Summary of Consideration s 7
Installationshaving these devices could use sequential caching as an installation
option. Installationswith a m ixtureof devices with large, small, or no cache can
benefit from the bypasscache opti on.
2.2.3 Log Structured Fil e
Devicesusing the log structured file technique (like the RVA) do not maintain data
location during data updates. Fort hese devicesthere exists a concept of logical
location of data, independentf romthe the physical location. The logical location
is used by the devicet o presentt he data to the application: the user sees a
contiguous extent on a 3390 volume,while the data is in reality scattered across
the LSF.
A REORG of a DB2 table space provides a logical sequence of records which
could not be corresponding to a physical sequence. Thisi s a function of the
space management of the storage server.
Worrying about reorganizingdata to reclaim space extents is now much less
criticalwith the new disk architecture. REORG does not need to be run in order to
reclaim fragmented space in this case, only to reestablishthe clustering (logical)
sequence and the DB2 internalfree space. When the DB2 optimizer chooses
sequential prefetchas a valid access path, the storage server detects the logical
sequential access and initiatespre-staging of the logically sequenced tracks into
cache, providingimprovement to the I/O response time for the subsequent
prefetchac cesses.
2.2.4 RAMAC Architecture
Disk architecture defines each volume in a logicalway through tables. These tables
do an effectivemapping be tweenth elogical view of the volume onto the disk array
withda ta and rotating parity physical disks. This means that each I/O operation takes
place to or from several physical disks.H owever,the host still views only the logical
volume topology,and it bases its optimizing and scheduling strategies on this view,
as it used to do withn ative3380 and 3390.
2.2.5 SMS Storage G roups
Volumeseparation is easy when you have hundreds of volumes available. But
this separation isgood only if your volumes haveseparate access paths. Path
separation isimpor tant toachieve high parallel data transfer rates.
Without DFSMS, the user is responsiblefor distributing DB2 data sets among
disks. Thisproc essneeds to be reviewedperiodically, either when the workload
changes, or when the storage server configurationchanges.
With DFSMS,the user can distribute the DFSMS Storage Groups among storage
serverswith the pur poseof optimizing access parallelism. Another purpose could
be managing availability fordisa sterr ecoveryplanning. This can be combined
with the previouspurpose by letting DFSMS automaticallyfill in these Storage
Groups with datasets, by applying policies defined in the automatic class
selection routines.
Changes to the topology of the Storage Group can be managed to minimize the
application outages. Thiscan be done simply by adding new volumes to the
Storage Group,then m anagingt he allocationenablement (opening it on new