linux-kernel - [RFC 0/2] Increase accuracy and precision of sampled io

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <20200609040724.448519-1-joshs@netflix.com>
Date:   Mon,  8 Jun 2020 21:07:22 -0700
From:   Josh Snyder <joshs@...flix.com>
To:     Jens Axboe <axboe@...nel.dk>,
        Mikulas Patocka <mpatocka@...hat.com>,
        Mike Snitzer <snitzer@...hat.com>
Cc:     linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
        Josh Snyder <josh@...e406.com>
Subject: [RFC 0/2] Increase accuracy and precision of sampled io_ticks

Commit 5b18b5a73760 ("block: delete part_round_stats and switch to less
precise counting") introduces a sampling technique for calculating
io_ticks. The sampling algorithm introduces bias in the calculation of
I/O utilization. In my production system, this bias means that a
workload which previously reported 10% I/O utilization now reports 80%.
Patch 1 of this series eliminates the bias.

The sampling technique is also subject to statistical noise. Because it
infers io_ticks based on only 100 samples per second, io_ticks becomes
imprecise, and subject to swings when measuring both random and
deterministic workloads. Patch 2 of this series provides increased
precision by raising the sampling rate.