IT30069: Files processed by the Spectrum Protect Server Inline deduplication ratioshich can affect performance and

8.1.8.100-IBM-SPSRV-AIX
8.1.8.100-IBM-SPSRV-Linuxppc64le
8.1.8.100-IBM-SPSRV-Linuxs390x
8.1.8.100-IBM-SPSRV-Linuxx86_64
8.1.8.100-IBM-SPSRV-WindowsX64
8.1.8.100-IBM-SPCMS-Linuxx86_64
8.1.8.100-IBM-SPOC-AIX
8.1.8.100-IBM-SPCMS-WindowsI32
8.1.8.100-IBM-SPOC-LinuxPPC64le
8.1.8.100-IBM-SPCMS-WindowsX64
8.1.8.100-IBM-SPOC-Linuxs390x
8.1.8.100-IBM-SPOC-Linuxx86_64
8.1.8.100-IBM-SPOC-WindowsX64
8.1.9.000-IBM-SPOC-AIX
8.1.9.000-IBM-SPSRV-AIX
8.1.9.000-IBM-SPOC-LinuxPPC64le
8.1.9.000-IBM-SPOC-Linuxs390x
8.1.9.000-IBM-SPSRV-Linuxppc64le
8.1.9.000-IBM-SPCMS-Linuxx86_64
8.1.9.000-IBM-SPOC-Linuxx86_64
8.1.9.000-IBM-SPCMS-WindowsI32
8.1.9.000-IBM-SPSRV-Linuxs390x
8.1.9.000-IBM-SPCMS-WindowsX64
8.1.9.000-IBM-SPOC-WindowsX64
8.1.9.000-IBM-SPSRV-Linuxx86_64
8.1.9.000-IBM-SPSRV-WindowsX64
IBM Spectrum Protect Server V8.1.8.X interim fix downloads
IBM Spectrum Protect Server V8.1 Fix Pack 8 (V8.1.9) Downloads

APAR status

• Closed as program error.

Error description

• When a file is ingested, and processed by the inline
  deduplication engine (server-side), the file
  size is used to determine what chunking tier to use. This will
  dictate what the average chunk size
  will be for the given object. The following are defaults for the
  chunk tiers:
  
  99GB or less: Deduplication tier 1 is used which gives a minimum
  chunk size of 50KB
  100GB -> 399GB:  Deduplication tier 2 is used which gives a
  minimum chunk size of 250KB
  400GB and Larger:  Deduplication tier 3 is used which gives a
  minimum chunk size of 750KB
  
  What should be happening is that as each object is ingested and
  processed, the size for that given
  object should dictate what chunk size mask should be used. This
  APAR was opened to address the
  issue where the inline deduplication engine is using the size
  from the very first object processed
  on a given session as the chunk size mask to be applied to all
  subsequent objects processed on that
  given session.
  
  For example, if a 400GB file is processed, a 750KB minimum chunk
  size will be used for that object
  and any subsequent object processed on the session. If a 10MB
  object was processed right after, it
  would also have a 750KB minimum chunk size footprint as well. If
  the 10MB object was processed
  first, a 50KB minimum chunk size footprint would be assigned to
  both objects.
  
  If a large chunk sizing is applied to a smaller object, this can
  affect the overall deduplication
  rate for that object. If a smaller chunk size is applied to a
  larger object, this can affect
  performance for the management of that object (DB IOPS) and for
  the retrieval of that object (DISK
  IO).
  
  This is not a common problem since very large files are
  typically processed separately from smaller
  type files. In addition, very large files are typically
  deduplicated and compressed on the
  client-side for which this APAR does not apply. This exposure is
  for server-side inline
  deduplication only.
  
  The following SQL can be used to determine what chunk sizing has
  been used for a given object:
  
  db2 "select avg(cast(size as bigint)) as avgchunksize,
  min(cast(size as bigint)) as
  minchunksize,max(cast(size as bigint)) as maxchunksize from
  sd_chunk_locations sdcl where exists
  (select 1 from sd_recon_order sdro where
  sdro.chunkid=sdcl.chunkid and sdro.objid=<object id>) for
  read only with ur"
  
  The RABIN trace class can be used to determine what chunk size
  mask is being applied as an object
  is being processed. The following trace point will be found:
  
  Using Tier X for file of size XXXXXXXXXXXXX
  
  In the above example, X will be either 1,2 or 3.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * All IBM Spectrum Protect server users of container type      *
  * storage pools.                                               *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * See error description.                                       *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Apply fixing level when available. This problem is currently *
  * projected to be                                              *
  * fixed in levels 8.1.7.100, 8.1.8.100 and 8.1.9. Note that    *
  * this is subject to change at the discretion of IBM.          *
  ****************************************************************
  

Problem conclusion

• This problem was fixed.
  Affected platforms for reported release:  AIX, Linux, and
  Windows.
  Platforms fixed:   AIX, Linux, Windows.
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  TSM SERVER

• Fixed component ID

  5698ISMSV

Applicable component levels