On Sun, 17 Apr 2005, Robert Watson wrote:
I'd like to confirm that for the first two patches, for interesting
workloads, performance generally improves, and that stability doesn't
degrade. For the third partch, I'd like to quantify the cost of the
changes for interesting
es applied.
Because this diff was generated by p4, patch may need some help in
identifying the targets of each part of the diff.
Robert N M Watson
On Sun, 17 Apr 2005, Robert Watson wrote:
Attached please find three patches:
(1) uma.diff, which modifies the UMA slab allocator to use critical
sections
On Mon, 25 Apr 2005, Robert Watson wrote:
I now have updated versions of these patches, which correct some
inconsistencies in approach (universal use of curcpu now, for example),
remove some debugging code, etc. I've received relatively little
performance feedback on them, and would appre
On Mon, 25 Apr 2005, Robert Watson wrote:
On Mon, 25 Apr 2005, Robert Watson wrote:
I now have updated versions of these patches, which correct some
inconsistencies in approach (universal use of curcpu now, for example),
remove some debugging code, etc. I've received relatively l
On Sat, 30 Apr 2005, Arne WXrner wrote:
3. The man page geom(4) of R5.3 says "The GEOM framework
provides an infrastructure in which "classes" can per-
form transformations on disk I/O requests on their path
from the upper kernel to the device drivers and back.
Could it be, that geom slows so
On Sun, 1 May 2005, Mike Tancsa wrote:
A somewhat obvious question to some perhaps, but what server application
mix on FreeBSD today sees an improvement using 64bit CPUs ? In my ISP
centric world, my big apps are BIND, IMAP/POP3, httpd via apache, SMTP,
AV and SPAM scanning, and firewalls/routi
Now that the UMA changes to use critical sections instead of per-cpu mutexes
for per-cpu caches are merged, I'd like to move on to the next piece of the
memory allocation patch previous posted. The attached patch does a couple of
things:
- Introduce per-cpu malloc type statistics, protecting t
On Mon, 2 May 2005, Poul-Henning Kamp wrote:
I'm quite willing to test and optimise things but so far no one has
had any concrete suggestions on that to try.
First thing I heard about this was a few hours ago. (Admittedly, my
email has been in a sucky state last week, so that is probably my own
On Sat, 28 May 2005, Matthew D. Fuller wrote:
On Sat, May 28, 2005 at 01:17:24PM +1000 I heard the voice of
Michael VInce, and lo! it spake thus:
On one of my web servers I have a really high usage of mbuf clusters in
use on a web server that does about 3million hits a day.
4294914731/262144 m
On Fri, 17 Jun 2005, David Sze wrote:
FreeBSD/amd64 5.4-RELEASE (libpthread, system and process scope)
FreeBSD/amd64 6.0-CURRENT (libpthread, libthr, system and process
scope)
CentOS/amd64 4.0 (i.e. RHEL4.0)
I couldn't get libthr to work on FreeBSD/amd64 5.4-RELEASE, mys
On Mon, 20 Jun 2005, David Sze wrote:
I'll be re-running super-smack against an InnoDB table. Any additional
requests for configurations to test, or other tweaking suggestions?
This is what I'll be trying:
FreeBSD/amd64 4.11-RELEASE, linuxthreads
FreeBSD/amd64 5.4-RELEASE, lib
On Thu, 23 Jun 2005, Steve Roome wrote:
The different threading libraries are more for completeness. In my
last test I saw <10% difference between them on amd64.
Well, I finally got some tests out for FreeBSD/i386 with -current, Here
we go with a bunch of results of FreeBSD 6 with mysql and
In 2003, Jonathan Lemon added initial support for direct dispatch of
netisr handlers from the calling thread, as part of his DARPA/NAI Labs
contract in the DARPA CHATS research program. Over the last two years
since then, Sam Leffler and I have worked to refine this implementation,
removing
On Wed, 5 Oct 2005, Robert Watson wrote:
In 2003, Jonathan Lemon added initial support for direct dispatch of
netisr handlers from the calling thread, as part of his DARPA/NAI Labs
contract in the DARPA CHATS research program. Over the last two years
since then, Sam Leffler and I have
On Wed, 12 Oct 2005 [EMAIL PROTECTED] wrote:
At Tue, 11 Oct 2005 15:01:11 +0100 (BST),
rwatson wrote:
If I don't hear anything back in the near future, I will commit a
change to 7.x to make direct dispatch the default, in order to let a
broader community do the testing. :-) If you are setup t
On Fri, 14 Oct 2005, Ben Laurie wrote:
JG wrote:
Now I say, what about
# cpio -i < mysql-m.tgz
(assuming that mysql-m.tgz is in "tar" format)?
Hello,
idea is good, but the result is the same :/
The filesystem is _much_ slower in 5 than it is in 4. You can benchmark
it with postmark (in t
On Tue, 25 Oct 2005, Lukas Razik wrote:
http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/i386/i386/mp_machdep.c?only_with_tag=RELENG_5_4
"Add a knob for disabling/enabling HTT,
"machdep.hyperthreading_allowed".
Default off due to information disclosure on multi-user systems."
Does turnin
On Wed, 9 Nov 2005, Mike Jakubik wrote:
I have done many tests to try to determine the poor performance on my
systems (FreeBSD-current connected directly to Windows XP via identical
Intel Pro 1000 cards) and my only conclusion is that Samba on FreeBSD
when talking to a Windows box is simply fu
On Wed, 9 Nov 2005, Joao Barros wrote:
I tried using a single drive, an IDE and a SCSI-2 and on 2 machines at
work both with a RAID1. Even better, there is a part in my initial email
where I mention that having a 700MB file cached (iostat reported no
reads) the results were the same. With thi
On Wed, 23 Nov 2005, Michael Conlen wrote:
Sorry if this is the wrong place for this, but I haven't been getting
answers elsewhere.
I'm trying to tune the system to allow very large mmap()'s in a single
process space, something on the order of 1.5 GB so I can pass very large
values for -Xms
On Wed, 25 Jan 2006, Ivan Voras wrote:
Thomas Krause (Webmatic) wrote:
But with static linked libraries, I cannot switch between
libthread and libthr - right? Could somebody give me an inspiration/
recommendation?
Unless you are ready to count individual CPU cycles, you won't find a
notica
Dear all,
Attached, please find a patch implementing more fine-grained locking for the
POSIX local socket subsystem (UNIX domain socket subsystem). In the current
implementation, we use a single global subsystem lock to protect all local IPC
over the PF_LOCAL socket type. This has low overh
On Sat, 6 May 2006, Robert Watson wrote:
Attached, please find a patch implementing more fine-grained locking for the
POSIX local socket subsystem (UNIX domain socket subsystem). In the current
implementation, we use a single global subsystem lock to protect all local
IPC over the PF_LOCAL
On Sat, 6 May 2006, Attila Nagy wrote:
On 2006. 05. 06. 16:16, Robert Watson wrote:
In local measurements, I have observed a 0% change in MySQL performance on
uniprocessor systems, and on a dual-processor system I have observed a
4%-5% performance improvement with two client MySQL threads
On Sun, 7 May 2006, Sven Petai wrote:
I performed tests on a 4 * dualcore 2Ghz opteron system (so 8 cores in
total).
In general with 10 parallel smacker threads the performance seems to go up
with your patch by ~44% and with 100 parallel threads it goes down ~25%
This is an interesting eff
On Sun, 7 May 2006, Mike Jakubik wrote:
The difference in performance is just ridiculous. Is mysql written to be
slow on freebsd or is there a problem with freebsd?
In past discussion, I think a reasonable conclusion has been some amount of
both. We've identified a few particular areas wher
On Sun, May 07, 2006 at 10:00:41PM +0300, Sven Petai wrote:
The results in my mail were mean values over 2 runs,
only once did I see really huge (more than 10%) differences between several
subsequent runs with same settings, this case was clearly mentioned in the
results.
FYI, 2 is not real
On Sun, 7 May 2006, Kris Kennaway wrote:
Typically, I do 12 runs of supersmack in each configuration, and discard
the first 2 runs in which the cache and scheduler (etc) are still settling,
as I'm interested in the steady state.
Yeah, forgot to mention that too. Also keeping in mind that my
On Mon, 8 May 2006, Kris Kennaway wrote:
unp contention has risen a bit. The other big gain is to sleep mtxpool
contention, which roughly doubled:
In the general case, you can increase the size of the mutex pool. However,
since this is per-uid for the socket buffer resource limits, that wo
On Wed, 10 May 2006, David Xu wrote:
Fixing one of big lock contentions is not enough, you have to fix them all,
it is easy to see that a second contention becomes a top one. :-)
So I guess the real question is: do we want to merge the UNIX domain socket
locking work? The MySQL gains sound
On Thu, 11 May 2006, Scott Long wrote:
So I guess the real question is: do we want to merge the UNIX domain socket
locking work? The MySQL gains sound good, the performance drop under very
high load seems problematic, and there are more general questions about
performance with other workload
On Fri, 19 May 2006, Vlad GALU wrote:
On 5/19/06, Chad Leigh -- Shire.Net LLC <[EMAIL PROTECTED]> wrote:
On May 18, 2006, at 3:27 PM, Ivan Voras wrote:
>
> Using lighttpd or ab on other host is not really applicable - I
> wasn't trying to configure the box for production, only ran the
> benc
On Fri, 19 May 2006, Ivan Voras wrote:
If I was working on a project such as Apache I know I wouldn't be
concentrating of performance with prefork in this day when threaded modes
are where Apache has been aiming for modern high performance web serving
for a fair while now.
Did you miss disc
On Fri, 19 May 2006, Ivan Voras wrote:
Michael Vince wrote:
What I am trying to say here is you are expecting good performance out of
things like CGI/PHP and prefork,
Ok, did anybody read my initial post?
I'm NOT setting up a production machine. I'm NOT using PHP - it was
mentioned as a r
On Sun, 11 Jun 2006, Kris Kennaway wrote:
* The postgres processes seem to change their proctitle hundreds or
thousands of times per second. This is currently done via a Giant-locked
sysctl (kern.proc.args) so there is enormous contention for Giant. Even
when this is fixed (thanks to a patc
On Mon, 12 Jun 2006, Danial Thom wrote:
first, why is the default for HZ now 1000? It seems that 900 extra clock
interrupts aren't a performance enhancement.
This is a design change that is in the process of being reconsidered. I
expect that HZ will not be 1000 in 7.x, but can't tell you whe
On Mon, 12 Jun 2006, Danial Thom wrote:
This is a design change that is in the process of being reconsidered. I
expect that HZ will not be 1000 in 7.x, but can't tell you whether it will
go back to 100, or some middle ground. There are a number of benefits to a
higher HZ, not least is more
On Mon, 12 Jun 2006, Scott Long wrote:
I run a number of high-load production systems that do a lot of network and
filesystem activity, all with HZ set to 100. It has also been shown in the
past that certain things in the network area where not fixed to deal with a
high HZ value, so it's poss
On Tue, 13 Jun 2006, David Xu wrote:
On Tuesday 13 June 2006 04:32, Kris Kennaway wrote:
On Mon, Jun 12, 2006 at 09:08:12PM +0100, Robert Watson wrote:
On Mon, 12 Jun 2006, Scott Long wrote:
I run a number of high-load production systems that do a lot of network
and filesystem activity, all
On Tue, 13 Jun 2006, Danial Thom wrote:
I'm sorry if I missed it, but I don't believe anyone answered this question:
Lastly, is there a utility similar to cpustat in
DragonflyBSD which shows the per-cpu usage stats?
I need to gauge the efficiency of SMP for a particular application, and
On Tue, 13 Jun 2006, Danial Thom wrote:
I didn't answer it because I don't know what output cpustat provides. What
output does cpustat provide on DragonflyBSD?
Its a simple output such as:
CPU-0 state: 14.00% user, 0.00% nice, 2.00%
sys, 6.00% intr, 78.00% idle
CPU-1 state: 4.00% us
On Tue, 13 Jun 2006, Danial Thom wrote:
Maybe someone can explain this output. The top line shows 99.6%idle. Is it
just showing CPU 0s stats on the top line?
Two types of measurements are taken: sampled ticks regarding whether the
system as a while is in {user, nice, system, intr, idle}, and
On Tue, 13 Jun 2006, Danial Thom wrote:
Two types of measurements are taken: sampled ticks regarding whether the
system as a while is in {user, nice, system, intr, idle}, and then sampling
for individual processes. Right now, the system measurements are kept in a
simple array of tick counter
Scott asked me if I could take a look at the impact of changing HZ for some
simple TCP performance tests. I ran the first couple, and got some results
that were surprising, so I thought I'd post about them and ask people who are
interested if they could do some investigation also. The short
On Sat, 17 Jun 2006, Danial Thom wrote:
At some point you're going to have to figure out that there's a reason that
every time anyone other than you tests FreeBSD it completely pigs out.
Sqeezing out some extra bytes in netperf isn't "performance". Performance is
everything that a system can
Attached, and at the below URL, find an updated copy of the UNIX domain socket
fine-grained locking patch. Since the last revision, I've updated the patch
to close several race conditions historically present in UNIX domain sockets
(which should be merged regardless of the rest of the patch),
On Mon, 3 Jul 2006, David Xu wrote:
I found 5% performance decrease on dual P4, maybe P4 is quite bad when doing
atomic operation. ;-) Thanks,
When I've measured, generally, yes, P4 performance has been abysmal for
synchronization operations, both atomic operations and CPU-local interrupt
di
On Mon, 3 Jul 2006, Robert Watson wrote:
On Mon, 3 Jul 2006, David Xu wrote:
I found 5% performance decrease on dual P4, maybe P4 is quite bad when
doing atomic operation. ;-) Thanks,
When I've measured, generally, yes, P4 performance has been abysmal for
synchronization operations,
On Sat, 12 Aug 2006, Nikolas Britton wrote:
After reading man tuning it said to change vfs.hirunningspace to a larger
value. What should I set it to?
My RAID controller, an Areca ARC-1220, has a 256MB cache and each disk has
16MB of cache.
What other vfs.* sysctls can I change?
The defaul
On Wed, 13 Sep 2006, Ingo wrote:
I`ve some problems with the network performance on my Soekris NET 4801.
(Freebsd 6.1 release-p3)
When I start "netio" on the soekris and do a "netio localhost", I get about
8.4 MB/sec, and when I start with "netio 192.168.0.11"(it´s localhost
address) I get
use on its loopback interface as compared to the
network interface?
Robert N M Watson
Computer Laboratory
University of Cambridge
Greetings
Am 18.09.2006, 15:52 Uhr, schrieb Robert Watson <[EMAIL PROTECTED]>:
On Wed, 13 Sep 2006, Ingo wrote:
I`ve some problems with the n
On Mon, 30 Oct 2006, Marcelo Gardini do Amaral wrote:
Dave, could you please describe you test set?
I've posted some results months ago and they were kind different.
I have done some tests [1] [2] with bind and queryperf and my result on
FreeBSD 6.1 was very poor if compared with 4.11.
Besi
On Wed, 29 Nov 2006, Mike Tancsa wrote:
Did some more tests, this time using a single NIC interface in trunking
mode. Strangely enough, the speed is a little faster on HEAD. Perhaps less
interrupt processing ? Results in the usual place
http://www.tancsa.com//blast.html
You may want to dat
As part of Kris and Jeff's recent work on improving MySQL scalability on
FreeBSD, I've updated my fine-grained locking patch for UNIX domain sockets to
a more recent 7-CURRENT:
http://www.watson.org/~robert/freebsd/netperf/20070224-uds-fine-grain.diff
This patch replaces the global mutex w
On Sat, 24 Feb 2007, Steven Hartland wrote:
- Original Message - From: "Robert Watson" <[EMAIL PROTECTED]>
As part of Kris and Jeff's recent work on improving MySQL scalability on
FreeBSD
Are there any results / info on what's been done that we can look at
puter Laboratory
University of Cambridge
-- Forwarded message --
Date: Mon, 26 Feb 2007 20:47:52 + (UTC)
From: Robert Watson <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED], [EMAIL PROTECTED], cvs-all@FreeBSD.org
Subject: cvs commit: src/sys/sys unpcb.h src/sys/kern uipc_usrreq
On Wed, 28 Feb 2007, Peter Losher wrote:
Ivan Voras wrote:
I agree in general, but MySQL performance is very exposed as an advocacy
issue - it has traditionally been the source of statements like "FreeBSD's
threading implementation is weak/bad/broken".
And these days ISC can't consciously
On Thu, 1 Mar 2007, Kip Macy wrote:
We recently put a stock Fedora Core 6 and a stock FreeBSD 6.2 on the same
HW (HP ProLiant DL320 G5 Dual Core Xeons w/ 16GB RAM) and running BIND
9.4.0 and a well known ccTLD zone that we slammed a query stream to. On a
single threaded BIND, there was a 20%
On Sun, 18 Feb 2007, Lars Erik Gullerud wrote:
On Fri, 16 Feb 2007, Robert Watson wrote:
I can certainly investigate doing this -- since 6.2 is safely out the door
it's a good time to do so. I'll follow up by e-mail in a few days -- would
it be possible for you to help with tes
On Fri, 2 Mar 2007, O. Hartmann wrote:
On my lab's FreeBSD 6.2/i386 box (ASUS P4P800, ICH5 with two SATA 150 ports,
two SATA 300 drives attached) I copied big files (~ 5GB) from one drive to
another while the box didn't do anything else than copying. I watched the
copy process via 'systat -vms
On Mon, 5 Mar 2007, David Gilbert wrote:
"Dinesh" == Dinesh Nair <[EMAIL PROTECTED]> writes:
Dinesh> On Wed, 28 Feb 2007 01:06:33 -0800, Peter Losher
Dinesh> <[EMAIL PROTECTED]> wrote:
Ivan Voras wrote:
I agree in general, but MySQL performance is very exposed as an advocacy
issue - it has
On Wed, 7 Mar 2007, [EMAIL PROTECTED] wrote:
For reference, the sosend_copyin.diff applies these changes:
src/sys/kern/uipc_socket.c:1.253, 1.254, 1.255
Sorry, It's seem doesn't work ... . The function sosend_copyin() where Have
I to declare it?
socketvar.h:1.144 is missing from this pa
On Thu, 15 Mar 2007, Marcelo Gardini do Amaral wrote:
I repeated that performance test done with bind [1] using now Robert's MFC
[2]. Another tweak that I was supposed to do was to use libthr instead of
libpthread (via libmap.conf) and build bind with threads option. In the new
test I did thi
Dear all,
The attached patch moves file descriptor locks from being a custom mutex/sleep
lock implemented using msleep() to an sx lock. With the new sx lock
optimizations in place, this is now sensible, avoiding both a custom lock type
and significantly improving performance. Kris has repor
On Sun, 1 Apr 2007, Robert Watson wrote:
The attached patch moves file descriptor locks from being a custom
mutex/sleep lock implemented using msleep() to an sx lock. With the new sx
lock optimizations in place, this is now sensible, avoiding both a custom
lock type and significantly
On Wed, 4 Apr 2007, Hans Petter Selasky wrote:
Just a small comment:
@@ -60,10 +60,7 @@
u_short fd_cmask; /* mask for file creation */
u_short fd_refcnt; /* thread reference count */
u_short fd_holdcnt; /* hold count on structure +
ed message --
Date: Wed, 4 Apr 2007 09:11:34 + (UTC)
From: Robert Watson <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED], [EMAIL PROTECTED], cvs-all@FreeBSD.org
Subject: cvs commit: src/sys/compat/linux linux_file.c src/sys/compat/svr4
svr4_filio.c src/sys/dev/streams streams.c
On Tue, 10 Apr 2007, Kevin Way wrote:
Kris Kennaway wrote:
If so, then your task is the following:
Make SYSV semaphores less dumb about process wakeups. Currently whenever
the semaphore state changes, all processes sleeping on the semaphore are
woken, even if we only have released enough r
On Sun, 22 Apr 2007, Stefan Lambrev wrote:
I'm using my laptop to run ab in both test, and 2 totally different servers
:) That's why I do not pretend that the benchmark is done the right way.
The linux host is with pentium 4 single core processor, The freebsd host is
amd64 athlon 3200+ (2GHz
On Wed, 5 Dec 2007, Philipp Wuensche wrote:
we are running a FreeBSD 7-BETA4 with SCHED_4BSD on a Intel Core2Dual E6600
2.4GHz system for our bittorrent Opentracker.
The system handles about 20Kpps (18Mbit/s) incoming and 15kpps (22 Mbit/s)
outgoing traffic serving 4000 connections/sec using
On Wed, 23 Jan 2008, Erik Cederstrand wrote:
I'd like to send a small update on my progress on the Performance Tracker
project.
I now have a small setup of a server and a slave chugging along, currently
collecting data. I'm following CURRENT and collecting results from
super-smack and unixb
On Wed, 23 Jan 2008, Erik Cederstrand wrote:
One way to do this would be a matrix of each metric with its change
compared to recent samples. e.g. you could do a student's T comparison of
today's numbers with those from yesterday, or from a week ago, and
colour-code those that show a significa
On Wed, 23 Jan 2008, Erik Cederstrand wrote:
Robert Watson wrote:
This looks really exciting!
Do you plan to add a way so that people can submit performance data?
I.e., if I set up my own test box and want to submit a result once a week
for that, will there be a way for me to get set up
On Fri, 1 Feb 2008, Alexander Motin wrote:
That was actually my second question. As there is only 512 items by default
and they are small in size I can easily preallocate them all on boot. But is
it a good way? Why UMA can't do just the same when I have created zone with
specified element siz
On Sat, 2 Feb 2008, Alexander Motin wrote:
Robert Watson wrote:
I guess the question is: where are the cycles going? Are we suffering
excessive cache misses in managing the slabs? Are you effectively "cycling
through" objects rather than using a smaller set that fits better in
On Sat, 2 Feb 2008, Kris Kennaway wrote:
Alexander Motin wrote:
Robert Watson wrote:
Hence my request for drilling down a bit on profiling -- the question I'm
asking is whether profiling shows things running or taking time that
shouldn't be.
I have not yet understood why does
On Sun, 3 Feb 2008, Alexander Motin wrote:
Robert Watson wrote:
Basically, the goal would be to make the pcpu cache FIFO as much as
possible as that maximizes the chances that the newly allocated object
already has lines in the cache.
Why FIFO? I think LIFO (stack) should be better for this
On Sat, 1 Mar 2008, Chris wrote:
Ironically the latest server I got last night has a intel pro 1000 a rarity
:)
I am just giving feedback as when I speak to people in the datacentre and
hosting business the biggest gripe with freebsd is hardware compatability,
as I adore freebsd I ignore th
On Sat, 8 Mar 2008, Steven Hartland wrote:
We've been suffering on our stats box for some time now where by the machine
will just stall for several seconds preventing everything from tab
completion to vi newfile.txt.
I was hoping an upgrade to 7.0 and ULE may help the situation but
unfortuna
On Mon, 17 Mar 2008, Aminuddin Abdullah wrote:
I have just upgraded 5 of my machines to V7 from 6.3 and then realized that
all the machines has a high CPU usage. Almost all of them using 80%-90% CPU
with more than 8000 connections. Using previous 6.3, it only uses 40-50% CPU
with the same kin
On Tue, 18 Mar 2008, John Baldwin wrote:
'+' is used in a swi name to indicate that the names of the interrupts to
put in the thread name are too long, and the code looks like it was written
under the assumption that at least one name would fit. It sounds like in
this case, none fit. We shou
On Wed, 30 Jul 2008, Alexander Strange wrote:
On Jul 21, 2008, at 3:53 PM, Ivan Voras wrote:
Alexander Strange wrote:
And there's no firewalls or packet shapers in front of it.
How about on it? Do you run ipfw?
No, I wouldn't answer a question so specifically like that.
We didn't see th
On Tue, 12 Aug 2008, Tim Traver wrote:
I have recently had the opportunity to upgrade a few servers from old
versions of 5.4 to 7.0, and have seen some interesting data. Before doing
this, I wanted to take some benchmarks to see how the scripts that I would
run would fare between the two vers
On Sun, 25 Jul 2010, Alexander Motin wrote:
The numbers that you are showing doesn't show much difference. Have you
tried buildworld?
If you mean relative difference -- as I have told, it's mostly because of my
CPU. It's maximal boost is 266MHz (8.3%), but 133MHz of them is enabled most
of t
84 matches
Mail list logo
