Messages in this thread | | | Date | Tue, 23 Feb 1999 03:30:08 -0800 | From | Hans Reiser <> | Subject | Re: (reiserfs) Re: Reiserfs is order of magnitude faster for fsyncof large files than ext2 |
| |
<!doctype html public "-//w3c//dtd html 4.0 transitional//en"> <html> David, I agree that you ask the right question, and I asked <br>Yura to make the almost the same measurement you asked for <br>5 minutes after I posted to linux-kernel about it. <br>It shouldn't take Yura more than 30 minutes to do. <p>Best, <p>Hans <p>David C Niemi wrote: <blockquote TYPE=CITE>On Tue, 23 Feb 1999, Vladimir V. Saveliev wrote: <br>> David C Niemi wrote: <br>> > A question: how about fsync() when there is actually a modest amount of <br>> > real change between in-core and on-disk? The benchmark you describe is <br>> > really a "null fsync()", which while promising might not be completely <br>> > representative of real-world usage. <br>> <br>> Hi <br>> I would say, that the most significant thing in such test is not amount <br>> of changes (if you just have read whole file, fsync costs nothing, as <br>> all indirect blocks are in cache). The amount of un-cached indirect <br>> blocks is the bottleneck for file system like ext2 in such test. <p>I would expect so, but given wildly different implementations I would be <br>more convinced with tests showing similar results when there are a few <br>dirty pages. <p>> What is typical usage pattern for fsyncing of big files of logs or <br>> databases? 1% or 10%? We really would love to measure it. <p>Both would be good data points, but 1% is probably much more typical. If <br>you think about it, fsync is just writing the dirty blocks that are within <br>the last two automatic sync() intervals worst case (sync traditionally <br>writes all currently scheduled blocks to disk, and schedules all currently <br>schedulable blocks, hence the merits of doing 2-3 sync commands). With a <br>sync interval of 20-30 seconds, I'd expect you would seldom dirty more than <br>1% of a large file within a minute unless you were in the midst of truly <br>torrid activity (which would be a totally separate test case). <p>--- David C Niemi ---niemi at tux.org--- Reston, Virginia, USA --- <br> But only the man who cares about something in itself, who loves <br> it and does it *con amore*, will do it in all seriousness. The <br> highest achievement has always been that of such men, and not of <br> the hacks who serve for pay. -- Arthur Schopenhauer</blockquote>
<pre>-- Don't be locked out of the source, and doomed to life in the slow lane. Dump NT! Get Linux (<A HREF="http://www.kernel.org">http://www.kernel.org</A>) plus ReiserFS (<A HREF="http://devlinux.org/namesys">http://devlinux.org/namesys</A>). If you sell an OS or internet appliance, buy a port of ReiserFS! Speed matters. Trees are fast. Go faster!</pre> </html>
| |