[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <yq1d1rvvstn.fsf@sermon.lab.mkp.net>
Date: Wed, 17 Feb 2016 18:08:04 -0500
From: "Martin K. Petersen" <martin.petersen@...cle.com>
To: Jan Kara <jack@...e.cz>
Cc: Changho Choi-SSI <changho.c@....samsung.com>,
"lsf-pc\@lists.linux-foundation.org"
<lsf-pc@...ts.linux-foundation.org>, linux-fsdevel@...r.kernel.org,
LKML <linux-kernel@...r.kernel.org>
Subject: Re: [Lsf-pc] [LSF/MM TOPIC] multi-stream IO hint implementation proposal for LSF/MM 2016
>>>>> "Jan" == Jan Kara <jack@...e.cz> writes:
Jan> So a key question for a feature like this is: How many stream IDs
Jan> are devices going to support? Because AFAIR so far the answer was
Jan> "it depends on the device". However the design how stream IDs can
Jan> be used greatly differs between "a couple of stream IDs" and
Jan> e.g. 2^32 stream IDs. Without this information I don't think the
Jan> discussion would be very useful. So can you provide some rough
Jan> numbers?
Both SCSI and NVMe have a 16-bit stream ID field. So you are capped at
64K unique identifiers.
However, the number of streams a device can have open concurrently is
expected to be much smaller than that. This leaves us with the
interesting problem of how to divide up the available streams. Whether
to support application- and kernel allocated streams concurrently. And
how we deal with explicitly closing streams after use.
--
Martin K. Petersen Oracle Linux Engineering
Powered by blists - more mailing lists