5.0.5 MP system running slowly with 1Gb RAM

0 views
Skip to first unread message

Graham Nicholls

unread,
Dec 8, 1999, 3:00:00 AM12/8/99
to
I have a client who has a system which had 512M RAM, 2 PII/333
processors, SCSI, etc. It ran out of memory, so they upgraded. It took
a while to get the memory working in the ASUS motherboard, but now it
is, but the system runs very slowly.

My question is: Is this a SCO or a cache issue? (or something else)


I am coming to the conclusion that this is a cache issue. PII chips
have 512k of L2 cache. Is this enough for 1Gb RAM, or am I confusing
issues?

Any ideas gratefully received.

Marco Ce.

unread,
Dec 9, 1999, 3:00:00 AM12/9/99
to
I was know that OSR 5.05 does not support more than 512 MB Ram.
Try to verify this.


--

Graham Nicholls <gra...@rockcons.co.uk> wrote in message
news:384EACCE...@rockcons.co.uk...

Ken Wolff

unread,
Dec 9, 1999, 3:00:00 AM12/9/99
to

Not true. I'm not sure of the limit off the top of my head, but we have
1GB on our 5.0.5 machine with 4 Pentium Pro 200's (256k cache on each) and
it runs like a champ. We originally had 512MB and 2 processors and added
another 512MB along with the 3rd and 4th processor about 4-5 months ago and
saw a dramatic increase in performance.


--------------------------------------------------------------
Ken Wolff
Phone: 616-957-4949 Ext: 111
FAX: 616-957-1614
--------------------------------------------------------------

L Janda

unread,
Dec 9, 1999, 3:00:00 AM12/9/99
to
Did you do anything about your swap file?

Marco Ce.

unread,
Dec 9, 1999, 3:00:00 AM12/9/99
to
Had you searched in kernel parameter? scoadmin, harware, kernel manager,
tune parameters?
I was wrong about 512 MB, was true for 5.00 5.02.


--
Marco

Ken Wolff <ke...@cscc.maximus.com> wrote in message
news:4.2.0.58.1999120...@scogr1.cscc.maximus.com...


> At 11:54 AM 12/9/99 +0100, Marco Ce. wrote:
> >I was know that OSR 5.05 does not support more than 512 MB Ram.
> >Try to verify this.
> >
> >
> >--
> >

Tony Lawrence

unread,
Dec 9, 1999, 3:00:00 AM12/9/99
to
L Janda wrote:
>
> Did you do anything about your swap file?

What is it you'd want him to do about his swap file?

If he needed more swap, he'd have more problems thaan just
running slowly- with 1 GB of memory he might not even need
any swap: see http://www.aplawrence.com/Boot/swap.html

Something else is afoot here.

>
> Graham Nicholls <gra...@rockcons.co.uk> wrote in message
> news:384EACCE...@rockcons.co.uk...
> > I have a client who has a system which had 512M RAM, 2 PII/333
> > processors, SCSI, etc. It ran out of memory, so they upgraded. It took
> > a while to get the memory working in the ASUS motherboard, but now it
> > is, but the system runs very slowly.
> >
> > My question is: Is this a SCO or a cache issue? (or something else)
> >
> >
> > I am coming to the conclusion that this is a cache issue. PII chips
> > have 512k of L2 cache. Is this enough for 1Gb RAM, or am I confusing
> > issues?
> >
> > Any ideas gratefully received.

--
Tony Lawrence (to...@aplawrence.com)
SCO articles, help, book reviews, tests,
job listings and more : http://www.ApLawrence.com

John Gray

unread,
Dec 9, 1999, 3:00:00 AM12/9/99
to
Graham Nicholls wrote:

> I have a client who has a system which had 512M RAM, 2 PII/333
> processors, SCSI, etc. It ran out of memory, so they upgraded. It took
> a while to get the memory working in the ASUS motherboard, but now it
> is, but the system runs very slowly.
>
> My question is: Is this a SCO or a cache issue? (or something else)
>
> I am coming to the conclusion that this is a cache issue. PII chips
> have 512k of L2 cache. Is this enough for 1Gb RAM, or am I confusing
> issues?
>
> Any ideas gratefully received.

OSR 5.05 will support up to 4 GB of RAM. Make sure that you have enough
disk buffers by default you will get only 6K buffers ( 6MB) and if you do
raise
the buffers be sure to reduce the PLOWBUFS. You need only about 300-400
below 16 MB. You may be right on the money with the CACHE size.

Going from the 64k cache 16 MB RAM
128k 32
256k 64
512k 128
1024k 256
2048k 512
4096k 1024 MB RAM.

How big is your swap area? You might not be using all the new memory.
You might start using sar to see if there are any obvious bottlenecks.

Good luck
-john


Tony Blackwell

unread,
Dec 10, 1999, 3:00:00 AM12/10/99
to
Tony Lawrence wrote:
>
<snip>

> Something else is afoot here.
<snip>

> >
> > Graham Nicholls <gra...@rockcons.co.uk> wrote in message
> > news:384EACCE...@rockcons.co.uk...
> > > I have a client who has a system which had 512M RAM, 2 PII/333
> > > processors, SCSI, etc. It ran out of memory, so they upgraded. It took a while to get the memory working in the ASUS motherboard, but now it is, but the system runs very slowly.
<snip>

Could you tell us just what the problem was configuring the memory =
what did you do to get it working? How does it differ from the memory
already in the unit?

Tony Blackwell

Graham Nicholls

unread,
Dec 10, 1999, 3:00:00 AM12/10/99
to
Tony Blackwell wrote:
>
> Tony Lawrence wrote:
> >
> <snip>
> > Something else is afoot here.

My conclusion exactly.


BTW I certainly hope it isn't swapping. The extra memory was installed
in order to prevent swapping which was slowing the system drastically.

>>, but now it is, but the system runs very slowly.
> <snip>
>
> Could you tell us just what the problem was configuring the memory =
> what did you do to get it working? How does it differ from the memory
> already in the unit?

Well, at first we couldn't get 1Gb working at all - the machine would
just not boot at all. (IIRC 3 beeps on boot, but thats secondhand, as I
wasn't on site, and therefore may be a red herring.) Then ASUS told us
to use registered RAM, which I had never heard of. (anyone fancy
explaining that - I'd guess an extra address line to access more memory
?? ) They got hold of some registered RAM, and that didn't work. In
each case, the memory was replaced, rather than added to the existing
system. Now, there are two very similar servers on site - one live, and
one backup system.
Unfortunately, they both have different motherboards. One is an ASUS,
the other a Gigabyte. Neither works properly. As soon as we add the
512Mb of RAM, the system either fails to run, except with the L2 cache
turned off (it then _seems_ to fly, until lots of users are on, when it
runs out of processing steam - idle times around zero), or on the other
motherboard (the ASUS, IIRC) it crawls along. In the old days ;-) I
remember that adding memory could cause a system to be unable to cache
all of memory, and it ran slowly. PIII chips have 512k cache - perhaps
enough for 512Mb, but not 1Gb - is this a recurrence of that problem, or
am I confusing issues? ( I had thought that the on chip cache was
entirely different)

Anyway, there's the problem in some more detail.
Thanks all for your responses, and if you've any further ideas, such as
which kernel parameters to tune... (its surely too drastic a problem to
be this?)
Graham

Roberto Zini

unread,
Dec 10, 1999, 3:00:00 AM12/10/99
to
Graham Nicholls wrote:
>
> I have a client who has a system which had 512M RAM, 2 PII/333
> processors, SCSI, etc. It ran out of memory, so they upgraded. It took
> a while to get the memory working in the ASUS motherboard, but now it

> is, but the system runs very slowly.
>
> My question is: Is this a SCO or a cache issue? (or something else)
>
> I am coming to the conclusion that this is a cache issue. PII chips
> have 512k of L2 cache. Is this enough for 1Gb RAM, or am I confusing
> issues?
>
> Any ideas gratefully received.

I saw other good replies on the group, to which I'd dare to add;
please check carefully against Intel if the amount of cache your
system is equipped with it properly sized to handle that amount
of RAM. We've witnessed some cache related problems which caused
the machine to crawl when adding more memory.

Also, you could try by using the iddeftune(ADM) command which will
tune some kernel parameters accordingly to the amount of memory
you have; please notice that, as per the manual, this command
is effective only when you increase the amount of memory in a system
from less than 32MB so I'm not sure about the impact it will have
on your systen, but give it a try.

Best,
Roberto
--
---------------------------------------------------------------------
Roberto Zini email : fr...@strhold.it
Technical Support Manager -- Strhold Sistemi EDP Reggio Emilia(ITALY)
---------------------------------------------------------------------
"Has anybody around here seen an aircraft carrier?"
(Pete "Maverick" Mitchell - Top Gun)

Tony Lawrence

unread,
Dec 10, 1999, 3:00:00 AM12/10/99
to
John Gray wrote:
>
> How big is your swap area? You might not be using all the new memory.

I don't know if you meant to imply that he'd need swap to
see the memory, but someone could certainly read it that
way, and that's not the case. See
http://www.aplawrence.com/Boot/swap.html

Tony Lawrence

unread,
Dec 10, 1999, 3:00:00 AM12/10/99
to
Graham Nicholls wrote:
>
>
> BTW I certainly hope it isn't swapping. The extra memory was installed
> in order to prevent swapping which was slowing the system drastically.

Well, you don't have to guess: sar will tell you if you are
swapping and what your memory usage is: see
http://www.aplawrence.com/Boot/swap.html


> Unfortunately, they both have different motherboards. One is an ASUS,
> the other a Gigabyte. Neither works properly. As soon as we add the
> 512Mb of RAM, the system either fails to run, except with the L2 cache
> turned off (it then _seems_ to fly, until lots of users are on, when it
> runs out of processing steam - idle times around zero),

Is this from cpusar? And what's sucking the time- user or
sys?

>or on the other
> motherboard (the ASUS, IIRC) it crawls along. In the old days ;-) I
> remember that adding memory could cause a system to be unable to cache
> all of memory, and it ran slowly. PIII chips have 512k cache - perhaps
> enough for 512Mb, but not 1Gb - is this a recurrence of that problem, or
> am I confusing issues? ( I had thought that the on chip cache was
> entirely different)

It's an SMP system- does cpusar show that both processors
keep performing well after the memory is added? And how
does that compare to it before adding the memory?

You said you added this memory because it was swapping and
running slowly. Was that somebody's guess as to why it was
running slowly or was it actually swapping?

Graham Nicholls

unread,
Dec 10, 1999, 3:00:00 AM12/10/99
to
Tony Lawrence wrote:
>
> Graham Nicholls wrote:
> >
> >
> > BTW I certainly hope it isn't swapping. The extra memory was installed
> > in order to prevent swapping which was slowing the system drastically.
>
> Well, you don't have to guess: sar will tell you if you are
> swapping and what your memory usage is: see
> http://www.aplawrence.com/Boot/swap.html
>
> > Unfortunately, they both have different motherboards. One is an ASUS,
> > the other a Gigabyte. Neither works properly. As soon as we add the
> > 512Mb of RAM, the system either fails to run, except with the L2 cache
> > turned off (it then _seems_ to fly, until lots of users are on, when it
> > runs out of processing steam - idle times around zero),
>
> Is this from cpusar? And what's sucking the time- user or
> sys?
>
> >or on the other
> > motherboard (the ASUS, IIRC) it crawls along. In the old days ;-) I
> > remember that adding memory could cause a system to be unable to cache
> > all of memory, and it ran slowly. PIII chips have 512k cache -
>
>
> It's an SMP system- does cpusar show that both processors
> keep performing well after the memory is added? And how
> does that compare to it before adding the memory?
>
Idle times of zero on both processors, apparently.

> You said you added this memory because it was swapping and
> running slowly. Was that somebody's guess as to why it was
> running slowly or was it actually swapping?
>

No. It was swapping. Not now. Just running slowly - with very low idle
times.

If it is a cache issue, then why have some respondants reported very
good performance with MP systems with 1Gb RAM and only 256k
cache chips ?

Mystified.... (may turn out to be an obscure motherboard issue)


Thanks.

Graham.

Graham Nicholls

unread,
Dec 10, 1999, 3:00:00 AM12/10/99
to
Tony Lawrence wrote:
>

> Well, you don't have to guess: sar will tell you if you are
> swapping and what your memory usage is: see
> http://www.aplawrence.com/Boot/swap.html


A very useful page, BTW. I have long suspectetd that swap wasn't
strictly necessary, and this seems to confirm my opinion. SCO still
offer guidelines of 2.5 x RAM, don't they - which would be a little
silly in this instance. I want my systems never to swap, except maybe
under extreme provocation, when it is better that they swap than crash.
My attitude is: memory's cheap, bung some more in. Swapping is
expensive - don't do it. Especially, dont do it lots, so theres no
point in having very large swap spaces.

Graham

Bela Lubkin

unread,
Dec 10, 1999, 3:00:00 AM12/10/99
to
Graham Nicholls wrote:

> If it is a cache issue, then why have some respondants reported very
> good performance with MP systems with 1Gb RAM and only 256k
> cache chips ?

There is no linear relationship between size of cache and amount of RAM
it can cache. This is a common misconception.

Cache designs vary greatly. One of the design parameters is the width
of tag RAM entries. Tag RAM entries identify, for each location in the
cache, which location in main memory is represented.

One common tag RAM width is 8 bits. An 8-bit tag can specify one of 256
locations. An 8-bit tag combined with a 512K direct-mapped cache could
cache 256 * 512K == 128MB of RAM. However, most caches today are not
direct; 2-way and 4-way associative caches are more common. An N-way
associative cache is like N direct-mapped caches of (total size/N). So
a 512K 4-way associative cache with 8-bit tags could cache 256 * (512K /
4) == 32MB of RAM. Such a design is improbable; it would almost
certainly be designed with wider tags. 4-way 512K with 12-bit tags
would be able to cache 2^12 * 512K / 4 == 512MB.

You didn't have a performance problem at 512MB, and you do at 1GB. So
your cache parameters are some combination which tops out at 512MB. We
know one parameter (512K), and we can guess the associativity (1
(direct), 2- or 4-way). The possible combinations are:

associativity tag length cache size memory coverage
============= ========== ========== ===============
direct-mapped 10 512K 512MB
2-way associative 11 512K 512MB
4-way associative 12 512K 512MB

You might be able to fix this. Go into BIOS setup, look for "advanced
chipset setup" or whatever it offers. See if it has a "way-ness" knob.
If so, reduce it from 4-way to 2-way, or 2-way to direct. With the same
tag length, this should double the amount of memory covered.

The Pentium III's L2 cache is inside the CPU. I'm not really sure
whether you get to control any of this; and if you do, I'm not sure your
BIOS setup will give you access to it. But it's worth a try.

I tried to find documentation on cachability ranges for Intel CPUs, but
gave up after a while. It doesn't seem to be very public information.

>Bela<

Tony Lawrence

unread,
Dec 11, 1999, 3:00:00 AM12/11/99
to
Bela Lubkin wrote:
>
> Graham Nicholls wrote:
>
> > If it is a cache issue, then why have some respondants reported very
> > good performance with MP systems with 1Gb RAM and only 256k
> > cache chips ?
>
> There is no linear relationship between size of cache and amount of RAM
> it can cache. This is a common misconception.
>
> Cache designs vary greatly. One of the design parameters is the width
> of tag RAM entries. Tag RAM entries identify, for each location in the
> cache, which location in main memory is represented.

If Bela's explanation didn't make your brain hurt, and you
want to learn even more about caches, I strongly recommend
Unix Systems for Modern Architecture by Curt Schimmel (
http://www.aplawrence.com/Books/smp.html ). It's
unfortunately out of print right now, but it will tell you
more about caches than any sane human would want to know.

Personally, I can only grok this stuff for brief periods of
time- and the older I get, the more brief the time..

Graham Nicholls

unread,
Dec 11, 1999, 3:00:00 AM12/11/99
to
Tony Lawrence wrote:
>
> Bela Lubkin wrote:
> >
> > Graham Nicholls wrote:
> >
> > > If it is a cache issue, then why have some respondants reported very
> > > good performance with MP systems with 1Gb RAM and only 256k
> > > cache chips ?
> >
> > There is no linear relationship between size of cache and amount of RAM
> > it can cache. This is a common misconception.
> >
> > Cache designs vary greatly. One of the design parameters is
> If Bela's explanation didn't make your brain hurt,

It did! (but was also very interesting, and useful - especially in
explaining why other similarly cached systems perform better).

> Personally, I can only grok this stuff for brief periods of
> time- and the older I get, the more brief the time..
>

Likewise ;-(
Graham

Roberto Nunin

unread,
Dec 14, 1999, 3:00:00 AM12/14/99
to
At our site, we have 5 OSE 505 MP servers with 1 GB of memory and all
works fine.

Roberto Nunin
Comifar Service SPA Italy
"Marco Ce." <marc...@yahoo.nonlometto.it> scrisse:

>I was know that OSR 5.05 does not support more than 512 MB Ram.
>Try to verify this.

Roberto Nunin
Comifar Service S.p.A. ITALY
e-mail : Robert...@comifar.it

TINGB

unread,
Dec 19, 1999, 3:00:00 AM12/19/99
to
We too are having problems with an SMP box: 4 P-Pros with 1gig RAM on an HP.
Used to run great, but gradually slowing, with high load averages when nothing
is going on. But since it DID run well, we know it's possible with 1gig of RAM.

Good luck!

Ting Barrow
ti...@adm.com

Reply all
Reply to author
Forward
0 new messages