[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20180914142309.6789-1-paolo.valente@linaro.org>
Date: Fri, 14 Sep 2018 16:23:06 +0200
From: Paolo Valente <paolo.valente@...aro.org>
To: Jens Axboe <axboe@...nel.dk>
Cc: linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
ulf.hansson@...aro.org, linus.walleij@...aro.org,
broonie@...nel.org, bfq-iosched@...glegroups.com,
oleksandr@...alenko.name, Paolo Valente <paolo.valente@...aro.org>
Subject: [PATCH BUGFIX/IMPROVEMENT 0/3] bfq: one fix and two important improvements for throughput
Hi Jens,
the second and third patch in this series provide two important
improvements in bfq's ability to boost throughput with random I/O. The
benefits of the second patch concern I/O control, and are described in
detail in this LWN article [1] (and briefly in the commit message
itself). The benefits of the other patch should be
straightforward. Finally, the first patch fixes an I/O-control bug,
found while making the second patch.
These patches modify somehow core operations of bfq, so, before
submitting them, I have tested them a lot, and have had them tested by
other people too. In particular, during these months, they have been
tested in systems ranging from PCs to development platforms.
Thanks,
Paolo
[1] https://lwn.net/Articles/763603/
Paolo Valente (3):
block, bfq: correctly charge and reset entity service in all cases
block, bfq: inject other-queue I/O into seeky idle queues on NCQ flash
blok, bfq: do not plug I/O if all queues are weight-raised
block/bfq-iosched.c | 78 +++++++++++++++++++++++++++++++++++++++++++++++------
block/bfq-iosched.h | 26 ++++++++++++++++++
block/bfq-wf2q.c | 13 ++++++---
3 files changed, 106 insertions(+), 11 deletions(-)
--
2.16.1
Powered by blists - more mailing lists