BFQ, fairness and low latency in block I/O – The Two Towers – SFO17-302

Session ID: SFO17-302
Session Name: BFQ, fairness and low latency in block I/O – The Two Towers – SFO17-302
Speaker: Paolo Valente
Track: Kernel

★ Session Summary ★
The BFQ I/O scheduler has been finally ported to blk-mq, and merged
into the kernel (from 4.12). We start the first part of this
presentation by recalling motivations and benefits of BFQ. Then we
report the developments occurred since BFQ has been ported to blk-mq
and upstreamed: fixes of important bugs, further throughput
improvements for flash-based storage, public performance reports. In
the second part of the presentation, we deal with the near future of
BFQ, characterized by two great challenges. The first is to keep
throughput high, while, at the same time, providing strong guarantees
on fairness, bandwidth distribution and latency. The second great
challenge is to achieve this combined goal with the low overhead
imposed by modern devices. As a last part, we try to have a look at
the future of BFQ, and of the latency issues that still affect many,
if not most Linux-based systems
★ Resources ★
Event Page:

★ Event Details ★
Linaro Connect San Francisco 2017 (SFO17)
25-29 September 2017
Hyatt Regency San Francisco Airport

Follow us on Social Media

Event Date: 27 Sept 2017

Speakers: Paolo Valente