Settings>Advanced>Performance Test

[Rilo Ravestein]

Cyberjock: that's not goofy, silly. It's hungry!

 

[cyberjock]

Haha. Classic song!

And a classic movie!

 

[scurrier]

So did this button get removed or something? I just expanded my pool and wanted to test performance again. I guess I could do command line but I'm wondering where the button went. I'm running latest version.

 

[dlavigne]

From http://doc.freenas.org/9.3/freenas_intro.html#what-s-new-since-9-3-release:

The “Performance Test” button has been removed from System ‣ Advanced.

 

[JeremyOne]

Hello, the “Performance Test” button was a great way to get some quick and basic disk/raid/zfs bandwidth diagnostics. I am sad that is has been removed with little explanation.

I used this (easy and quick!) feature quite a few times while configuring my system disks.

 

[cyberjock]

I can give you an explanation. ;)

The performance test button really didn't do what people thought it did, and the numbers provided were really not useful unless you happened to have a workload that exactly matched the iozone parameters that FreeNAS used. Since virtually nobody even knew what those parameters were (since you'd only know if you looked at the FreeNAS code yourself) it was a useless test to perform.

Benchmarking ZFS is an art. Even I try to avoid doing it as much as possible because there's so many caveats that, if not understood and worked around, yield numbers that are completely meaningless. And I mean 'completely' meaningless.

 

[Donny Davis]

I can give you an explanation. ;)

The performance test button really didn't do what people thought it did, and the numbers provided were really not useful unless you happened to have a workload that exactly matched the iozone parameters that FreeNAS used. Since virtually nobody even knew what those parameters were (since you'd only know if you looked at the FreeNAS code yourself) it was a useless test to perform.

Benchmarking ZFS is an art. Even I try to avoid doing it as much as possible because there's so many caveats that, if not understood and worked around, yield numbers that are completely meaningless. And I mean 'completely' meaningless.

Can we sticky that. I have been looking for a good way to bench mark my system before I get started. Guess I will just put a workload on it and see how it does

 

[cyberjock]

Can we sticky that. I have been looking for a good way to bench mark my system before I get started. Guess I will just put a workload on it and see how it does

That is, quite literally, the best way to do things.

Simply determine what workload you need to deal with, then build the system to be optimal for that workload. Once you are in a position to do real-world tests, then you can see if the L2ARC is useful, whether you need more RAM, etc.