xfs
[Top] [All Lists]

Re: XFS w/RT

To: tubaranja@xxxxxxxxx
Subject: Re: XFS w/RT
From: Eric Sandeen <sandeen@xxxxxxx>
Date: Fri, 03 Jun 2005 07:55:00 -0500
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <42A014E8.000008.28676@pantene.yandex.ru>
References: <42A014E8.000008.28676@pantene.yandex.ru>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mozilla Thunderbird 1.0.2 (Macintosh/20050317)
Arthur wrote:
Hello!
This message about the realtime feature. Now there is a question about to include it in the kernel of one distribution (CRUX). Me interests opinion of developers. How you think, how much it's stable at present? How often there are problems with it and what character of this problems? Somebody spent tests on latency with inclusion of this opportunity and without it? If yes, can will show me links to results of tests? In advance thanks.

Realtime should be in decent shape but it's not heavily tested. That's why it's still under CONFIG_EXPERIMENTAL


And FWIW, it's not 'realtime' in a strict sense, but the allocator is more deterministic than standard. It also allocates larger chunks at a time.

I'm not aware of any published analysis of latency.

Very few users/applications have a need for this option, to tell you the truth. But in some cases it can be good; direct IO media streaming might be one case.

You might ask James Chapman why they chose to use it in their embedded system, for starters. :)

-Eric


<Prev in Thread] Current Thread [Next in Thread>
  • XFS w/RT, Arthur
    • Re: XFS w/RT, Eric Sandeen <=