Chapter 7 HPSS User Interface Configuration
HPSS Installation Guide September 2002 441
Release 4.5, Revision 2

7.6.2.6.2 DMAP Gateway Server

The DMAP Gateway is a conduit and a translator between HDM and HPSS. HPSS servers use
DCE/RPCsto communicate, however the DMAP Gateway encodes requests using XDR and sends
these requests via sockets to HDM. In addition, it translates XDR from the HDM to DCE/TRPC/
Encinacalls to the appropriate HPSS servers. When a connection between the HDM and Gateway
is made, mutual authentication occurs.
The DMAP Gateway keeps a record of all the managed filesystems and HPSS filesets it manages.
For scalability, multiple DMAP Gateways are supported. However, a given DMAP Gateway will
only operate on the filesets it manages. At this time, the DMAP Gateway only supports filesets
managed by DFS and filesystems managed by XFS. Additionally, a DMAP Gateway can only
interact with the Name Server(s) within its HPSS site.
In addition to translating requests between HDM and HPSS servers, the DMAP Gateway keeps
fileset request statistics and internal DMAP Gateway resource utilization statistics. Thus, heavily
usedfilesets can be identified and management of these filesets could be distributed across multiple
DMAP Gateways to improve performance.
7.6.2.7 HPSS/DFS Usage Guide and Limitations
This section of the document provides information to help system administrators determine the
best way to use the HPSS/DFS interface. Both DFS and HPSS have limitations that will impact
performance when using the HPSS/DFS interface. Some of these limitations are the result of
implementation decisions, others are hardware dependent. This section will address these
limitations, explain why they exist, when and if the limitation will change, and how to configure
your system to minimize the impact of these limitations.

7.6.2.7.1 Fileset Anode Limitations

Thecurrent implementation of Episode allows a maximum of 2GB of space per fileset for anode and
fileattribute storage. This space includes anodes (file headers), space allocation, ACL’s, and space
used to keep track of DMAPI attributes (regions and extended attributes). In a DFS fileset, that is
not also managed by HPSS, this provides storage space for approximately 4 million files. The
number of files in a fileset is greatly reduced when the fileset is also managed by HPSS.
Theinformation stored to keep track of migration generally uses an extra fragment of anode space.
Sincewe recommend using a fragment size equivalent to the block size (8K), each file migrated into
HPSS will use a little more than 8K (assuming no extra ACL space). This puts the upper limit on
the number of files per fileset at about 250,000.
This limit may change in the future. If the fragment size can be reduced to 1K then the number of
files increases to about 1.5 million. However, using a 1K fragment size is not currently
recommended for two reasons:
1. Itis possible for an aggregate to appear to have plenty of space but that space could be allo-
cated as fragments that cannot be used for files greater than the block size. This leads to
problems when trying to determine how much data needs to be purged to free up DFS
space.
2. Fragments can not be purged. Only files using blocks can be purged.