Andrea's notes.
Present: ZM, AK, AV
Over the phone: AS, RB
New release
- ZM just circulated 0.1.5
- 0.1.5 includes only the transport layer improvement
- differently from what was proposed last week, 0.1.5 does not yet
include caching (had some MT problems, will send that later as 0.1.6)
Discussion about writing to sockets
- ZM to AS: is proxy cache writing packets to sockets all at once?
- AS: only if it is a new payload; if the payload is cached, writing
to sockets is done in two steps, first the (new) header, then the
(cached) payload
- AV to AS: try to write all at once (combine header and payload in
memory), because ZM had observed performance problems in a similar case
- RB: may be useful to look at paper on 'Nagle vs Delayed Ack'
if not yet done (http://www.stuartcheshire.org/papers/NagleDelayedAck)
AOB
- AS created a new tag 0.0.6, we can start using it
- AK mainly working on the thesis
- AK had no time to work on profiling stress tests this week
Plans
- ZM release 0.1.6 with caching (after solving MT issues)
- ZM plan stress tests with multiple clients after 0.1.6;
AK will also help writing stress tests
- ZM use stress tests to investigate the hang discussed last 2 weeks
- AK continue to write thesis
- Application Area talk postponed to october 8th
There are minutes attached to this event.
Show them.