 
> > regarding the SIP protocol?
> >
> > Are these metrics depicted accurately from a SIP protocol
perspective?
> >
> > In addition to these questions, I have a couple of tasks from the 
> > PMOL group to include in the next revision.
> >
> > Here is a link to the draft:
> > http://www.ietf.org/internet-drafts/draft-malas-performance-metrics-
> > 08.txt
> >
> > Thanks...
> >
> > Daryl
> >
> >
> >
> > _______________________________________________
> > Sipping mailing list  https://www1.ietf.org/mailman/listinfo/sipping
> > This list is for NEW development of the application of SIP Use 
> > [e-mail] for questions on current sip Use 
> > [e-mail] for new developments of core SIP
>
> ______________________________________________________________________
> This email has been scanned by the MessageLabs Email Security System.
> For more information please visit http://www.messagelabs.com/email 
> ______________________________________________________________________
>


_______________________________________________
Sipping mailing list  https://www1.ietf.org/mailman/listinfo/sipping
This list is for NEW development of the application of SIP Use
[e-mail] for questions on current sip Use
[e-mail] for new developments of core SIP


_______________________________________________
Sipping mailing list  https://www1.ietf.org/mailman/listinfo/sipping
This list is for NEW development of the application of SIP Use
[e-mail] for questions on current sip Use
[e-mail] for new developments of core SIP


_______________________________________________
PMOL mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/pmol