We have a server running Tru64 Unix, which is our main production server for a single application our organisation uses. The software vendor has complete control of the hardware and software (we still administer the software, but have no root access).
however the vendor has allowed us to run vmstat, which will produce output every 15 seconds for 10 intervals then exit.
I was going to setup an automated process that would run vmstat, logging the output. I thought this would be quite useful information, especially considering we have had performance issues lately.
Mangement have told me that I cannot do this as vmstat chews up a lot of resources and will slow the system down if it is constantly running.
Can anyone tell me if this is actually true?
Answer
You can continually run vmstat without fear of chewing up your resources.
vmstat outputs all it's performance statistics in text form that is printed to standard output, nothing more. The overhead is incredibly small. As a test I ran vmstat on two different servers and in both cases it required approximately:
- 456k to 485k usage
Additional superfluous information
On both servers I ran it at 1 second intervals for 50 intervals and it averaged
- 485k
- about 0.03% of the overall system CPU during that time period
I then ran it at 1 second intervals for 500 intervals and it averaged (1GB Ram - Intel(R) Xeon(TM) CPU 3.00GHz)
- 485k
- 0.38% of the overall system CPU during that time period
And I ran it at 1 second intervals for 500 intervals and it averaged (12GB Ram - Quad Core Intel(R) Xeon(R) CPU 5130 @ 2.00GHz)
- 485k
- 0.26% of the overall system CPU during that time period
Note: One server was a high performance server, the other an email server. Both functioned with barely a thought to vmstat running on the terminal. It'll take your server more resources to find out how much load vmstat creates than actually running vmstat.
No comments:
Post a Comment