[tor-bugs] #1819 [Tor Relay]: Implement new metric on bidirectional use of connections

[Tor Bug Tracker & Wiki]

#1819: Implement new metric on bidirectional use of connections
-------------------------+--------------------------------------------------
 Reporter:  karsten      |       Owner:  karsten     
     Type:  enhancement  |      Status:  needs_review
 Priority:  normal       |   Milestone:              
Component:  Tor Relay    |     Version:              
 Keywords:               |      Parent:              
-------------------------+--------------------------------------------------

Comment(by karsten):

 Replying to [comment:1 nickm]:
 > What's the motivation on this one?  And why 10 seconds?

 Björn's and Florian's assumption is that bidirectional use of connections
 screws Tor's performance. For a more general design discussion of the
 motivation and the parameters, please ask them by replying to my or-dev
 message from July 12:

 http://archives.seul.org/or/dev/Jul-2010/msg00016.html

 > There's probably a better name than ConnStatistics here; lots of our
 other statistics could fairly be called connection statistics.
 ConnDirectionStats?  ConBidiStats?

 Going with !ConnDirectionStatistics.

 > BIDI_THRESHOLD and BIDI_INTERVAL and BIDI_FACTOR are uncommented.

 Done.

 > Doing a round trip to disk here feels a little kludgy; it could probably
 use unit tests too.

 I don't understand what you mean by "round trip to disk".

 Yes, this code could use unit tests. If you say that the unit tests in
 #1825 look good, I'll add similar tests for connection stats.

 > Looking at the code that generates the extrainfo, it looks like if we
 leave ConnStatistics on for a while, then turn if off for a year, then
 turn it on again, there could be a window when the old ConnStatistics from
 a year ago get written in the extrainfo.  Is that true?  If so, is that
 really what we want?

 No, load_stats_file() only returns a string if the stats have been written
 in the past last 25 hours.

 > In any case, it looks like a new feature; is there a good reason to add
 it to 0.2.2.x rather than waiting for 0.2.3.x?

 Waiting for 0.2.3.x is fine.

 But before we ignore this feature for the next two months, do you have an
 answer to my question from the task description?

 > An open question is whether we should distinguish between connections to
 other relays and to clients. I wonder if there's an easy way to tell the
 two connection types apart.

 See updated branch bidistats in my public repository. I also added a
 ChangeLog entry and a man page entry.

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/1819#comment:2>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online