Reads             (VSAM)

DESCRIPTION
The number of "Reads" represents the total number of physical reads issued to 
a "Dataset" since a Sun MTP system was initiated. A physical read is a read not 
accommodated by Sun MTP's internal buffers. This read may, however, be handled by
the operating system buffer cache and not actually require a physical disk access.


VALUES
"Reads" is an integer number that starts at zero when a Sun MTP system initiates
and increments for each physical read from (for KSDS, the data portion of) a 
"Dataset". Note that this datapoint is not set for remote datasets (those with 
a value set for the "RSys" field) because Sun MTPScan does not communicate with the 
remote region.


INTERPRETATION
This datapoint, together with "Idx Reads", "Writes", and "Idx Writes", provides 
an accurate depiction of dataset activity. Having multiple highly active 
datasets on the same physical device decreases system throughput. All four of 
these datapoints should be used to determine the extent of a dataset's activity.
Note that the "Writes" and "Idx Writes" entail more overhead than their read 
counterparts. 


CONCLUSIONS
There are a few ways to optimize dataset access performance. There are two
kinds of configurable cache involved with a VSAM file local to a host machine:
the operating system buffer cache and the Sun MTP buffer pool. Consult your
machines System Administrator's Guide for details on how to modify the systems
buffer cache.  For instance, on a UNIX system, the UNIX kernel must be rebuilt
with a new configuration in order to change the size of its buffer cache.
Remember that increasing the buffer cache chews up valuable memory. As for the
Sun MTP cache, the size of it is easily configured in the Sun MTP VSAM Control
Table (VCT). Change the value of "Number of shared buffers", write the VCT to
disk, and restart the system to activate a change.

There are also a few ways to distribute dataset files across hardware devices
to achieve performance gains. The simplest way is to physically place entire
datasets on separate disks. If the disks happen to have separate disk
controllers, so much the better.

Another method to take advantage of I/O bandwidth is to use disk striping. This
file system facility automatically distributes files across multiple devices.
Again, consult your machines System Administrator's Guide for details.

Sun MTP provides another method to boost performance by striping a dataset across
multiple segments, each of which may reside on a separate device. Sun MTP allows
a dataset to have at most 8 segments.  The size of each file segment depends on 
the type of system where Sun MTP resides.  On UNIX systems, each segment can 
contain up to 2 Gigabytes (G) of information (this limit is imposed by the 
UNIX file system). On Windows NT systems, the maximum size of each segment 
depends on the maximum file size offered by the file system in place. For the 
FAT file system, the maximum file size is 4 Gigabytes. For NTFS, the maximum 
file size is 2 Terabytes, although this limit is constrained by hardware limits
rather than the NT operating system. This method proves to be more efficient 
than the other two. You may choose to combine Sun MTP segmentation with operating
system striping, but the results are indeterminate since the file organization 
becomes extremely complex.
