2. PERFORMANCE REPORTING ANALYSIS › 2.3 Auxiliary Storage Management Analysis › 2.3.4 MVS Concepts › 2.3.4.1 Auxiliary Storage Management Overview

2.3.4.1 Auxiliary Storage Management Overview

The purpose of the Auxiliary Storage Manager is to move
virtual memory pages between real storage and auxiliary
storage.  The Auxiliary Storage Manager divides virtual
memory pages into three logically distinct groups:  PLPA,
common, and local.  At least one page data set is required
for each group of pages in a minimal ASM configuration,
though there are multiple local page data sets in most
systems.  In addition, swap data sets can be allocated to
decrease swap times.

The importance of allocating swap data sets varies, depending
on the level of the system and the nature of the
configuration and workload.  Except when you are using a
fixed head device, these data sets should be isolated on
separate volumes that are dedicated to ASM.  To every degree
possible, contention for the paths to these volumes should be
minimized.

ASM logic arranges its data sets in circular queues ordered
from those devices with the fastest service times (that is,
cache and solid state device (SSD)) to those with the slowest
(that is, fixed head then moveable head).  When selecting a
swap or page data set, the "fastest" queue is always scanned
first.  Other data sets, arrayed on the "slower" queues, are
scanned only when no fast queue device can handle the
request.

Within a queue, data sets are selected in round-robin
sequence.  In MVS/XA and MVS/ESA systems, an effort is made
to load balance the paging subsystem by scheduling page-out
requests to those devices that are projected to provide fast
response times and bypassing those that are projected to
provide slow response times.
ASM allocates available slots and builds paging channel
programs in such a way that head movement is minimized.  The
generated channel programs contain SEEK and SET SECTOR
commands as required.

When processing multiple paging requests, channel programs
are constructed to run for a targeted period of time, called
the service burst.  Avoiding long-running channel programs
allows ASM to spread the paging I/O workload, increase
parallelism, and minimize forcing an otherwise innocent
address space that is waiting for page fault resolution to
wait for a long channel program to complete.  The service
burst concept is discussed in more detail in Section
2.3.4.2.2.