Mercurial > hg
view CONTRIBUTORS @ 30206:d105195436c0
wireproto: compress data from a generator
Currently, the "getbundle" wire protocol command obtains a generator of
data, converts it to a util.chunkbuffer, then converts it back to a
generator via the protocol's groupchunks() implementation. For the SSH
protocol, groupchunks() simply reads 4kb chunks then write()s the
data to a file descriptor. For the HTTP protocol, groupchunks() reads
32kb chunks, feeds those into a zlib compressor, emits compressed data
as it is available, and that is sent to the WSGI layer, where it is
likely turned into HTTP chunked transfer chunks as is or further
buffered and turned into a larger chunk.
For both the SSH and HTTP protocols, there is inefficiency from using
util.chunkbuffer.
For SSH, emitting consistent 4kb chunks sounds nice. However, the file
descriptor it is writing to is almost certainly buffered. That means
that a Python .write() probably doesn't translate into exactly what is
written to the I/O layer.
For HTTP, we're going through an intermediate layer to zlib compress
data. So all util.chunkbuffer is doing is ensuring that the chunks we
feed into the zlib compressor are of uniform size. This means more CPU
time in Python buffering and emitting chunks in util.chunkbuffer but
fewer function calls to zlib.
This patch introduces and implements a new wire protocol abstract
method: compresschunks(). It is like groupchunks() except it operates
on a generator instead of something with a .read(). The SSH
implementation simply proxies chunks. The HTTP implementation uses
zlib compression.
To avoid duplicate code, the HTTP groupchunks() has been reimplemented
in terms of compresschunks().
To prove this all works, the "getbundle" wire protocol command has been
switched to compresschunks(). This removes the util.chunkbuffer from
that command. Now, data essentially streams straight from the
changegroup emitter to the wire, possibly through a zlib compressor.
Generators all the way, baby.
There were slim to no performance changes on the server as measured
with the mozilla-central repository. This is likely because CPU
time is dominated by reading revlogs, producing the changegroup, and
zlib compressing the output stream. Still, this brings us a little
closer to our ideal of using generators everywhere.
author | Gregory Szorc <gregory.szorc@gmail.com> |
---|---|
date | Sun, 16 Oct 2016 11:10:21 -0700 |
parents | c29efd272395 |
children |
line wrap: on
line source
[This file is here for historical purposes, all recent contributors should appear in the changelog directly] Andrea Arcangeli <andrea at suse.de> Thomas Arendsen Hein <thomas at intevation.de> Goffredo Baroncelli <kreijack at libero.it> Muli Ben-Yehuda <mulix at mulix.org> Mikael Berthe <mikael at lilotux.net> Benoit Boissinot <bboissin at gmail.com> Brendan Cully <brendan at kublai.com> Vincent Danjean <vdanjean.ml at free.fr> Jake Edge <jake at edge2.net> Michael Fetterman <michael.fetterman at intel.com> Edouard Gomez <ed.gomez at free.fr> Eric Hopper <hopper at omnifarious.org> Alecs King <alecsk at gmail.com> Volker Kleinfeld <Volker.Kleinfeld at gmx.de> Vadim Lebedev <vadim at mbdsys.com> Christopher Li <hg at chrisli.org> Chris Mason <mason at suse.com> Colin McMillen <mcmillen at cs.cmu.edu> Wojciech Milkowski <wmilkowski at interia.pl> Chad Netzer <chad.netzer at gmail.com> Bryan O'Sullivan <bos at serpentine.com> Vicent SeguĂ Pascual <vseguip at gmail.com> Sean Perry <shaleh at speakeasy.net> Nguyen Anh Quynh <aquynh at gmail.com> Ollivier Robert <roberto at keltia.freenix.fr> Alexander Schremmer <alex at alexanderweb.de> Arun Sharma <arun at sharma-home.net> Josef "Jeff" Sipek <jeffpc at optonline.net> Kevin Smith <yarcs at qualitycode.com> TK Soh <teekaysoh at yahoo.com> Radoslaw Szkodzinski <astralstorm at gorzow.mm.pl> Samuel Tardieu <sam at rfc1149.net> K Thananchayan <thananck at yahoo.com> Andrew Thompson <andrewkt at aktzero.com> Michael S. Tsirkin <mst at mellanox.co.il> Rafael Villar Burke <pachi at mmn-arquitectos.com> Tristan Wibberley <tristan at wibberley.org> Mark Williamson <mark.williamson at cl.cam.ac.uk>