• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

What is the correlation between IOPS and Throughput?

X

x-cimo

Limp Gawd
Joined
Mar 26, 2012
Messages
209
Basically, I am trying to understand why low IOPS (such as SATA 2TB drive) are able to get a fairly fast data transfer (150 MB/s +) with around 150 IOPS while fast SSD drive can only do maybe 2 or 3 time more (400 MB/s) with 20000 IOPS?

Why do Database / ZFS sync write, need so many IOPS if a drive can do 150MB/s while reading a large file?

Is the answer all in the cache? Because the drive still need to read the data to write it to cache...
 
I think they are bottlenecked by both their interfaces (SATA / SAS 6Gbps) and also by the NAND interfaces or raw throughput of the flash chips back to the SSD controller on the drive.
 
cause your comparing two totally different things.

For data transfer, they are measuring sequentual throughput, and disks are good at that, delivering 266k iops. But the iops of that really don't matter at all, since your only concerned with how fast you can read/write a large file.

When you access the drive randomly, you want to know how many random operations it can do, that is what normally the iops figure tells you, and for a 7200rpm disk, that is going to be around 80iops.
 
Basically, I am trying to understand why low IOPS (such as SATA 2TB drive) are able to get a fairly fast data transfer (150 MB/s +) with around 150 IOPS while fast SSD drive can only do maybe 2 or 3 time more (400 MB/s) with 20000 IOPS?
Click to expand...

Disks because of their mechanical nature are bad at iops because the heads have to move and the disk has to rotate for the drive to access the next random location. SSDs do not have this mechanical limitation thus jumping to a random spot is orders of magnitude faster. IOPS measures this speed of jumping to random spots on a drive. The STR (100+ MB/s numbers you posted) are the performance you get if you are sequentially loading a large file and not jumping around the drive. Hard Drives are not bad if their heads do not need to move or only move to the next track. However in the real world you generally do not load large files sequentially also generally have more than 1 application (or OS + application) competing for the drive so this is less useful.
 
Last edited:
No, 150MB/s at 4MB transfer size is only in the 10s of IOPS for a single drive. Even on a SATA III SSD, at 4MB you only get 79 iops at larger transfer sizes. Since the transfer size is so big, you don't need many IOPS to get to 500MB/s
 
patrickdk said:
cause your comparing two totally different things.

For data transfer, they are measuring sequentual throughput, and disks are good at that, delivering 266k iops. But the iops of that really don't matter at all, since your only concerned with how fast you can read/write a large file.

When you access the drive randomly, you want to know how many random operations it can do, that is what normally the iops figure tells you, and for a 7200rpm disk, that is going to be around 80iops.
Click to expand...

totally nailed it right here.

nice response. +1 rep.
 
You must log in or register to reply here.
Back
Top