How is linux load average calculated

In case of hyperthreaded CPUs there is an interesting side effect: loading a cpu makes its hyperthreaded pairs slower. But this happens on a deeper layer what the normal task scheduling handles, although it can and should influence the process-moving decisions of the scheduler. Hyperthreaded CPUs aren't widely used today. On Windows, a different method is used for the load calculation: there load 1. It is an exponentially weighted time average with a time constant of 1, 5, 15 minutes.

It is more effective to calculate, and reacts better for recent changes. Sign up to join this community. The best answers are voted up and rise to the top. Stack Overflow for Teams — Collaborate and share knowledge with a private group. Create a free Team What is Teams? Learn more. How is the load average is calculated in this case? Ask Question. Asked 4 years, 1 month ago. Active 1 year, 5 months ago. Viewed 11k times. How can I show such numbers with CPU only?

Improve this question. Lolika padmanbhan Lolika padmanbhan 31 1 1 gold badge 1 1 silver badge 2 2 bronze badges. Is this only a single processor system? Are the loads increasing over time?

Also, a complete profile of the jobs currently running would be useful. Now, what about those three numbers? Which brings us to the question:. For the numbers we've talked about 1.

Frankly, if your box spikes above 1. It's when the minute average goes north of 1. So of cores is important to interpreting load averages Note: not available on OSX, Google for alternatives. Adding servers can be a band-aid for slow code. Scout APM helps you find and fix your inefficient and costly code. Ready to optimize your site? Sign up for a free trial. All Engineering Performance Community. Andre Lewis on July 28, Most popular New Relic vs. Subscribe for the latest APM news and tech tips.

It simply means that a bit more than three processes were ready to execute and CPU cores were present to handle their execution. To translate this to a percent-based system load, you could use this simple, if not obtuse, command:. This command sequences isolates the 1-minute average via cut and echos it, divided by the number of CPU cores, through bc, a command-line calculator, to derive the percentage. This value is by no means scientific but does provide a rough approximation of CPU load in percent.

The system administrator must keep in mind that:. Because of this, getting a handle of CPU load on a Linux system is not entirely an empirical matter. Even if it were, CPU load alone is not an adequate measurement of overall system resource utilization. Facebook's Josef Bacik first did this with his kernelscope tool, which also uses bcc and flame graphs. In my examples, I'm just showing the kernel stacks, but offcputime.

As for the flame graph above: it shows that only ms out of 60 seconds were spent in uninterruptible sleep. That's only adding 0. Here is a more interesting one, this time only spanning 10 seconds SVG :. I've highlighted those functions in magenta using the flame graph search feature. The interruptible versions allow the task to be interrupted by a signal, and then wake up to process it before the lock is acquired.

Time in uninterruptible lock sleeps usually don't add much to the load averages, but in this case they are adding 0. Does it make sense for these code paths to be included in the load average? Yes, I'd say so. Those threads are in the middle of doing work, and happen to block on a lock. They are demand on the system, albeit for software resources rather than hardware resources. Can the Linux load average value be fully decomposed into components? Here's an example: on an idle 8 CPU system, I launched tar to archive some uncached files.

It spends several minutes mostly blocked on disk reads. Here are the stats, collected from three different terminal windows:. That adds up to 1. I'm still missing 0. Prior to 1. How much? That's pretty close to the 1.

This is a system where one thread tar plus a little more some time in kernel worker threads are doing work, and Linux reports the load average as 1.

If it was measuring "CPU load averages", the system would have reported 0. Likewise, which load averages? CPU load averages? System load averages? Clarifying it this way lets me make sense of it like this:. Perhaps one day we'll add additional load averages to Linux, and let the user choose what they want to use: a separate "CPU load averages", "disk load averages", "network load averages", etc.

Or just use different metrics altogether. Some people have found values that seem to work for their systems and workloads: they know that when load goes over X, application latency is high and customers start complaining. But there aren't really rules for this. It's somewhat ambiguous, as it's a long-term average at least one minute which can hide variation. One system with a ratio of 1.

I once administered a two-CPU email server that during the day ran with a CPU load average of between 11 and 16 a ratio of between 5. Latency was acceptable and no one complained. As for Linux's system load averages: these are even more ambiguous as they cover different resource types, so you can't just divide by the CPU count. It's more useful for relative comparisons: if you know the system runs fine at a load of 20, and it's now at 40, then it's time to dig in with other metrics to see what's going on.

When Linux load averages increase, you know you have higher demand for resources CPUs, disks, and some locks , but you aren't sure which. You can use other metrics for clarification. For example, for CPUs:. The first two are utilization metrics, the last three are saturation metrics. Utilization metrics are useful for workload characterization, and saturation metrics useful for identifying a performance problem.