Re: [sip-overload] draft-ietf-soc-overload-rate-control-04.txt

["NOEL, ERIC C (ERIC C)" <[e-mail]>]

James,

My apologies for having missed your comment.

"For the rate control, the server could calculate the arrival rate from each communicating client so that it could allocate the overall target SIP request rate to the clients based on their arrival rates known to the server. "
[EN] Yes, instead of uniformly distributed target SIP request rate across clients, the overloaded server could distributed the request rate function of the requests it gets from the clients.
In our propose we eluded to your suggestion through the following sentence (section 3.4):
"The server must allocate a portion of the target SIP request rate to each of its client. The server may set the same rate for every client, or may set different rates for different clients."

"But another option is for the client to "optionally" include its calculated arrival rate in its request to the server when rate control related parameters are present.  Should this option be evaluated/included to relieve the server from doing the arrival rate calculations.  This would be beneficial to a server when it receives the requests from many clients."
[EN] Your suggestion is currently not supported by draft-ietf-soc-overload-control (there are no parameters that would allow clients to inform server of their respective arrival rates) from which draft-ietf-soc-overload-rate-control depends upon. You may want to propose this option to draft-ietf-soc-overload-control authors. But in practice, servers estimate percent blocking or rate dynamically based on their overload state so as to have clients throttle at a level close to servers overload onset level.

Thanks,

Eric Noel
AT&T Labs, Inc.
Rethink Possible

Network Design and Performance Analysis
200 South Laurel Avenue, D5-3D19
Middletown, NJ 07748
P: 732.420.4174
[e-mail]<mailto:[e-mail]>

From: Yu, James [mailto:[e-mail]]
Sent: Wednesday, July 10, 2013 9:01 AM
To: NOEL, ERIC C (ERIC C)
Cc: [e-mail]; [e-mail]; [e-mail]
Subject: RE: [sip-overload] draft-ietf-soc-overload-rate-control-04.txt

Noel,

I mentioned in my message to Janet on 7/2:

For the rate control, the server could calculate the arrival rate from each communicating client so that it could allocate the overall target SIP request rate to the clients based on their arrival rates known to the server.  But another option is for the client to "optionally" include its calculated arrival rate in its request to the server when rate control related parameters are present.  Should this option be evaluated/included to relieve the server from doing the arrival rate calculations.  This would be beneficial to a server when it receives the requests from many clients.

I suggest that the I-D adds that option to allow the client to indicate its current arrival rate (towards the receiving server) in the SIP request.

James

From: NOEL, ERIC C (ERIC C) [mailto:[e-mail]]
Sent: Tuesday, July 09, 2013 4:44 PM
To: Yu, James; Janet P Gunn
Cc: [e-mail]; [e-mail]; [e-mail]
Subject: RE: [sip-overload] draft-ietf-soc-overload-rate-control-04.txt

Janet, James,

Thanks you for your valuable comments and discussion.  I tried to capture all resolution in the following text (based on James word document with track changes and imbedded comments enabled).

+ Abstract section:
- Agreed with suggested changes
- Please note I will also need to make further changes to remove all references per Christer Holmberg comment

+ Section 1:
- Agreed with suggested changes

+ Section 3.1:
- Agreed with suggested changes

+ Section 3.2: - Please note the section title will become "Via header field parameters for overload control " per Christer Holmberg comment
- Agreed with suggested changes excluding title that will change per previous bullet

+ Section 3.3:
- Agreed with suggested changes

+ Section 3.4:
- ID:  "Note that the target SIP request rate is a max rate that may not be attained by the arrival rate at the client, and the server cannot assume that it will."
   [JY] Not clear what value this paragraph tries to add.  Is it saying that the client's arrival rate may be lower than the target SIP request rate?
   [JG] Yes + supporting example (see below in thread)
  Agreed with Janet and supporting example. Per James request, I will add the following text (inspired from Janet's illustrative example):
   "In other words, when multiple clients are being controlled by an overloaded server, at any given time some clients may receive requests at a rate below its target SIP request rate while others above that target rate. But the resulting request rate presented to the overloaded server will converge towards the target SIP request rate."
- Agreed with other suggested changes

+ Section 3.5.1:
- ID: "And the larger the difference between TAU1 and TAU2, the closer to the control is to strict priority."
  [JY] Suggest changing into "And the larger the difference between TAU1 and TAU2, the closer the control is to strict priority queuing."
  [JG] Suggest changing into "And the larger the difference between TAU1 and TAU2, the closer the control is to strict priority treatment."
  [EN] Agreed with Janet's suggestion
- Agreed with other suggested changes

+ Section 4:
- Agreed with suggested changes

+ Section 5:
- Please note that based comments from per Christer Holmberg and Janet Gunn, the following was tentatively agreed
   Replace oc-value = "NaN" / oc-num    by    oc = "oc" [EQUAL oc-num]
- Agreed with other suggested change

+ Section 7:
- [JY] Should "rate" in oc-algo be registered with IANA?
  This issue needs to be addressed by draft-ietf-soc-overload-control authors

Please note I will wait until expiration of WGLC prior updating our draft RFC. Once again your comments and/or suggestions are most appreciated.

Thanks,

Eric Noel
AT&T Labs, Inc.
Rethink Possible

Network Design and Performance Analysis
200 South Laurel Avenue, D5-3D19
Middletown, NJ 07748
P: 732.420.4174
[e-mail]<mailto:[e-mail]>

From: [e-mail]<mailto:[e-mail]> [mailto:[e-mail]] On Behalf Of Yu, James
Sent: Tuesday, July 02, 2013 9:42 AM
To: Janet P Gunn
Cc: [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Subject: Re: [sip-overload] draft-ietf-soc-overload-rate-control-04.txt

Janet,

For my comment on section 3.4 (the first one below), the current text does not provide any value.  The client's arrival rate could be well below the target SIP request rate when its load is light so the fact that the client may not achieve the target SIP request rate (the max. rate it is allowed to send to the server) is well understood.  But with your explanation on the "delta" part, the text then makes sense.  Please add some discussions on the "delta" aspect so that even if the average arrival rate at the client is higher than the target SIP request rate the client at times may not send more than what the target SIP request rate allows due to the fluctuation of the arriving requests at the client.

For the comment on section 3.5.1, I agree with your proposed change.

For the rate control, the server could calculate the arrival rate from each communicating client so that it could allocate the overall target SIP request rate to the clients based on their arrival rates known to the server.  But another option is for the client to "optionally" include its calculated arrival rate in its request to the server when rate control related parameters are present.  Should this option be evaluated/included to relieve the server from doing the arrival rate calculations.  This would be beneficial to a server when it receives the requests from many clients.

James

From: Janet P Gunn [mailto:[e-mail]]
Sent: Friday, June 28, 2013 3:34 PM
To: Yu, James
Cc: [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Subject: Re: [sip-overload] draft-ietf-soc-overload-rate-control-04.txt

James,

It is a little hard to respond in email when your comments are in a separate document, but here goes.


section 3.4
ID says:
"Note that the target SIP request rate is a max rate that may not be
   attained by the arrival rate at the client, and the server cannot
   assume that it will."

Your comment :
"Not clear what value this paragraph tries to add.  Is it saying that the client's arrival rate may be lower than the target SIP request rate?  "

Yes.

Suppose the server want to limit the total rate of arriving SIP messages to 100 / sec, and has 10 clients.  Each client has a high variance in its message rate, but together they are well above 100 messages per sec

If it sets the rate at 10 messages per second for each of the clients, it will almost certainly end up  with an overall average of less than 100 messages per sec, because some clients will be in a "lull" while others are busy.  This is good from a throttling perspective, but, assuming messages are correlated with revenue, bad/wasteful from a revenue, or overall productivity perspective.  So the server might want to set the rate per client to 10 + delta, where delta is going to be very specific to operating environment.

---
In section  3.5.1, bottom of page 8
ID says:
"And the larger
   the difference between TAU1 and TAU2, the closer to the control is
   to strict priority."

You propose changing it to:
"And the larger
   the difference between TAU1 and TAU2, the closer  the control is
   to strict priority queuing."

I agree with taking out the redundant "to".  But I disagree wit adding "queuing".  There is no queuing, priority or otherwise involved.

You could say:
"And the larger
   the difference between TAU1 and TAU2, the closer  the control is
   to strict priority treatment."

where " strict priority treatment" would refer to the case where non-priority messages are restricted to a total (priority + non-priority) rate of 10 messages per second, but priority messages can continue to be sent as long as the total (priority + non-priority) rate is less than 12 messages per second.

At least, I think that is what Eric and  Philip are trying to say.

Janet





This is a PRIVATE message. If you are not the intended recipient, please delete without copying and kindly advise us by e-mail of the mistake in delivery. NOTE: Regardless of content, this e-mail shall not operate to bind CSC to any order or other contract unless pursuant to explicit written agreement or government initiative expressly permitting the use of e-mail for such purpose.



From:        "Yu, James" <[e-mail]<mailto:[e-mail]>>
To:        "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Date:        06/28/2013 08:10 AM
Subject:        [sip-overload] draft-ietf-soc-overload-rate-control-04.txt
Sent by:        [e-mail]<mailto:[e-mail]>
________________________________



Salvatore,

Please see the attachment for my comments.

I pasted the text to a word document to trace/show the proposed changes and comments.

Regards,

James
 [attachment "comments on draft-ietf-soc-overload-rate-control-04.docx" deleted by Janet P Gunn/USA/CSC] _______________________________________________
sip-overload mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/sip-overload
________________________________
No virus found in this message.
Checked by AVG - www.avg.com<http://www.avg.com>
Version: 2013.0.2904 / Virus Database: 3204/6452 - Release Date: 06/30/13
________________________________
No virus found in this message.
Checked by AVG - www.avg.com<http://www.avg.com>
Version: 2013.0.2904 / Virus Database: 3204/6478 - Release Date: 07/09/13