Hardware is ready, What tests shall we run?

dvl@

dvl@

Developer
I think we have all the hardware assembled, and I've run a few tests, but now it's time to selected a group of tests to run on each set of HDD and see how they compare.

I've added a blog post which outlines the hardware so you know what we have to test. In short, we have a 4x2TB Seagate, raidz1 array, 3 individual 3TB HDD (Toshiba, WD, and Seagate), one 2TB Segate, one SSD, and the ability to create an 8x2TB raidz2 array of Seagates.

I can run a simple bonnie++ test.

But I think what I'm looking for is some fio tests to run. On a simple search, I didn't find any 'standard' set of tests.

Got ideas?
 
A test that might or might not have interesting results: use dump(8)/restore(8) to duplicate the FreeBSD system on each of the individual drives. Boot from each and do a buildworld.

The times might not be that interesting because the hard drives are all probably similar in performance. The SSD will win, but by how much? That could illustrate how much a buildworld depends on I/O rather than CPU. Doing the same buildworld from a root-on-ZFS setup could be interesting also. Even if the results are not surprising, they would at least verify the common assumptions.
 
Sounds feasible. Something along the line of what the handbook says?

e.g.

Code: 
cd /mnt/root && dump -0b 512 -f - / | buffer -S 2048K -p 75 | restore -rb 512 -f -

Oddly enough, I found that reference on an obscure website. ;)

Repeat for each mount point.
 
When searching for ways to speed up dump(8), I found reports that -b64 was roughly twice as fast as -b32. However, the next logical step, using even larger values, resulted in failures. Unfortunately, I don't think I saved a pointer, but I'm sure they were on one of the FreeBSD mailing lists.
 
Hmmm, do you have an easy way to change booting remotely? boot from da0 this time, da1 next time, etc.
 
Sounds like a way to brick my machine when I'm away.... Perhaps I'll wait until next week.
 
Idea: put the OS onto an SSD, use that as the source. This would eliminate the source for the copy as a possible bottleneck.
 
I have copied the base os to both the SSD (mounted at /sdd) and to a 5-disk raidz2 array (mounted at /mnt). From there I'll do two make buildworlds. The first, from /sdd/usr/src, the second from /mnt/usr/src.

I'm hoping that'll be enough of a test without necessitating booting from the respective drives.
 
Rebooting can be done. That's not a problem. But it'll be from the same gmirror each time.
 
Are you looking for an easy way to change boot device temporarily?
If so,
Code: 
gpart set -a bootonce -i <n> <device>
might help.
 
tingo said:
Are you looking for an easy way to change boot device temporarily?
If so,
Code: 
gpart set -a bootonce -i <n> <device>
might help.
Click to expand...

That is so sexy!
 
I did post it once here, not sure which thread it was. It's worth mentioning these ZFS best practises.

Especially:

Code: 
RAIDZ Configuration Requirements and Recommendations

A RAIDZ configuration with N disks of size X with P parity disks can hold approximately (N-P)*X bytes and can withstand P device(s) failing before data integrity is compromised.

    Start a single-parity RAIDZ (raidz) configuration at 3 disks (2+1)
    Start a double-parity RAIDZ (raidz2) configuration at 6 disks (4+2)
    Start a triple-parity RAIDZ (raidz3) configuration at 9 disks (6+3)
    (N+P) with P = 1 (raidz), 2 (raidz2), or 3 (raidz3) and N equals 2, 4, or 6

Recently I too built a new server and wanted to post some tests to compare. I've 6 2TB REDs in raidz2 with one Intel's 520 SSD as cache device, all hooked up on LSI SAS 9211 on S1200bts board. One pool is partially encrypted, HW AES supported.

But I'm not that familiar with any "standard" test either; I can share my bonnie++ tests to compare though.
 
I can do an 8 x2TB disk raidz1 or raidz1. What I'm testing with now is constricted by having the 3 x 3TB HDD in the server.

At this point, it seems clear which HDD to buy, based on the tests so far.

Or would any of you do more tests before making a purchasing decision?
 
I am now of the opinion that I don't need any more tests to find which HDD of the three 3TB disks I've been looking at.

Anyone think I'm wrong? I think the Toshiba is the fastest.
 
dvl@ said:
I am now of the opinion that I don't need any more tests to find which HDD of the three 3TB disks I've been looking at.

Anyone think I'm wrong? I think the Toshiba is the fastest.
Click to expand...

Did you post the results anywhere?
 
dvl@ said:
I can do an 8 x2TB disk raidz1 or raidz1.
Click to expand...
But neither of that would be in line with recommendation pasted above. If you can't do more disks, maybe you can go with 6x3TB.
Once you have those disks you can check and compare performance difference with raidz1 on 3x and 4x setup (and raidz2 with 6x and 8x).
 
matoatlantis said:
But neither of that would be in line with recommendation pasted above. If you can't do more disks, maybe you can go with 6x3TB.
Once you have those disks you can check and compare performance difference with raidz1 on 3x and 4x setup (and raidz2 with 6x and 8x).
Click to expand...

The recommendation pasted above in message #17 doesn't indicate to me that an 8 disk raidz2 is not recommended.

Can you elaborate please as to how it goes against recommendation?
 
You must log in or register to reply here.
Back
Top