[bind10-dev] statistics plan

[Naoki Kambe]

From: JINMEI Tatuya / 神明達哉 <jinmei at isc.org>
Date: Thu, 26 Jul 2012 17:24:58 -0700
> Could you provide a list of these things in this thread?  My memory is
> poor and I already forgot some details on what we agreed to.

Regarding items to do in statistics, we three basically refered from
this minutes:
http://bind10.isc.org/wiki/April2012MeetingMinutes#Estimates

    differences - only sending changes since last statistics query
    hybrid model (push/pull) - have push at the moment (modules send to statistics daemon). Processes will push (but not wait for an answer) if a table overfills. Daemon will pull otherwise.
    support aggregate reporting - can collect statistics from multiple modules (e.g. multiple auto processes)
    per-zone statistics (tracking and querying)
    Adding counters into the auto server 

Also statistics plan in 2011 is seen in:
http://bind10.isc.org/wiki/Year3Plan#ReleasePlanforStatistics
This is what we three originally planned.

Based on the above items and our planning, we three planned
mainly following four tasks by end of Sept.:

  1. Reverse the direction of the statistical request:

     Stats module asks each module for statistics instead of
     listening. This is for preventing the timeout issue which
     occasionally happens in Auth.  This is based on the second and
     third item above, but we don't think the hybrid model is always
     necessary. Because we three think the implement would become
     complected and either of them would become unused.  But we might
     need to discuss more each other about this. This task is related
     to the tickets #2135, #2136, #2137 and #2138.


  2. Implement the rest of counters to Auth, Xfrout and Xfrin:

     This is based on the fifth one of agreed items and the plan of
     2011.  This task is related to the tickets #2155, #2156, #2157
     and #2158.
     
  3. Implement per-zone counters to Auth, Xfrout and Xfrin:

     This is based on the fourth one of agreed items and the plan of
     2011.

  4. Support incremental updates between Stats and each module:

     This is aimed for reducing amount of data to be sent. This is
     based on the first one of agreed items. 

The order depends on both what we three think as important and what we
three can easily do. This priority can be matched with your
imagination?  Please let us know if you have any comment to the
priority.

> I appreciate that, and I have no doubt about your seriousness.  Still,
> we developers (not excluding myself) are always over optimistic.
> However sure we are, our history and experiences have proved we are
> always too optimistic:-). 

Thank you for the consideration. We three think we should also
consider thoughts of other developers. We three don't absolutely want
to do something high-risk. But we three think we should do our best
efforts to others until end of Sept.. Otherwise we three might regret.

Regards,

Naoki Kambe