Very Computer

by **Chin Fa** » Wed, 25 Aug 1999 04:00:00

A developer here complained to me this afternoon that his compilation
jobs were too slow. Since it's impossible for me to find a faster
machine for him right now, so this evening, I hooked up a spare
7200 RPM external HD to his SS5. The machine has 256MB RAM, two internal
1G 5400 HDs, on a switched local network, and runs Solaris 7 5/99.
The compiler is EGCS 1.1.2.

I attempted to improve the IO with Casper Dik's fastfs. However,
when I attempt to do the following

fastfs /work fast

I kept getting:

/src/system/fastfs /work fast
FAILED: /work is now slow

Feeling curious, I compiled fastfs.c using -g version, and traced the
entire execution in gdb. As it turns out, it failed in the following
section (line numbers included):

140 if (ioctl(fd, FIODIOS, &newflag) == -1) {
141 perror("status ioctl");
142 errors ++;
143 } else {
144 if (cmd != CMD_STATUS && flag != newflag)
145 printf("FAILED: ");
146 if (cmd != CMD_STATUS)
147 how = nochange ? "already " : "now ";
148 printf("%s\tis %s%s\n", path, how, cmds[newflag]);
149 }

I looked at this section of code for a while, but so far I haven't
figured out what's wrong. Perhaps it's late for me. In anycase,
if anyone has any ideas and is willing to share, I would be quite
appreciative.

Regards,

Chin Fang

by **Reardon, Andrew** » Thu, 26 Aug 1999 04:00:00

Hi

For speeding compilations, point the users TMPDIR environment variable to a
tmpfs filesystem (sometimes /tmp, depending on /etc/vfstab).
The compiler will then be using a memory based filesystem for all it's
temporary files, which should speed it up significantly.

fastfs would speed it up too, but not as much as using a tmpfs.

This is explained in the "notes" section in man tmpfs(7FS).

I myself haven't tested fastfs on Sol7. I hope it does work on, it's a great
utility.Can you let the ng know if you do get it working, TIA.

Cheers

Andrew Reardon
UNIX/Informix Administrator
Aircraft Systems, Boeing Aust. Ltd.
Ph: +61 7 3306 3346 Mob: +61 0419 745 831

> ----------

> Sent: Wednesday, August 25, 1999 2:03 PM

> Subject: fastfs doesn't work for Solaris 7?

> Message from the Deja.com forum:
> comp.unix.solaris
> Your subscription is set to individual email delivery

> A developer here complained to me this afternoon that his compilation
> jobs were too slow. Since it's impossible for me to find a faster
> machine for him right now, so this evening, I hooked up a spare
> 7200 RPM external HD to his SS5. The machine has 256MB RAM, two internal
> 1G 5400 HDs, on a switched local network, and runs Solaris 7 5/99.
> The compiler is EGCS 1.1.2.

> I attempted to improve the IO with Casper Dik's fastfs. However,
> when I attempt to do the following

> fastfs /work fast

> I kept getting:

> /src/system/fastfs /work fast
> FAILED: /work is now slow

> Feeling curious, I compiled fastfs.c using -g version, and traced the
> entire execution in gdb. As it turns out, it failed in the following
> section (line numbers included):

> 140 if (ioctl(fd, FIODIOS, &newflag) == -1) {
> 141 perror("status ioctl");
> 142 errors ++;
> 143 } else {
> 144 if (cmd != CMD_STATUS && flag != newflag)
> 145 printf("FAILED: ");
> 146 if (cmd != CMD_STATUS)
> 147 how = nochange ? "already " : "now ";
> 148 printf("%s\tis %s%s\n", path, how, cmds[newflag]);
> 149 }

> I looked at this section of code for a while, but so far I haven't
> figured out what's wrong. Perhaps it's late for me. In anycase,
> if anyone has any ideas and is willing to share, I would be quite
> appreciative.

> Regards,

> Chin Fang

> _____________________________________________________________
> Deja.com: Share what you know. Learn what you don't.
> http://www.deja.com/
> * To modify or remove your subscription, go to
> http://www.deja.com/edit_sub.xp?group=comp.unix.solaris
> * Read this thread at
> http://www.deja.com/thread/%3C7pvpuf%242gp%40saga20.Stanford.EDU%3E

Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.

by **Holger Burd** » Thu, 26 Aug 1999 04:00:00

> A developer here complained to me this afternoon that his compilation
> jobs were too slow. Since it's impossible for me to find a faster
> machine for him right now, so this evening, I hooked up a spare
> 7200 RPM external HD to his SS5. The machine has 256MB RAM, two internal
> 1G 5400 HDs, on a switched local network, and runs Solaris 7 5/99.
> The compiler is EGCS 1.1.2.

> ....

> Regards,

> Chin Fang

hi;

Turning on logging on the Filesystem seems to have a posive effect on the FS
performance. If you have enough RAM
you can also try tmpfs which is a memory based FS. Since the SS5 is not the
fastest machine around and egcs is quite a resource (CPU) hog i think your
problem is more CPU than an IO bound.

hb
---------------------------------------------------------------------
UNIX SysAdmin - BCI GmbH

by **Martin Recktenwal** » Thu, 26 Aug 1999 04:00:00

> I kept getting:

> /src/system/fastfs /work fast
> FAILED: /work is now slow

Well, I have one filesystem with fastfs running:

/squid2/cache is fast

uname -a:

SunOS wjpserver 5.7 Generic_106542-05 i86pc i386 i86pc

At least with Solaris/x86 it does work.

Martin.
--
"They are marked 'dangerous' because they eat filesystems for breakfast."
Linus Torvalds about 2.3.7 pre-patches, LKM

by **Walter Stein** » Thu, 26 Aug 1999 04:00:00

I assume that fastfs does not work on filesystems mounted ufs logging.

-ws

by **Philip Bro** » Thu, 26 Aug 1999 04:00:00

Quote:>I assume that fastfs does not work on filesystems mounted ufs logging.

actually, it half works, and corrupts the filesystem.

--
[Trim the no-bots from my address to reply to me by email!]
[ Do NOT email-CC me on posts. Pick one or the other.]
--------------------------------------------------
The word of the day is mispergitude

by **Casper H.S. Dik - Network Security Engine** » Fri, 03 Sep 1999 04:00:00

[[ PLEASE DON'T SEND ME EMAIL COPIES OF POSTINGS ]]

>fastfs /work fast
>I kept getting:
>/src/system/fastfs /work fast
>FAILED: /work is now slow

fastfs does not work on file systems mounted 'logging'

Casper
--
Expressed in this posting are my opinions. They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.