ADSM-L

db space required for multiple versions, etc

1996-01-22 14:18:54
Subject: db space required for multiple versions, etc
From: Greg Tevis <gtevis AT VNET.IBM DOT COM>
Date: Mon, 22 Jan 1996 11:18:54 PST
The adsm server database requires the same amount of db space to save
each extra (inactive) version of a file as it does for the active
version.  So the amount of db space required is (almost) proportional to
the number of versions kept...(further discussions below)

db space (for backup) = v*n*A + K

where v = # of versions kept
      n = # of files backed up to server
      A = db space req'd for each db entry..varies with filenames, ACLs, etc
          typically around 500 bytes.
      K = constant representing db storage used for policies/schedules/
          node/admin definitions, etc...obviously this isn't a true
          constant since this changes but it is relatively constant
          (and insignificant) compared to the data storage info.


I must mention that this does not hold true for file copies made using
v2 copy storage pools.  Copy storage pool files do NOT take the
same amount of db space as active/inactive primary copies....they take
around 150-200 bytes of db space (1/2 or less the space required
for regular db entries).

Archive files take around the same as backup files (maybe a little
more for description info).

Also, your data reference patterns affect db space...for example,
files that don't change very often may have inactive versions
deleted due to retention period...this means fewer versions are
actually kept than is specified by the number of versions.

Also, other things affect db space such as caching...each file
kept in cache (disk pool after already migrated to tape) costs
about the same as a copy pool file entry (ie, about 1/2 a regular
db entry).

so...here's a modified equation (but still with lots of caveats...
like assuming only 1 copy file for each backup or archive file)...

db space = v*n*A + a*A + (n+a)*B + c*C + K

where v = # of versions kept
      n = # of files backed up to server
      a = # of files archived to server
      c = # of files cached in disk storage pool
      A = db space req'd for each db entry..varies with filenames, ACLs, etc
          typically around 500 bytes.
      B = db space req'd for each copy of a file in a copy storage pool  etc
          typically around 150-200 bytes.
      C = db space req'd for each cached file...around 200 bytes?        etc
      K = constant representing db storage used for policies/schedules/
          node/admin definitions, etc...obviously this isn't a true
          constant since this changes but it is relatively constant
          (and insignificant) compared to the data storage info.


...Greg Tevis, adsm technical support
<Prev in Thread] Current Thread [Next in Thread>