Hi Simon, I beleive your suspicions are correct - AFAIK (and I'm pretty certain) it is the load average as a %age of capacity over the time period. Load snapshots taken in the same way as performance monitor/task manager under Winnt. I was a bit confused as to the OID though - do you know who has the enterprise OID 2010 ? - sorry - I could look this up but hoped you would know anyway. Cheers Earl

-----Original Message----- From: main-admin@lists.alug.org.uk [mailto:main-admin@lists.alug.org.uk]On Behalf Of Simon Parkes Sent: Wednesday 28 November 2001 12:41 To: ALUG Subject: [Alug] SNMP Load averages...

Hi, I'm hoping someone can help lift the mist from my eyes... doing an snmpwalk of my PC on .1.3.6.1.4.1.2010.10.1.3 gives me 3 results (as expected), these are the load averages of 1, 5 and 15 mins.. Can someone tell me what might be a high value ? My 1 min load average is 3.66 most of the time.... 3.66 out of what? 100, is this a percentage? Any ideas anyone...

Simon

---

main@lists.alug.org.uk http://www.anglian.lug.org.uk/ http://lists.alug.org.uk/mailman/listinfo/main Unsubscribe? See message headers or the web site above!

[earl.brannigan@lindenhouse.co.uk] www.lindenhouse.co.uk

on Wed, Nov 28, 2001 at 12:40:54PM -0000, Simon Parkes wrote:

Hi, I'm hoping someone can help lift the mist from my eyes... doing an snmpwalk of my PC on .1.3.6.1.4.1.2010.10.1.3 gives me 3 results (as expected), these are the load averages of 1, 5 and 15 mins.. Can someone tell me what might be a high value ? My 1 min load average is 3.66 most of the time.... 3.66 out of what? 100, is this a percentage? Any ideas anyone...

The Design and Implementation of 4.4BSD defines the load average as follows:

load average A measure of the CPU load on the system. The load average in 4.4BSD is defined as an average of the number of processes ready to run or waiting for disk I/O to complete, as sampled over the previous 1-minute interval of system operation.

although page 94 of the same book says:

...where the load is a sampled average of the sum of the lengths of the run queue and of the short-term sleep queue over the previous 1-minute interval of system operation.

and a more recent NetBSD man page states:

The kvm_getloadavg() function returns the number of processes in the sys- tem run queue of the kernel indicated by kd, averaged over various peri- ods of time.

The run queue is a list of processes that are ready to be executed whenever there's some cpu time for them to run. That is, the process isn't waiting for any kind of I/O or other operation that could cause the process to sleep, waiting for the operation to complete.

Linux appears to compute the load average in a similar way in /usr/src/linux/fs/proc/array.c line 229, in the function "get_loadavg". I don't know if/how much this relates to your situation though. :)

3.66 strikes me as a fairly busy machine. Generally I find as the load average approaches the 10-20 range, it can get quite slow with interactive operation (eg, typing things in at the console, ssh, etc).

On Wed, 28 Nov 2001 12:40:54 Simon Parkes wrote:

Hi, I'm hoping someone can help lift the mist from my eyes... doing an snmpwalk of my PC on .1.3.6.1.4.1.2010.10.1.3 gives me 3 results (as expected), these are the load averages of 1, 5 and 15 mins.. Can someone tell me what might be a high value ? My 1 min load average is 3.66 most of the time.... 3.66 out of what? 100, is this a percentage? Any ideas anyone...

Simon

My understanding of these figures is that are not out of anything, i.e. the do not represent a percentage of system capacity because system capacity is not fixed - you can always give a system more work to do if you are content for all work to be done more slowly.

The figures are averages of the number of processes that are ready to run and waiting for a CPU time slot, i.e. could make immediate use of the CPU rather than waiting for I/O.

For comparison with figures like %age CPU utilisation, consider the case where you have a single process that can always use the CPU because it spends its time calcuating rather the doing I/O, but that all other processes in the system mostly sit waiting for I/O. In this case I would expect CPU utilisation of 100% and a load average of 1.

For lower load averages the CPU utilisation would, on average, be less than 100% with the implication that if you gave the system more work (more work for the CPU) you could get more throughput.

Once the load average goes above it means the CPU is not only permanently busy but there is competion for the CPU so those programs that make most use of the CPU are not longer getting it whenever they want. At this point, if you gave the system more work the overall throughput should stay the same - the system would do more things at once and each one more slowly.

As for what is a typical or a high value it rather depends on the application the machine is being put to.

Steve.