SMP buffer management test question

Started by Nonamealmost 21 years ago4 messages
#1Noname
lsunley@mb.sympatico.ca

There was a mention of a "context-swap-storm" test in the posts on the ARC
patent. Where might I obtain a copy of this test procedure?

Thanks

--
-----------------------------------------------------------
lsunley@mb.sympatico.ca
-----------------------------------------------------------

#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: Noname (#1)
Re: SMP buffer management test question

lsunley@mb.sympatico.ca writes:

There was a mention of a "context-swap-storm" test in the posts on the ARC
patent. Where might I obtain a copy of this test procedure?

http://archives.postgresql.org/pgsql-performance/2004-04/msg00280.php

(The archiver seems to have lost the message's separation into parts,
but the "setup" script is the part down through the checkpoint command.)

Read the whole thread for context.

regards, tom lane

#3Noname
lsunley@mb.sympatico.ca
In reply to: Tom Lane (#2)
Re: SMP buffer management test question

Thanks

Well I tried it out and on OS/2 using 8.0 it has the same problems as on
Linux, with two clients running on a dual PIII 750 both CPUs go to 100%
and the system is madly switching between the two processes. With a
single client, about 35% CPU is used.

Should the new buffer manager patch eliminate this behaviour? That is what
I gathered from the discussions.

Lorne

In <15517.1108926080@sss.pgh.pa.us>, on 02/20/05
at 02:01 PM, Tom Lane <tgl@sss.pgh.pa.us> said:

lsunley@mb.sympatico.ca writes:

There was a mention of a "context-swap-storm" test in the posts on the ARC
patent. Where might I obtain a copy of this test procedure?

http://archives.postgresql.org/pgsql-performance/2004-04/msg00280.php

(The archiver seems to have lost the message's separation into parts, but
the "setup" script is the part down through the checkpoint command.)

Read the whole thread for context.

regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 7: don't forget to increase your free space map settings

--
-----------------------------------------------------------
lsunley@mb.sympatico.ca
-----------------------------------------------------------

#4Tom Lane
tgl@sss.pgh.pa.us
In reply to: Noname (#3)
Re: SMP buffer management test question

lsunley@mb.sympatico.ca writes:

Well I tried it out and on OS/2 using 8.0 it has the same problems as on
Linux, with two clients running on a dual PIII 750 both CPUs go to 100%
and the system is madly switching between the two processes. With a
single client, about 35% CPU is used.

Should the new buffer manager patch eliminate this behaviour?

It does for me ...

regards, tom lane