load value in 'top' command

[Jim Dyches]

What is the range of possible values for the LOAD value output by the top command? What would be considered a heavily loaded system? What exactly does LOAD represent?

[Prashun Gupta]

Load actually is a CPU utilization by a particular process. Just 4-5% utilization is Okay for medium class systems.
Prashun

[Philip Chan_1]

For the top command in HPUX, the three load average numbers are referencing to the average number of jobs in the run queue over the last 1, 5, and 15 minutes. So, the range of load average value will vary depending on how many active jobs are there in your system.

Hope this answer your question.

Rgds,
Philip

[Kherwin Cheng Chua]

No one can really say what is a "heavy load" for another person's machine without knowing the specifications, behavior and use of that machine. To know what is "heavy load" for you, try baselining. Get the values of your system's load during its quiet state (no one logged in and only the most important processes running) and use that as your basis. What may seem like a heavy load for one system may just be normal load to another.

[Philip Chan_1 · Accepted solution]

Hi Jim,

For any load averages that is equal to or greater than 1, that means your system is under heavy load. Load averages of 2, 3, or 5 don't really matter if you only want to know whether your system is busy or not. Because that just mean how many jobs were splitting the cpu time, the cpu was fully utilized anyway.

You can imaging how busy your system was for the past 15 minutes by looking at the "load averages" line from the top command.

Load averages: x, y, z

If x is equal to or greater than 1, that mean there was x number of jobs in the run queue for the past minute. That also mean your cpu was utilized because there was at least 1 job in the run queue for the whole period.

If y is equal to or larger than 1, that mean your system was very busy and being tied up by y number of jobs for the past 5 minutes.

If z is >= 1, you have an extremely busy system, because the system was under load (prbably full load) for the past 15 minutes.

Rgds,
Philip

[Paula J Frazer-Campbell]

Hi Jim

As many have said load as displayed by the TOP command is dependant upon your servers use.

One of my main business servers a 4 way N class runs around 3 for the main part of the working day -

I tend to allow a load of 1 per processor so 4x1 = 4 and if my system goes over 4 I consider it busy in that response to my users is degraded slightly.

Whereas another of the main servers running a different database again a 4 way N class is I consider busy if it reaches 0.75 as this is above normal for it.

Each server has approx 800 concurrent users from a users base of approx 14000 so as previously said, "it depends".

What is more important that the "Load" figure is response to the users and therefore the viability of the server to carry out its prime role. This in my case is selling holidays.

HTH

Paula

[Jim Dyches]

Thanks Philip and Paula, your last answers was what I was looking for.

Jim

[David Hixson]

LOAD numbers can differ from how busy the system is under some conditions.

For example, I was working with a machine the other day with an idle around 80% and a LOAD of 400-something and rising slowly.

LOAD indicates the number of processes read to run at a given time. Some tools divide it by the number of processors, some don't... but having a system load of higher than your number of processes means things are getting worse.

Extremely high load values indicate that you probably have a deadlocked condition of a critical resource, causing lots of processed to be 'ready to run' but sleeping on some event that will never happen. In my case it was a veritas get_inode bug... 400 some processes all trying to access a file by name and hanging. The great thing is that virtualy every tool you will run to analysis the problem hangs and just adds to the traffic jam. q4 is your friend.

Just wanted to make people aware than when the numbers get WAY out of line, this is something to check for.

[Suhas_2]

Jim,
My two cents...
In the top as you can see the load on the CPU's which is the average load on the system. I would like to draw your attention to the following output....
Cpu states:
CPU LOAD USER NICE SYS IDLE BLOCK SWAIT INTR SSYS
0 0.45 4.8% 0.0% 2.9% 92.3% 0.0% 0.0% 0.0% 0.0%
1 0.04 1.9% 0.0% 2.9% 95.2% 0.0% 0.0% 0.0% 0.0%
--- ---- ----- ----- ----- ----- ----- ----- ----- -----
avg 0.25 3.8% 0.0% 2.9% 93.3% 0.0% 0.0% 0.0% 0.0%

Here as a sysadmin, I would observe the values of
LOAD on each CPU (in a multiple CPU environment), CPU utilized by the USER processes, CPU utilized by the system. There should not be an erratic utilization on either part.
Secondly the load on each CPU should be nearly same. In some environments, some programs can adhere to the same CPU. In that scenario we would find that one CPU is being utilized 100% and the other one is idle. Patches are available to avoid this.

Hope this helps.....
Suhas.