The following is an explanation of memory statistics in By Cluster and By Node formats:
This is equal to (Total Memory - Free Pages / Total Memory).
The average number of processes waiting for available memory. The value of Memory Queue is equivalent to the count of processes in the computable outswapped queue (COMO).
The number of processes in the system (including SWAPPER).
The number of processes resident in the balance set.
The total number of page faults per second during the example interval.
The percentage of Page Faults that required a read from disk. This is (Read I/Os)/(Page Faults) expressed as a percentage.
The percentage of Page Faults that were resolved from memory, that is, without reading from disk. This is equal to (100 - Hard Faults) %.
The number of page faults per second resolved as valid global pages (already in memory) during the example interval.
The number of page faults incurred in system space per second during the example interval.
The number of process inswaps performed during the last example interval.
The following CPU statistics are reported in By Cluster and By Node formats:
Percentage of time the CPU time spent in interrupt stack, busy wait, kernel, executive, supervisor, user, and compatibility modes. This is the sum of System CPU % and Task CPU %.
The sum of the interrupt, kernel, executive and busy wait CPU percentages.
The sum of the supervisor, user, and compatibility mode busy percentages.
The average number of processes waiting for the CPU. There is a queue if this number is greater than one (1) (the process that would have run in absence of the collection process). Equivalent to the sum of processes in the computable queue (COM).
The percentage of time that the CPU was idle.
The percentage of time that the CPU and all disk devices (selected for data collection) were idle.
The percentage of time (non-overlapped CPU time) that the CPU was busy and no disk device (selected for data collection) was busy.
The percentage of time (non-overlapped I/O time) that the CPU was idle and at least one disk device (selected for data collection)was busy.
The percentage of time that two (2) or more disk devices (selected for data collection) were busy.
The percentage of time (overlapped CPU and I/O time) that both the CPU and at least one disk device (selected for data collection) were busy.
The following I/O statistics are reported in By Cluster and By Node formats.
The number of direct I/O operations performed per second during the example interval, exclusive of page and swap I/O.
The number of buffered I/O operations performed per second during the example interval.
The following lock statistics are reported in By Cluster and By Node formats:
This metric is the Host-Originated Locking Activity per second. This is the amount of locking activity generated by the host node. It is equal to the sum of local and outgoing enqueue operations, plus local and outgoing converted enqueue operations, plus local and outgoing dequeue operations.
This is the percentage of the host originated locking activity (above) which had to be serviced by other nodes in the cluster. It is equal to the sum of the outgoing enqueue, outgoing converted enqueue, and outgoing dequeue operations, divided by the total host originated lock activity.
This is the percentage of enqueue and converted enqueue operations that were forced to wait. It is equal to the enqueue wait rate, divided by the sum of local and outgoing enqueue and local and outgoing converted enqueue, operations.
This metric is the Directory Operations Incoming per second. This is the number of lock directory operations per second being requested of the host node.
This is the number of locking operations performed on the host node on behalf of other nodes in the cluster. It is equal to the sum of incoming enqueue, incoming converted enqueue, and incoming dequeue operations.
|
Copyright © 2008 CA.
All rights reserved.
|
|