IBM Hub/Switch Installation Manual page 440

Chapter 7 HPSS User Interface Configuration
To support persistent DM-related metadata, XFS utilizes its standard extended attribute facility.
DM attributes, event masks, managed regions, and attribute change times (dtime values) are stored
as extended attributes. These extended attributes are treated as file metadata.
The xfsdump and xfsrestore utilities include extended attributes and migrated regions. Migrated
data is not recalled when a dump is taken, producing an abbreviated dump.
7.6.2.6 HPSS Components Required to Support DFS and XFS
An HPSS Data Management Application (HPSS/DMAP or HDM) and a DMAP Gateway must be
configured into the HPSS system if an integrated HPSS/DFS or HPSS/XFS system is to be
supported at your site.
7.6.2.6.1 HPSS/DMAP (HDM) Server
HDM is responsible for initiating and coordinating name space and data interactions between the
managed file system and HPSS. It catches and processes desired name and data space events
generated by the managed file system, migrates file data to HPSS, purges unneeded file data from
the managed file system after migration, and processes requests originating in HPSS interfaces that
require either name or data resources from the managed file system. HDM resides on the same
machine where the managed file system is running. HDM communicates with HPSS components
via XDR over TCP sockets.
HDM registers to receive name and data space events originating in the managed file system. After
catching a name space event involving a mirrored fileset, the appropriate requests are made to
HPSS to keep the name spaces synchronized. However, for archived filesets, only create and
destroy name space events are processed, but no HPSS resources are utilized until the file is
migrated to HPSS. HDM receives data space events when a file is read, written, or truncated and
the data region involved is registered via XDSM. HDM is responsible for caching data to the
managed file system that is not present; invalidating HPSS data, if necessary; and manipulating
data regions to minimize the occurrence of events involving the same region. HDM can cache
partial files with DFS but must cache whole files for XFS due to a constraint on the number of
managed regions supported by XFS.
HDM provides an interface for HPSS to request that an action occur in DFS to keep the HPSS and
DFS name and data spaces synchronized. This mechanism is only used with mirrored filesets when
an HPSS client requests to create, delete or modify a name space object. This interface is also used
by HPSS to migrate or purge data from Episode disks. Before file data can be altered through HPSS
interfaces, the data must first be purged from Episode disks. The capability to forward DCE
credentials is provided, enabling HDM to make DFS requests on behalf of the end user.
HDM migrates file data from the managed file system to HPSS. It also purges file data from the
managed file system to free disk resources. For DFS, only data that has been modified on the
managed file system is migrated to HPSS, thus, a minor modification to a very large file will not
result in re-migrating the entire file. Because policies for migrating and purging data are separately
configurable, file data migrated from a managed file system is not automatically purged. In many
cases, data for a given file will be present in both the managed file system and HPSS and can be
read from either interface without any data migration or staging.
440 September 2002 HPSS Installation Guide
Release 4.5, Revision 2