Load (computing)


In UNIX computing, the system load is a measure of the amount of computational work that a computer system performs. The load average represents the average system load over a period of time. It conventionally appears in the form of three numbers which represent the system load during the last one-, five-, and fifteen-minute periods.

Load

The Unix load number refers to the number of processes using or waiting for CPU; i.e., the number of processes in the ready queue or run queue. An idle computer has a load number of 0. Each running process increments the load number by 1. Each process that terminates decrements it by 1. Most UNIX systems count only processes in the running or runnable states.
In addition to processes in "R" state, Linux also includes processes in uninterruptible sleep states, which can lead to markedly different results if many processes remain blocked in I/O due to a busy or stalled I/O system. This, for example, includes processes blocking due to an NFS server failure or too slow media. Such circumstances can result in an elevated load average, which does not reflect an actual increase in CPU use. The idea behind its inclusion is that while disk wait is not the same as CPU-wait, it still reflects how long a user needs to wait.
On modern UNIX systems, the treatment of threading with respect to load averages varies. Some systems treat threads as processes for the purposes of load average calculation: each thread waiting to run will add 1 to the load. However, other systems, especially systems implementing so-called M:N threading, use different strategies such as counting the process exactly once for the purpose of load, or counting only threads currently exposed by the user-thread scheduler to the kernel, which may depend on the level of concurrency set on the process. Linux appears to count each thread separately as adding 1 to the load.
There is no standard way to obtain the length of the run queue across different Unix-like systems, but a commonly-available way is through parsing the output of the ps command, specifically ps -ax -o stat, and counting the number of lines starting with "R". If desired, one can also add uninterruptible "disk" wait state, labelled "D" on Linux and FreeBSD, "U" on macOS. -M may be used to get per-thread information on Linux and macOS, but not FreeBSD, where the option is instead -H.
On Linux specifically, the procfs file /proc/stat contains two lines and, corresponding to scheduling entities in "R" and "D" states respectively. This can be used to read the current load instead of ps. As with before, the load reported includes the program currently reading the procfs file, so subtract one from the sum to obtain the true load.

Compared to CPU utilization

The comparative study of different load indices carried out by Ferrari et al. reported that CPU load information based upon the CPU queue length does much better in load balancing compared to CPU utilization. The reason CPU queue length did better is probably because when a host is heavily loaded, its CPU utilization is likely to be close to 100%, and it is unable to reflect the exact load level of the utilization. In contrast, CPU queue lengths can directly reflect the amount of load on a CPU. As an example, two systems, one with 3 and the other with 6 processes in the queue, are both very likely to have utilizations close to 100%, although they would obviously differ in terms of process wait-times.

Load average

All Unix and Unix-like systems generate a dimensionless metric of three "load average" numbers in the kernel. Users can easily query the current result from a Unix shell by running the uptime command:

$ uptime
14:34:03 up 10:43, 4 users, load average: 0.06, 0.11, 0.09

The w and top commands show the same three load average numbers, as do a range of graphical user interface utilities. The underlying interface is, a C function present on most UNIX systems since 4.3BSD-Reno of 1990. On Linux specifically, one can also read /proc/loadavg for this information. This file also provides instantaneous information on the number of processes in "R" state, the total number of processes, and the process ID of the most-recently-created process.
Systems calculate the load average as the exponentially damped/weighted moving average of the load number. The three values of load average refer to the past one, five, and fifteen minutes of system operation.
Mathematically speaking, all three values always average all the system load since the system started up. They all decay exponentially, but they decay at different speeds: they decay exponentially by e after 1, 5, and 15 minutes respectively. Hence, the 1-minute load average consists of 63% of the load from the last minute and 37% of the average load since start up, excluding the last minute. For the 5- and 15-minute load averages, the same 63%/37% ratio is computed over 5 minutes and 15 minutes, respectively. Therefore, it is not technically accurate that the 1-minute load average only includes the last 60 seconds of activity, as it includes 37% of the activity from the past, but it is correct to state that it includes mostly the last minute.

Interpretation

For single-CPU systems that are CPU bound, one can think of load average as a measure of system utilization during the respective time period. For systems with multiple CPUs, one must divide the load by the number of processors in order to get a comparable measure.
For example, one can interpret a load average of "1.73 0.60 7.98" on a single-CPU system as:
  • During the last minute, the system was overloaded by 73% on average.
  • During the last 5 minutes, the CPU was idling 40% of the time, on average.
  • During the last 15 minutes, the system was overloaded 698% on average.
This implies that this system could have handled all the work scheduled for the last minute if it were 1.73 times as fast.
In a system with four CPUs, a load average of 3.73 would indicate that there were, on average, 3.73 processes ready to run. Because this is fewer than 4, we know each one could be scheduled into a CPU, and that no overloading is present.

Reckoning CPU load

Linux as an example

On Linux systems, the load-average is not calculated on each clock tick, but driven by a variable value that is based on the frequency setting and tested on each clock tick. This setting defines the kernel clock tick rate in hertz, and it defaults to 100 for ticks. Kernel activities use this number of ticks to time themselves. Specifically, the function, which calculates the load average, nominally runs every ticks, i.e. a tiny bit over.

extern unsigned long avenrun; /* Load averages */
extern void get_avenrun;
  1. define FSHIFT 11 /* nr of bits of precision */
  2. define FIXED_1 /* 1.0 as fixed-point */
  3. define LOAD_FREQ /* 5 sec intervals */
  4. define EXP_1 1884 /* 1/exp as fixed-point */
  5. define EXP_5 2014 /* 1/exp */
  6. define EXP_15 2037 /* 1/exp */
/* a1 = a0 * e + a * */
static inline unsigned long
calc_load
extern unsigned long calc_load_n;
  1. define LOAD_INT
  2. define LOAD_FRAC LOAD_INT & ) * 100)

The avenrun array contains 1-minute, 5-minute and 15-minute average. The function provides the correct update of the load-average for the default update rate of. It is used like such in loadavg.c :

void calc_global_load

Note this handling of NO_HZ CPUs. NO_HZ is a mode meant to reduce the number of scheduling-clock interrupts on idle processors, which improves power efficiency and reduces clock jitter. This would cause processors to miss their update ticks, however. As a result, the counting of tasks is done using atomic operations. The function handles the calculation in case of needing to catch up to multiple ticks, making use of this function:

/* application of the geometric series:
* n 1 - x^
* S_n := \Sum x^i = -------------
* i=0 1 - x */
unsigned long
calc_load_n

Here raises exp to its nth power in fixed-point arithmetic.

Sampling and precision

The "sampled" calculation of load averages is a somewhat common behavior; FreeBSD, too, only refreshes the value every five seconds. The interval is usually taken to not be exact so that they do not collect processes that are scheduled to fire at a certain moment. This is the reason for "+1" in the Linux code from above; FreeBSD instead uses a pseudorandom offset added to the interval.
also mentions that the use of 11 fractional bits in the above fixed point calculation prevents reducing the interval much lower. For example, a two-second interval would result in the EXP values becoming 1981, 2034 and 2043, nearly saturating the precision available.
Ripke Klaus has shown in 2011 that the "+1" modification alone is not sufficient to avoid Moiré artifacts from regularly-scheduled processes. His experiments suggest 4.61 to be a better value: 0.61 is close to the golden ratio, which helps spread out the sample-point among fractional seconds. At the same time, 4.61 is close to, so the property of being an integer fraction of is maintained. Ripke's change is common among Android system kernels, although the exact expression used assumes an HZ of 100. would be more appropriate for varying values of HZ. The new values would be:

  1. define LOAD_FREQ /* 60/13 ~ 4.61 sec intervals */
  2. define EXP_1 1896 /* 1/exp = 1/exp as fixed-point */
  3. define EXP_5 2017 /* 1/exp = 1/exp */
  4. define EXP_15 2038 /* 1/exp = 1/exp */