| Subject | Re: [firebird-support] Firebird performance issues on VM platform |
|---|---|
| Author | Helen Borrie |
| Post date | 2010-09-10T03:04:40Z |
At 09:36 AM 10/09/2010, Mike Kenyon wrote:
-- Are you getting actual 8GB physical RAM now, or are you mostly in swap?
-- Did you have 7200 or 9600 rpm hard disk before on sub-terabyte HDDs, whereas now you are sharing space on a HDD >= 1 TB? (Curious about that as here in NZ, I haven't seen any TB-plus HDDs available w. rpm speeds faster or slower than this oddball "5700 rpm")
-- Is page cache configured on the assumption that Firebird still has oodles of RAM available? (Page cache is determined as page_size * no. of pages).
Old-hat stuff:
--hopefully your database partition is excluded from System Restore. If not, it needs to be.
--has anyone meddled with firebird.conf and tried to force the engine's cpu affinity on more than one cpu? Alternatively, if the engine is configured for the default (cpu 0) and the VM's affinity is forced to another (or others)...might be a mess. CPU Affinity MUST be to one and only CPU for SuperServer.
--was the database file-copied onto the colocation server, or gbak-restored? (Could be lots of garbage there if it was file copied.)
This is by no means an exhaustive checklist, btw.
Incidentally, though I cc-ed this to your tech colleague, don't expect this as normal for lists. If your colleague needs to see the responses, he'll need to subscribe himself...
./hb
>We recently moved/migrated our hosted server (physical server) to a VMCompared to
>platform - we are now running Windows Server 2008 R2 64 bit OS. We have
>found the performance of Firebird to be lacking on this OS / VM platform.
>We have tried increasing the physical resources, but this does not help.
>What we had before that worked:how do the resources stack up?
> Physical hosted server, quad core, 8Gb RAM, Windows Server 2008 R1
-- Are you getting actual 8GB physical RAM now, or are you mostly in swap?
-- Did you have 7200 or 9600 rpm hard disk before on sub-terabyte HDDs, whereas now you are sharing space on a HDD >= 1 TB? (Curious about that as here in NZ, I haven't seen any TB-plus HDDs available w. rpm speeds faster or slower than this oddball "5700 rpm")
-- Is page cache configured on the assumption that Firebird still has oodles of RAM available? (Page cache is determined as page_size * no. of pages).
Old-hat stuff:
--hopefully your database partition is excluded from System Restore. If not, it needs to be.
--has anyone meddled with firebird.conf and tried to force the engine's cpu affinity on more than one cpu? Alternatively, if the engine is configured for the default (cpu 0) and the VM's affinity is forced to another (or others)...might be a mess. CPU Affinity MUST be to one and only CPU for SuperServer.
--was the database file-copied onto the colocation server, or gbak-restored? (Could be lots of garbage there if it was file copied.)
This is by no means an exhaustive checklist, btw.
Incidentally, though I cc-ed this to your tech colleague, don't expect this as normal for lists. If your colleague needs to see the responses, he'll need to subscribe himself...
./hb