IBM Hub/Switch Installation Manual page 131

High performance storage system release 4.5
Table of Contents

Advertisement

Encina SFS Data
Actual HPSS metadata is stored in SFS data volumes. SFS data volumes store the actual SFS record
data as well as associated index and B-tree overhead information. SFS must have sufficient disk
space allocated for its data volumes in order to store the projected amount of HPSS metadata.
The metadata sizing spreadsheet, discussed in the previous section, needs to be used to calculate
the size of each SFS data volume and the distribution of the SFS data files across the data volumes.
Encina Transaction log
The Encina transaction log is a wrap-around log used to store transaction information. It records all
changes in SFS data on a per-transaction basis. In the event of a system or server crash, upon restart
the SFS server reads the transaction log to determine what, if any, transactions were in progress but
not committed and properly aborts all data changes associated with that transaction. Likewise, it
properly updates all SFS data for transactions that had been committed but not written to disk in
the SFS data volumes.
The optimal type of disk used for the transaction log is a low-latency disk since the I/O is
performed in many small chunks. To ensure full recoverability in the event of a media failure on
the transaction log disk, it is important to mirror the transaction log on a separate physical disk.
The size of the transaction log is dependent on the volume of HPSS transactions. Start with no less
than 256 MB (512 MB for larger installations). It is best to start small and increase the size if
necessary.
Since I/O delays associated with the transaction log can affect HPSS performance, it is
recommended that few, if any, other logical volumes be placed on the same disks used for the
transaction log. For example, it would be unwise to use the rootvg volume group for the transaction
log or put the transaction log on the same disk where paging space has been created.
Encina Media Recovery Archive Files (MRA files)
Because media failures may occur between SFS backups, it is important to be able to replay all SFS
transactions from the last SFS backup up to the point of a media failure. Encina "media archiving"
is the feature that enables this level of recovery. When media archiving is enabled, a running
backup of the transaction log is written to disk. The transaction log wraps around, but media
archive files do not. The administrator can then move these media archive files to offline storage for
safe keeping. If media archiving has not been enabled prior to a media failure, the SFS data cannot
be restored.
Because media failures can occur where these backup files are written, mirroring can be considered
for improved reliability. It is recommended that at least 2 GB of disk space be available for the
media recovery archive files.
HPSS Installation Guide
Release 4.5, Revision 2
Any SFS data file that is estimated to be greater than 1 GB needs to reside on its own
SFS data volume.
A large log volume will increase SFS start time.
September 2002
Chapter 2
HPSS Planning
131

Advertisement

Table of Contents
loading

This manual is also suitable for:

Hpss

Table of Contents