Optimizing OS Performance

The main causes of poor performance leading to unpredictable system response are due to access contention of memory and CPU resources.

You can use the real time features of the OS scheduler to manage system performance, but you should use this with caution as many processes in the HP OpenCall SS7 platform need easy access to processing resources (see “Platform and User Application Scheduling”).

A better way to ensure predictable and optimized performance is to avoid resource access contention by following these guidelines:

Always keep total CPU usage below 85% and keep spare CPU resources to minimize any CPU resource access contention.
Do not try to increase the priority of an application process with nice, rtprio or rtsched.
Make sure you have sufficient physical memory for all HP OpenCall SS7 and application processes. For a Front End running HP OpenCall SS7 processes only, with no application processes, 128 MB is the recommended minimum for the system. The recommended minimum per stack is also 128 MB. The system must have at least twice as much swap memory (also known as virtual memory) as physical memory.
Design and implement proactive overload handling mechanisms which will avoid the platform limiting the throughput due to it delaying the scheduling processes (that is, 100% CPU usage).
Minimize the number of processes involved in the performance path.
Beware of the effect of buffered I/Os. See the product Release Notes.
Avoid swapping end monitor swap usage.
Avoid dynamic memory allocation in application processes. Even when all the memory leaks are eliminated, the fragmentation effect can cause the process memory consumption to increase with time.
When using a FE/BE topology it is preferable to use high speed LANs such as FDDI or 100BASE-T.

Optimizing OS Performance

Technical documentation

» Table of Contents

» Index