Chapter 2 HPSS Planning
HPSS Installation Guide September 2002 121
Release 4.5, Revision 2
2.10.2.20 Metadata Constraints
The generic configurations for all HPSS servers must be contained in a single SFS file. SSM makes
this happen transparently. Also, all server-specific configuration entries for a given server type can
be in the same SFS file. For example, all Storage Servers should be defined in the single SFS file
named ss. The following metadata files should not be shared between servers; instead, each server
should use a distinct metadata SFS filename:
• MPS Checkpoints
• PVR Cartridges
• SS Disk Storage Maps
• SS Tape Storage Maps
• SS Disk Storage Segments
• SS Tape Storage Segments
• SS Disk Physical Volumes
• SS Tape Physical Volumes
• SS Disk Virtual Volumes
• SS Tape Virtual Volumes
2.10.2.21 Metadata Sizing Spreadsheet
To help with projecting disk space requirements for HPSS metadata, an Excel spreadsheet is
available on the HPSS web site at https://www4.clearlake.ibm.com/hpss/support/
ToolsRepository/md43.xls. To access the website, an account is required. To request an
account, use URL http://www4.clearlake.ibm.com/cgi-bin/hpss/requests/
id_request.pl. The spreadsheet is divided into two worksheets: the first defines the various
HPSS and SFS sizing assumptions; the second calculates the sizing estimates.
2.10.2.21.1 Metadata Sizing Assumptions
The metadata sizing assumptions are organized into multiple sections: HPSS dynamic variables,
HPSS dynamic variables for each subsystem, HPSS static configuration variables, and SFS sizing
assumptions. Note that all assumptions are with respect to sizing the maximum amount of
metadata required to support a planned configuration. It is recommended that sites use values for
multiple years into a project’s future to determine how the metadata sizing requirements will
increase with system growth.
HPSS Dynamic Variables. Table 2-5 defines the HPSS dynamic variables associated with the HPSS
system outside those specific to a subsystem. These variables will moderately impact the metadata
sizing estimates.
