OSI/AS Configuration and Management Manual

Performance Considerations
OSI/AS Configuration and Management Manual—424119-001
6-2
Factors That Affect Performance
Factors That Affect Performance
There are many factors in an OSI/AS subsystem that affect performance. They include
configuration attributes, hardware, software, NonStop Kernel operating system version,
applications design, traffic mix and volume of messages, line speeds, and OSI layers
used. You must consider not only these variables, but how these variables affect the
interrelationships among the components of the subsystem.
What to Measure
System performance is usually measured in terms of reliability, throughput, and delay.
When configuring or adjusting the configuration of a subsystem, you must consider the
tradeoffs between the functionality provided by the system, performance constraints,
and reliability. The primary concern of a user sitting at a terminal is typically good
response time or job turnaround time. The primary concern of the person managing the
system is usually reliable and satisfactory throughput.
The OSI/AS subsystem cannot be treated in isolation; the system environment must be
properly configured to provide for the needs of all the users of the network. You must
consider the importance of the OSI/AS subsystem in relation to other applications
operating on the network. Because programs share network hardware and software
resources, the whole system and all the demands on its resources must be taken into
consideration.
This section addresses those with sufficient knowledge of the OSI/AS subsystem, the
NonStop Kernel operating system, and performance analysis concepts to identify
potential sources of problems and areas that need further investigation.
Sources of Performance Problems
Since subsystem resources are limited by the capacity of the various hardware and
software components, only a limited number of users or programs can make use of any
given resource at a given time. The actual users of these resources are the system itself
(control information) and the applications running on the system. For example, only one
application at a time can run during each a CPU cycle, and the order in which that
application requires the necessary resources is usually rigid. If a resource is not
available, the application must wait until the resource has been made available again.
In the Compaq environment, this situation is improved. There are multiple CPUs in a
Compaq system, and work can be spread among the CPUs to achieve some simultaneity
of processing. Even with multiple CPUs, however, some CPUs may be overloaded,
while others are not. This results in the same resource availability problem that occurs
on single-CPU systems. To aggravate the problem, other resources, such as disks or
communications lines, may be overutilized.
This competition for resources leads to long queues for the use of a particular resource.
Congestion is typically due to the overutilization of such resources as memory, CPU
processing, disk processing, and communications line capacity. To achieve a substantial
improvement in your system performance, you may need to distribute the workload by
moving part of it to another CPU, disk, or communications line, or by scheduling parts
of the workload to run at different times of the day.