Chapter 2 HPSS Planning
HPSS Installation Guide September 2002 115
Release 4.5, Revision 2
bitfiles and bytes stored by the account index in the given COS, 2) if the storage class is not 0, this
is a statistics record and contains the total number of bitfile accesses and bytes transferred
associated with the account index, COSid, and referenced storage class. The number of records in
this file should be the number of users in the HPSS system multiplied by the average number of
levelsin a hierarchy plus 1. For most configurations, the average number of levels will be 2. For sites
with two copies configured, the average number of levels will be 3.
AccountingLogging Records. This file is used by the BFS to log updates to the (acctsum) records. It
consistsof values and flags plus the same information as defined in the (acctsum) file. The size of
thefile will depend upon load, but the maximum number of records for planning purposes should
be 3000.
2.10.2.6 Disk Storage Server Metadata
The Disk Storage Server is the primary user of the following SFS metadata files (the default SFS
filenames are shown in parentheses):
• SS Disk Storage Maps (storagemapdisk.#)
• SS Disk Storage Segments (storagesegdisk.#)
• SS Disk Physical Volumes (sspvdisk.#)
• SS Disk Virtual Volumes (vvdisk.#)
EachDisk Storage Server requires its own set of four metadata storage files: one file for disk storage
maps,one for disk storage segment metadata records, one for virtual volume metadata records, and
onefor physical volume metadata records. If a system has more than one Disk Storage Server, each
server must have its own set of files. Files must not be shared between servers.
SSDisk Storage Maps. There will be as many disk storage map records and virtual volume records
asthere are disk virtual volumes, and as many disk physical volume records as there are disk units,
butsince the number of volumes is likely to be relatively small (compared to the likely number of
tape volumes), these files will probably be small. Also, since increasing or decreasing the number
of disk virtual volumes is likely to be a rare event, these files will usually be a constant size. Even
for relatively large disk systems, the consumption of SFS storage resources by disk storage server
map, virtual volume, and physical volume metadata will be small.
SS Disk Storage Segments. The number of disk storage segment metadata records is likely to be
larger than any of the other disk storage server metadata types, but it too is bounded. In the
maximumcase, each disk virtual volume (VV) can support 16,384 disk storage segments, each with
itsown metadata record. The upper bound for a disk storage segment metadata file is 16,384 times
the number of VVs.
If the default storage segment size for the storage class is a multiple of the VV block size (and it
probably will be), the maximum number of disk storage segments is reduced by that factor. For
example, if the length of disk storage segments in a particular storage class is 4 VV blocks, the
maximumnumber of storage segments that can exist on such a VV is 16,384 divided by 4 or 4,096.
Theactual maximum number in this case is 4097, because the first VV block on each VV is reserved
for volume label information.