From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Subject: Re: [PATCH 4/6] blk-mq: use EWMA to estimate congestion threshold To: Ming Lei , linux-block@vger.kernel.org, Christoph Hellwig Cc: Bart Van Assche , Sagi Grimberg References: <20170711182103.11461-1-ming.lei@redhat.com> <20170711182103.11461-5-ming.lei@redhat.com> From: Jens Axboe Message-ID: Date: Tue, 11 Jul 2017 12:25:16 -0600 MIME-Version: 1.0 In-Reply-To: <20170711182103.11461-5-ming.lei@redhat.com> Content-Type: text/plain; charset=utf-8 List-ID: On 07/11/2017 12:21 PM, Ming Lei wrote: > When .queue_rq() returns BLK_STS_RESOURCE(BUSY), we can > consider that there is congestion in either low level > driver or hardware. > > This patch uses EWMA to estimate this congestion threshold, > then this threshold can be used to detect/avoid congestion. This whole patch set lacks some sort of reasoning why this makes sense. I'm assuming you want to reduce unnecessary restarts of the queue? I would much rather ensure that we only start when we absolutely have to to begin with, I'm pretty sure we have a number of cases where that is not so. What happens with fluid congestion boundaries, with shared tags? -- Jens Axboe