From: Werner Almesberger <werner@almesberger.net>
To: abiss-general@lists.sourceforge.net
Cc: linux-fsdevel@vger.kernel.org, celinux-dev@tree.celinuxforum.org
Subject: Active Block I/O Scheduling System (ABISS), version 1
Date: Thu, 21 Oct 2004 11:02:33 -0300 [thread overview]
Message-ID: <20041021110233.A31465@almesberger.net> (raw)
The Active Block I/O Scheduling System (ABISS) is an extension of the
hard-disk storage subsystem of Linux, whose main purpose is to provide
a guaranteed reading and writing bit rate to applications.
http://abiss.sourceforge.net/abiss-1.tar.gz
md5sum ae832bcb94d5f8f29bfe4f4a737b2831
sha1sum b57722be9f35b54806b17b896e0cdf48d1e0f9d1
The kernel parts are still for 2.6.7. (Upgrade to 2.6.9 coming soon.)
The highlights since the last release:
- added libabiss, a wrapper library that provides unistd/stdio-like
functions, prefixed with "abiss_", to open and access files with
ABISS.
- priorities are now per task, not per file, which greatly simplifies
things, and better aligns them with Jens' priority design.
- when refilling a buffer, the scheduler also tries to refill the
buffers of other files which are close to needing a refill. This
reduces the number of seeks if there are a few relatively slow
real-time readers competing with greedy best-effort readers.
- the whole "access prediction" mechanism for writes is gone. All
it did was to provide an estimate even Murphy would find too
pessimistic, and it needed a lot of file system specific code to do
this.
- switched from the "unofficial" UML patch 2.6.7-01 to the "official"
2.6.7-1
The detailed list of changes is below. For additional information,
please have a look at http://abiss.sourceforge.net/
- Werner
----------------------------------- CHANGES -----------------------------------
Version 1 (21-OCT-2004)
-----------------------
General:
- README: corrected Nick Piggin's name
- Makefile: added SHA1 sum
Tools:
- rdrt now measures the time to move the playout point separately, and sports
two new options, -p and -a, to print this time over run time, or in a
histogram
- rdrt now limits memory allocations (can be turned off with new option -x),
and its CPU priority can be lowered to normal with new option -o
abissd:
- abissd/var.c:var_assignment: forgot that strncpy does not always
NUL-terminate its strings
Kernel:
- switched from "unofficial" uml-patch-2.6.7-01 to "official" uml-patch-2.6.7-1
- removed per-file (really per-inode) priorities, because we already have
per-task priorities
- fs/abiss/sched_lib.c:abiss_hp_raise: set "sched_priority" to 1 instead of
(invalid) 0
- removed /sys/block/.../default_priority
- fs/abiss/sched_test.c: in an abiss_st_query message, msg.u.query.page_state
can now be NULL to omit retrieving the potentially large page state array
- removed access prediction for allocations
- added reserved space to "struct abiss_sched_test_prm", to allow for future
extensions without breaking binary compatibility
- sched_test.c now tries to refill buffers of different files at the same time,
to avoid excessive seeking
- fs/abiss/elevator.c: added some tracing to hunt down a heisenbug
--
_________________________________________________________________________
/ Werner Almesberger, Buenos Aires, Argentina werner@almesberger.net /
/_http://www.almesberger.net/____________________________________________/
reply other threads:[~2004-10-21 14:02 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20041021110233.A31465@almesberger.net \
--to=werner@almesberger.net \
--cc=abiss-general@lists.sourceforge.net \
--cc=celinux-dev@tree.celinuxforum.org \
--cc=linux-fsdevel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).