In article <20031014.112846.961220.NETNEWS AT VM.SAS DOT COM>, Rowan O'Donoghue
<R.ODonoghue AT UNITECH-IE DOT COM> says:
>
>All,
>
>Have a customer looking at subfile backups for two webservers which are
>hosted at a remote hosting facility and connected via a WAN.
>
>What have been your main issues? (apart from the obvious being expanded
>TSM processing time on the client, and slow restores).
We have not found an issue with the processing on the clients as
most newer pc's a fast and not used anywhere near 100% during off hours
when we are doing backups. Now the network is a different issue.

You can find a number of past comments on this subject. Even some from
IBM talking a little bit about how it works.

Our number one issue is that we have folks with greater then 2gig files
and TSM only does subfile backups on files less then 2gig in size.

The IBM folks gave the rules for when a new regular backup was used.
One reason was when the delta increased to x % of the whole file.

If you use this feature you should install the latest hot fix as there
is a fix for subfile restores when the combined size of the base and
delta exceed 2 gig.
>
>Also, how have you sized the subfile cache area? Was there a rule of thumb
>for this sizing?
We sized the cache for the max size. Note I believe that this does not
hold the delta's but the checksums for the blocks that make up the files.

>
>Rowan.

It looks like a number of other vendors are using incr forever and block
level backups.

len