Greg Ames wrote:
one of the things that inhibits our SMP scalability with
out-of-the-box Linux kernels is contention on the dcache spinlock.
oopss/dcache/dentry_cache/
The LTC guys use a "dcache RCU (read-copy-update)" patch
same here
Greg
> > When our specweb guys were whacking Zeus they would first run a
> > program that would walk the file set to try and fill up the cache.
Zues had some sort of internal cache that needed to be warmed on a
per-process basis (they would run with one process per cpu), as well
as warming the read ca
David Hill wrote:
When our specweb guys were whacking Zeus they would first run a
program that would walk the file set to try and fill up the cache.
hey, I like that idea! I wonder if:
find /spec_docroot/file_set/ -type f | xargs cat > /dev/null
...will do the job? I'll give it a try.
Greg
> > 1. most modern day os'es cache the files, and not do a disk io for
every
> > single file request. (duh !!.)
>
> yep. Yesterday I powered up wimp for the first time in ages and did
a
> mini-SPECweb experimental run in preparation for fiddling with the
stat() in
> mod_specweb99. I got really
MATHIHALLI,MADHUSUDAN (HP-Cupertino,ex1) wrote:
I had a couple of inputs here : I was talking to our specweb person, and he
had the following views :
1. most modern day os'es cache the files, and not do a disk io for every
single file request. (duh !!.)
yep. Yesterday I powered up wimp for the fir
> I had a couple of inputs here : I was talking to our specweb person,
and he
> had the following views :
>
> 1. most modern day os'es cache the files, and not do a disk io for
every
> single file request. (duh !!.)
Part of the design of specweb was to make it difficult (but not
imposible) to cac
e server is
idle.
-Madhu
>-Original Message-
>From: Greg Ames [mailto:[EMAIL PROTECTED]
>Sent: Tuesday, January 28, 2003 11:05 AM
>To: [EMAIL PROTECTED]
>Subject: Re: gettimeofday calls
>
>
>Greg Ames wrote:
>> Bill Stoddard wrote:
>
>>> Why
Greg Ames wrote:
Bill Stoddard wrote:
Why not use mod_file_cache?
On my wimpy 200MHz server, the SPEC file_set contains 5760 files and
uses .8G of disk. On more modern servers, the size of the file_set goes
up in proportion to the number of conforming connections you hope to
push thru it, so
Bill Stoddard wrote:
Infact, I tried this out yesterday (having the one global_time
variable), it
gives me around 3-4% improvement. But, occasionally I do get some
un-conforming results, and I'm trying to figure out if it's because of
the
time stamp.
You probably need to mutex updates to your g
Bill Stoddard wrote:
1. Why we need to do the apr_stat() for static files each time the
request
comes in - can it be done during the module_init() phase, and the
values put
in a array of some sort. ?.
Files change. Why not use mod_file_cache? It will (or should if it does
not have a bug) elimin
MATHIHALLI,MADHUSUDAN (HP-Cupertino,ex1) wrote:
I was referring to the time(), and thinking of
alternative ways of replacing the time() call.
I did that (without using the macros) - but didn't see much difference
though.. I think I was banging my head against the wall yesterday - by tring
to rem
>-Original Message-
>From: Bill Stoddard [mailto:[EMAIL PROTECTED]
[snip]
>You probably need to mutex updates to your global variable, which will
>probably suck out most of your performance gains.
That is correct.. The assumption I had is : timestamp is done once per
request, and since th
Infact, I tried this out yesterday (having the one global_time variable), it
gives me around 3-4% improvement. But, occasionally I do get some
un-conforming results, and I'm trying to figure out if it's because of the
time stamp.
You probably need to mutex updates to your global variable, which wi
> -Original Message-
> From: Greg Ames [mailto:[EMAIL PROTECTED]
>I'm a little confused. The gettimeofday()/apr_time_now() happens in the
httpd
>core AFAIK. Brian Pane said that it's cheaper than time() on some
platforms, I
>believe. If we want to discuss alternatives to that, the di
MATHIHALLI,MADHUSUDAN (HP-Cupertino,ex1) wrote:
I tried that, and I got back error from specweb99 stating that the responses
were not conforming. (or something like that)..
OK, if you used r->request_time to replace the time() calls in mod_specweb99,
you might have convert the units if the result
>> Is this against the spec or something ?.
>
> Which spec? If you are referring to either the SPECWeb99 run rules or to
> RFC2616, neither of them dictate which syscalls you use.
IIRC, the SPECWeb99 run rules just say that you have to treat ad expiration
correctly.
S.
--
Covalent Technologie
MATHIHALLI,MADHUSUDAN (HP-Cupertino,ex1) wrote:
I don't know if this has been discussed already, but I was thinking of the
following alternatives :
1. how about the listener thread in each of the child process keeps updating
a global time variable that each of the threads can refer to ?.
2. set t
() - in the ap_log_error(), I've
put it in "ifdef DEBUG" :-).. Trying this out now..
-Madhu
-Original Message-
From: David Hill [mailto:[EMAIL PROTECTED]
Sent: Thursday, January 23, 2003 1:56 PM
To: [EMAIL PROTECTED]
Subject: Re: gettimeofday calls
I would think that
03 4:05 PM
Subject: RE: gettimeofday calls
> I don't know if this has been discussed already, but I was thinking
of the
> following alternatives :
> 1. how about the listener thread in each of the child process keeps
updating
> a global time variable that each of the threads can refer to
ailto:[EMAIL PROTECTED]
Sent: Thursday, January 23, 2003 11:19 AM
To: [EMAIL PROTECTED]
Subject: Re: gettimeofday calls
MATHIHALLI,MADHUSUDAN (HP-Cupertino,ex1) wrote:
> The following is the tusc output of httpd (2.0.43) +
mod_specweb99.c
> on HP-UX.. Almost every single request has a g
MATHIHALLI,MADHUSUDAN (HP-Cupertino,ex1) wrote:
The following is the tusc output of httpd (2.0.43) + mod_specweb99.c
on HP-UX.. Almost every single request has a gettimeofday system call - is
there any way to avoid it ?.
The GET /file_set/* requests are just plain ol' static files served by the h
Hi,
The following is the tusc output of httpd (2.0.43) + mod_specweb99.c
on HP-UX.. Almost every single request has a gettimeofday system call - is
there any way to avoid it ?. I haven't searched the archives if this
question has already been asked several times - so, please excuse me here.
22 matches
Mail list logo
