All:

I see that that DMS tablespaces using raw devices have "File System
cache" enabled by *default*. My understanding is that file system cache
is applicable only for SMS.

Here is the tablespace snapshot:

Tablespace name = TYESDATA1
Tablespace Type = Database managed space
File system caching = Yes


Any ideas why file system cache this is enabled by *default* for DMS
raw?
Does it affect performance given that DMS does not do double buffering?

Thanks much for the help!

Vijay

UDBDBA wrote:
> All:
>
> I see that that DMS tablespaces using raw devices have "File System
> cache" enabled by *default*. My understanding is that file system cache
> is applicable only for SMS.

File system caching applies to DMS file, too, but not to DMS raw. Your
snapshot doesn't show it, but are your containers actually raw devices
or just files in a filesystem?

The default has *always* been for DB2 to use the filesystem cache, and
only recently (DB2 UDB V8.2) has added the ability to disable the cache
via an ALTER TABLESPACE command. Prior to this there were other options
such as using Direct IO or Concurrent IO (AIX), DB2_NT_NOCACHE (Win32),
etc. The ALTER TABLESPACE syntax has abstracted these methods into a
common statement usable across all platforms.

FYI When IBM adds new features such as this they generally must be
explicitly enabled to activate the feature for some time before they
become the default. For example, Hash Joins were introduced in V5.2,
but were not enabled by default until V8.1.

Thanks. The DMS are raw logical volumes. So, i guess caching is not
used although it is *always* enabled by default.

Vijay

UDBDBA wrote:
> Thanks. The DMS are raw logical volumes. So, i guess caching is not
> used although it is *always* enabled by default.
>

This is probably just an instrumentation problem - the snapshot monitor
probably looks at a bit in the tablespace header to determine whether
caching has been disabled for the tablespace, but of course this doesn't
apply if the containers are devices.

You could probably open a PMR to have this corrected.

Thanks Ian