Overview / Mem Info

The Overview / Mem Info screen presents information and stats about various aspects of CPU memory.

Virtual Memory is the amount of disk space available to extend physical memory when fully consumed. The Operating System encounters a shortage of free memory, it will page out or write to storage portions of memory then release it. When the Operating System detects a process encountered memory paged out, the process is transparently suspended, its memory returned (i.e. read) from storage, and re-activated. This situation is referred to as a page-fault. Typically, a NonStop System would not want to encounter frequent page faults as they result in waits on storage I/O. However, the alternative would be the Operating System expectantly terminating a process or denying a request for memory.

Virtual memory is configured via the utility NSKCOM. A "rule of thumb", the amount of virtual memory is at least 2.5 times physical memory. This is configured on a per CPU basis and generally places the swap file on a storage volume DP2 process also residing in the same CPU.

Physical Memory is the memory actually present in the CPU. For example, you have a 32 Gigabyte processor.

Locked Memory is memory that cannot be swapped to disk. Certain operations, such as I/O to/from the CPU, must be physically present for write or read operations. A certain amount of memory is always locked by the operating system at start up and the total mount varies as processes start, stop and I/O activity is performed.

Misc. Memory is a 'catch area' for other useful memory counters.

Unsponsored Memory is memory that is temporarily not owned by any process but must remain allocated.

Memory Pressure is a value from 0 to 7 where 0 is low. It is an indicator of the demand for memory. Systems with plenty of memory have a pressure of 0.

A page fault is the situation where a memory page is needed but not present. This occurs when a process is first started and continues until all needed memory is accessed at least once. Additionally, a page fault can occur when memory that was previously accessed is no longer present, perhaps due to another process needing memory.

Memory Queue is an indication of the number of processes waiting for memory. Newer systems do not seem to have a general memory queue. The servicing of page faults is now handled within the context of a user process (via system code in the background). On older machines, a memory queue was a second ready list because a process must first have memory accessible before it can wait to execute.