User's Manual

HP StorageWorks File System Extender Software user guide 125
Storage space allocation in an extended FSE disk buffer
In an FSE implementation where several file systems are assigned to the FSE disk buffer, the Resource
Manager tries to maximize usage of these file systems. This is necessary in order to execute FSE jobs in the
most efficient manner, and to speed up FSE administrative, maintenance, and backup tasks.
For all FSE job types, a round robin scheduler is used to allocate individual file systems to different FSE
jobs. When an FSE job requests FSE disk buffer resources, the Resource Manager allocates the first file
system which fulfills storage space requirement of the job. If no such file system is currently available, the
allocation is retried after a short interval.
Since recall (efficient recall), reorganization, and maintenance (redundant copy recreation) jobs are
flexible in terms of storage space requirement for FSE disk buffer, these jobs provide the Resource Manager
two values: maximum and minimum amount of the required storage space. If a file system containing the
maximum amount is not available, the FSE job is assigned the first file system which contains the minimum
amount, and only if the latter is also not available, the allocation is delayed.
Hierarchical Storage Manager lists
All migration and release jobs are carried out using Hierarchical Storage Manager (HSM) lists. These lists
provide information about the candidates to be migrated or released and gather them accordingly. FSE
executes migration or release once for a group of files rather than executing these jobs separately for each
file thus improving performance. However, recall is executed for each file separately in order to fulfill the
user’s request for a specific offline file without delay.
HSM maintains three types of lists: a list of changed files—a so-called “dirty files” list—a migration
candidate list, and a release candidate list. Each file entry can only appear on one list at a time. Either it
is a dirty file, and should not be migrated yet, or it is a migration candidate and is waiting for migration,
or it is a release candidate and is waiting to be released. Once a file is released, it is cleared from the
release candidate list. The file is then offline until the next recall. Offline files are not a part of any list. Their
status is recorded in the HSM file system in the HSM attributes (in the native file system’s extended
attributes).
The way transitions between HSM lists and other aspects of migration, recall, release, and deletion are
controlled is defined by parameters in the configuration file for the FSE partition. These parameters are
described in ”Migration” on page 127, ”Release” on page 129, ”Recall” on page 136, ”Deletion” on
page 137, and in ”System allocation and job priority policy” on page 142 and ”Partition allocation and
job priority policy” on page 144.
Table 2 Types of HSM lists
FSE HSM list Description
Dirty file list
(list of changed files)
Gathers files that were recently accessed and changed. A dirty file can
be either a new file or a file that has changed since the last migration or
recall. A file is added to this list once you start changing it.
Migration candidate list Once a changed file is not modified for a certain period of time, the file
is moved onto the migration candidate list.
Release candidate list Gathers online files that have already been migrated and have not
changed since then, and files that were brought online by recall.