Performance of 1.3 series

Performance of 1.3 series

Post by Joerg Pommni » Fri, 17 Nov 1995 04:00:00



This is a small table I compiled to compare the performance development
of the Linux kernel over the 1.3 series. Especially the xbench tests have
a few gaps, but I think the general trend towards an improved performance
is quite obvious.
My comments are mostly guesses to explain the changes in the numbers. If
somebody has better explanations, feel free to correct me.

Hardware:
 CPU   : i486/66
 RAM   : 16M
 SCSI  : NCR 53c810
 HD    : QUANTUM MAVERICK 540S
 video : ATI Mach64

Software:
 bm     : BYTE Unix Benchmarks Version 3.6 -- 5/15/91 (as found on tsx-11)
          statically linked ELF binaries compiled with gcc 2.6.3
 xbench : patched version
          dynamically linked ELF binary compiled with gcc 2.7.0 and
          gcc 2.7.1 snapshots
 X      : XFree86 3.1.2 self compiled with gcc 2.7.0

+---------------+--------+-----+----------------------------------------+
| OS  Version   | xbench | bm  | Comment                                |
+---------------+--------+-----+----------------------------------------+
| Linux-1.2.09  | 149882 | 4.1 | stable version, old UNIX domain sockets|
| Linux-1.3.08  |        | 3.9 |                                        |
| Linux-1.3.11+ |        | 4.3 | scheduler improved                     |
| Linux-1.3.14  |        | 4.7 | fork(2)/clone(2) improved ???          |
| Linux-1.3.17  |        | 5.0 |                                        |
| Linux-1.3.18  |        | 5.0 |                                        |
| Linux-1.3.29  | 121501 | 5.0 | new UNIX domain sockets                |
| Linux-1.3.30  |        | 4.9 |                                        |
| Linux-1.3.34  |        | 4.9 |                                        |
| Linux-1.3.35  | 145499 |     |                                        |
| Linux-1.3.36  |        | 4.9 |                                        |
| Linux-1.3.38  |        | 5.7 | kswap patches in ???                   |
| Linux-1.3.39  | 151284 | 5.5 |                                        |
| Linux-1.3.41  | 154599 | 5.6 | libc-5.2.13 with native readv/writev   |
+---------------+--------+-----+----------------------------------------+

Best Regards
        Joerg
--------------------------------------------------------------------------------
Joerg Pommnitz, IBM Scientific Center Heidelberg


 
 
 

Performance of 1.3 series

Post by Matthias Bruest » Sun, 19 Nov 1995 04:00:00


Mahlzeit


> | Linux-1.3.11+ |        | 4.3 | scheduler improved                     |
> | Linux-1.3.14  |        | 4.7 | fork(2)/clone(2) improved ???          |

I have noticed that the BSD-fork/exec-bench with 1.3.41 nearly twice
as fast as with 1.2.8 is.

Quote:> | Linux-1.3.38  |        | 5.7 | kswap patches in ???                   |

I have 8MB and I think the swaping is slower. The mp-byte is up to
16 users a little bit faster (again 1.3.41 vs. 1.2.8) but with 32 users
both begin to swap and now 1.2.8 is about twice as fast.

Mahlzeit

endergone Zwiebeltuete

i486DX33, 8MB RAM, 64kB cache, ISA

--
PGP:1024/0xDCB8D00F              I LOVE MY PDP-11/34A!
--
Remember, even if you win the rat race -- you're still a rat.

 
 
 

Performance of 1.3 series

Post by liuy » Tue, 21 Nov 1995 04:00:00



>|> | Linux-1.3.38  |        | 5.7 | kswap patches in ???                   |
>|I have 8MB and I think the swaping is slower. The mp-byte is up to
>|16 users a little bit faster (again 1.3.41 vs. 1.2.8) but with 32 users
>|both begin to swap and now 1.2.8 is about twice as fast.
>|
>strangely enough I applied the kswap patches to 1.2.9 and I saw a little
>slow down too??? Granted this benchmarking was not done in the most
>scientific way - I only use my machine as a personal host but I do
>stress the memory resources occasionally - X, emacs, compiling,
>netscape, etc. - I timed a few application load/startup's and saw 10-15%
>slowdown after the patch.

Hi,
  Same here.  After kswap patches to 1.2.13, our 386 w/4 meg (swapping
all the time :-)) is slower.  nn takes longer to load.  Compiling seems also
slower...
Any ideas why?...

Thanks.

 
 
 

Performance of 1.3 series

Post by Andrew Miles » Wed, 22 Nov 1995 04:00:00



> I have notice that the buffers are most time over 2MB. This means
> probably that it must swap more (7MB RAM -> 5MB RAM), but the
> file accesses are faster (0.xMB RAM -> 2MB RAM).

Buffers are reduced (to about 256k) before the kernel starts swapping.

Or so it has been in my experience.

-- Andrew E. Mileski --

-------------------------------------------------------
Dark Matter Technologies Inc. - Ottawa Ontario, Canada!

 
 
 

1. 1.3.60 / General state of 1.3 series ?


...

That raises a key question. With every new version in the 1.3 series we=20
soon hear about individual problems. Since the linux-kernel-digest list=20
at vger.rutgers.edu doesn't appear to work or noone seems to=20
bother sending reports to it, we, the users of linux, have no idea what=20
the current state is. I joined linux-kernel for a week and ended up=20
deleting up to 200 emails on it every day, most of them with a 24-48 hour=
=20
delay. Ok, this list is not really for users, I understand.

But what is there for users ?

Maybe one of you linux developers could give us a brief note on the state=
=20
of v1.3 ? Please ?
We do appreciate your efforts a lot as you well know !! We don't expect=20
you to waste a lot of your time on reports. Maybe just a threeliner every=
=20
ten issues ?

What is the current plan ?
More and more new things seem to be taken into the kernel.
When will there be the next stable one ? I realise that expecting an=20
accurate answer to this question would be foolish. However, maybe a vague=
=20
estimate might be possible ?

Please excuse my ignorance if I happen to have missed the one obvious=20
location to look for information on the current state of linux 1.3.x.
Just point me in that direction, please ?


=09Peter J=E4ckel

===============================================================================
    The above article is the personal view of the poster and should not be
       considered as an official comment from the JET Joint Undertaking
===============================================================================

2. fifo-buffer for ttyS01

3. XFree86 1.3->2.0 ... performance SUX

4. Third IDE disk on 2-HD motherboard ?

5. poor performance after SunOS4.1.3 to Solaris7 upgrade

6. First login of new user : passwd/nispasswd mismatch

7. sysstat-4.1.3 - System performance tools for Linux

8. AIX 3.2.5 and TCP/IP - anyone notice a slowdown?

9. rperf 1.3 -- performance monitoring of network hosts

10. 3c509 and 1.3.6x/1.3.7x

11. Diff. between SunOS4.1.3_U1 and SunOS4.1.3_U1 VB

12. NumberNine Imagine 128 (Series I) scrolling performance

13. Driver crashes on Alpha GS series, OK on DS series