Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

How to measure random read/write IOPs?

S

sevenhells

macrumors newbie
Original poster
Dec 13, 2015
17
0
Hello,

I'm using BlackMagicDisk, but unlike Samsung Magician it's not telling me my IOPs, only the sequential read/write and other benches.

Reason being is I'm suping up my 2010 MBP with 1tb ssds for traveling and getting ideas down, I can finish them on my power PC. I find that SATA 2 ports on various systems bottleneck certain SATA 3 drives differently (between 10-40k), and I'm really curious what the exacts are on this motherboard.

Any other software that can read the IOPs?
 
Last edited:
I need to know asap as I'm trying to buy some Crucial M4s and they are going out of stock
 
Thanks, I get lost after typing "mount" in terminal, these directions are not very clear. The name of my main OS drive is "OSX". What does I type after mount?
 
duervo said:
Post the output of the mount command.
Click to expand...


Anthonys-MacBook-Pro:~ Casalena$ mount

/dev/disk0s2 on / (hfs, local, journaled)

devfs on /dev (devfs, local, nobrowse)

map -hosts on /net (autofs, nosuid, automounted, nobrowse)

map auto_home on /home (autofs, automounted, nobrowse)

Anthonys-MacBook-Pro:~ Casalena$
[doublepost=1472241017][/doublepost]After putting mount, I typed iostat -d


Anthonys-MacBook-Pro:~ Casalena$ mount

/dev/disk0s2 on / (hfs, local, journaled)

devfs on /dev (devfs, local, nobrowse)

map -hosts on /net (autofs, nosuid, automounted, nobrowse)

map auto_home on /home (autofs, automounted, nobrowse)

Anthonys-MacBook-Pro:~ Casalena$ iostat -d

disk0

KB/t tps MB/s

662.48 10 6.78

Anthonys-MacBook-Pro:~ Casalena$
 
Ok. Type this, let it run for about 30 seconds, and then post the output:

iostat –d disk0 1

I almost forgot. Run your BlackMagic benchmark thingy while the iostat command is running.

 
Last edited:
Anthonys-MacBook-Pro:~ Casalena$ mount

/dev/disk0s2 on / (hfs, local, journaled)

devfs on /dev (devfs, local, nobrowse)

map -hosts on /net (autofs, nosuid, automounted, nobrowse)

map auto_home on /home (autofs, automounted, nobrowse)

Anthonys-MacBook-Pro:~ Casalena$ iostat -d disk0 1

disk0

KB/t tps MB/s

656.19 10 6.70

4.00 3 0.01

0.00 0 0.00

26.87 78 2.04

21.53 692 14.55

18.36 213 3.82

39.46 613 23.61

854.78 180 150.16

899.39 202 177.85

656.14 278 177.85

827.53 221 178.95

672.41 274 179.65

469.65 380 174.38

891.94 207 180.02

313.05 280 85.62

783.91 334 256.00

893.33 305 266.49

899.18 304 266.63

900.82 300 264.25

896.54 304 265.82

disk0

KB/t tps MB/s

895.00 204 178.11

895.76 153 133.74

901.77 139 122.73

879.83 69 59.20

908.73 71 62.88

886.63 67 57.96

900.00 72 63.21

801.69 104 81.29

901.28 290 255.67

899.18 303 266.15

879.19 301 258.18

868.30 40 33.76

893.45 98 85.76

900.00 60 52.63

902.03 122 107.19

893.30 74 64.28

891.36 75 64.96

890.76 97 84.27

851.61 131 109.35

901.23 302 266.12

disk0

KB/t tps MB/s

895.70 301 263.65

904.34 142 125.69

890.25 130 112.73

893.03 93 80.78

393.55 148 56.73

896.46 70 61.10

891.49 47 40.89

894.14 173 151.36

892.36 201 175.33

897.36 234 205.48

901.23 301 265.28

890.44 306 266.40

701.97 182 125.10

896.98 82 71.64

900.00 167 146.88

900.62 198 174.30

446.97 266 116.02

898.39 77 67.17

900.69 180 158.63

889.52 202 175.58

disk0

KB/t tps MB/s

868.01 295 250.27

899.18 304 266.66

899.18 303 266.47

901.63 303 266.40

898.25 141 123.95

901.20 205 180.82

896.12 128 111.87

901.85 201 176.72

899.39 202 177.73

900.86 144 126.89

903.35 74 64.90

899.41 208 182.75

899.59 304 267.24

871.90 298 254.05

617.71 215 129.64

896.25 304 266.03

901.33 278 244.32

897.59 205 179.78

901.85 200 176.39

901.24 199 175.31

disk0

KB/t tps MB/s

893.07 205 179.11

902.46 202 177.81

892.00 143 124.23

902.14 116 101.94

900.00 222 194.77

740.94 354 255.85

901.63 304 267.42

898.36 302 264.63

901.23 302 265.79

900.00 276 242.43

892.60 121 105.11

898.77 201 176.83

900.60 204 179.77

900.60 204 179.43

899.39 203 178.10

821.38 191 153.20

898.20 207 181.29

901.08 230 202.18

899.18 304 266.66

901.23 303 266.44

disk0

KB/t tps MB/s

899.18 304 266.66

892.43 306 266.54

897.29 228 199.89

831.16 104 84.04

604.18 166 98.17

900.00 203 178.81

900.62 198 174.07

900.62 198 174.38

896.98 205 179.26

900.62 199 174.71

900.00 227 199.72

901.63 303 266.80

899.18 303 266.25

898.77 303 265.79

898.28 303 265.41

900.00 227 199.63

897.90 176 154.37

902.46 201 177.23
 
Ok, all the values in the TPS column are your IOPS. The ones in the 200-300 range are most likely the read portion of the BlackMagic stressing. The lower TPS values are most likely during the write portion of the BlackMagic stressing.

To get an overall average, just average out the values of the TPS column. That will get you a combined average of read/write IOPS. (I'd use a spreadsheet for this part, but that's up to you.)
 
Last edited:
duervo said:
Ok, all the values in the TPS column are your IOPS. The ones in the 200-300 range are most likely the read portion of the BlackMagic stressing. The lower TPS values are most likely during the write portion of the BlackMagic stressing.

To get an overall average, just average out the values of the TPS column. That will get you a combined average of read/write IOPS. (I'd use a spreadsheet for this part, but that's up to you.)
Click to expand...

Wow, is there no way to just get the numbers right in front of my face? Really, apple doesn't have any software capable of this easy task?

Are we talking about the same IOPs? Lets assume the average is 250, how is that related to this:

4KB Random Read
Up to 45,000 IOPS

4KB Random Write
Up to 50,000 IOPS
 
Also
sevenhells said:
Wow, is there no way to just get the numbers right in front of my face? Really, apple doesn't have any software capable of this easy task?

Are we talking about the same IOPs? Lets assume the average is 250, how is that related to this:

4KB Random Read
Up to 45,000 IOPS

4KB Random Write
Up to 50,000 IOPS
Click to expand...

Those values are under optimal conditions. You most likely will not see them in real-world circumstances.

Apple caters to the lowest common denominator when they design, build, market, and sell their products.

"We want everyone to use our products." - Tim Cook, 2016. (I might be paraphrasing there, but he was quoted saying as much in a recent interview reported on here within the last week or two.)​

The average consumer will not understand any of this stuff, and most likely wouldn't even care to know about them in most circumstances.

So, both of those things combined gets you the product delivered in a manner that Apple intended. That said, they still include the tools (CLI commands and Terminal) to measure those metrics for those that still have a need or desire to do so.
 
I get it, I think, but 200-300 iops seems drastically less than 45 thousand. Am I understanding correctly?

The optimal is 40k read IOPs for my crucials, in SATA 3 ports. When it's in SATA 2, I would assume I would get at least 5-20K read IOPS, not 200.
 
sevenhells said:
I get it, I think, but 200-300 iops seems drastically less than 45 thousand. Am I understanding correctly?

The optimal is 40k read IOPs for my crucials, in SATA 3 ports. When it's in SATA 2, I would assume I would get at least 5-20K read IOPS, not 200.
Click to expand...

You are at the mercy of the settings used by BlackMagic tests in your case. If you want to measure things under the same circumstances used for the tests that Crucial ran in which they produced those results, then you'll have to try to find what tool they used to benchmark their drives, and what settings they used. Then, try to reproduce that environment on your system.

Some vendors do things like create a RAM cache on the system and use that in concert with the SSD, set the max size of the write and read requests to be aligned to a drive's 4K formatting, etc to achieve those numbers. (ie: They create optimal conditions to get those numbers.)

Realistically, and to be honest, I wouldn't worry so much about it. Rest assured, you are getting better performance out of that Crucial drive than any HDD you had in there. FWIW, my mid2012 cMBP gets numbers very close to yours (Crucial m4 256GB) ... maybe a tad higher on the reads (100 more IOPS here and there) while running the BlackMagic tool in macOS Sierra (latest dev beta), but for the most part they are very similar. The discrepancies could be explained by the difference in CPU there between our systems, but I didn't bother monitoring CPU. As drive speeds and IOPS capacities increase, the bottleneck in a system will move around to different areas that eventually you could potentially see the drive waiting on CPU to serve the IO requests. This is one of the primary reasons that at this time, I recommend an upgrade to anyone that is running a system from before 2011.
 
Last edited:
Is your model the exact same as mine? I have the older M4's, I think they improved them before going to M5.

If it's the same, then you're getting close to the advertised specs with your 2012 mbp with sata 3 ports, that means the SATA 2 ports can't be that bad and should perform to almost their max SATA 2 potential in general. Is there a universal and relative number that Sata 2 ports bottleneck when it comes to Sata 3 SSD/Iops?

Maybe if I plug this into my main PC, in a sata 2 port, then see what the average is, I'll just assume this macbook pro is similar.

duervo said:
my mid2012 cMBP gets numbers very close to yours (Crucial m4 256GB) ... maybe a tad higher on the reads (100 more IOPS here and there) while running the BlackMagic tool in macOS Sierra (latest dev beta), but for the most part they are very similar. The discrepancies could be explained by the difference in CPU there between our systems, but I didn't bother monitoring CPU. As drive speeds and IOPS capacities increase, the bottleneck in a system will move around to different areas that eventually you could potentially see the drive waiting on CPU to serve the IO requests. This is one of the primary reasons that at this time, I recommend an upgrade to anyone that is running a system from before 2011.
Click to expand...
 
IOPS are a terrible measure of performance. I like Wikipedia write up for why this is so.

Frequently mischaracterized as a 'benchmark', IOPS numbers published by storage device manufacturers do not relate to real-world application performance.[1][2]

To meaningfully describe the performance characteristics of any storage device, it is necessary to specify a minimum of three metrics simultaneously, IOPS, response time, and (application) workload. Absent simultaneous specifications or response-time and workload, IOPS are essentially meaningless. In isolation, IOPS can be considered analogous to "revolutions per minute" of an automobile engine i.e. an engine capable of spinning at 10,000 RPMs with its transmission in neutral does not convey anything of value, however an engine capable of developing specified torque and horsepower at a given number of RPMs fully describes the capabilities of the engine.

I prefer transfer rate as a better measurement of relative performance since it involves moving data rather than how many ready-set-ready you can do.
 
Yes, it all depends on the workload. This is why my standard reply to customers when they ask me these types of questions is, "It depends." ... followed by much tortured groaning (in jest, of course) by them, before I go on to explaining the details.

In my line of work IOPS is king though, as it involves a lot of architecture for virtualized servers. In those environments, you size for IOPS first, as storage workload patterns for virtualization hosts are generally random in nature, where high IOPS will present much greater benefits. For consumer products? It's not something to really worry about, but it can be interesting fodder for technical conversations sometimes.

I'm not a fan of that Wikipedia entry, BTW. It targets the consumer industry, and does nothing to consider the enterprise/corporate virtualization market. One of the reasons I take Wikipedia articles with a huge grain of salt.
 
Last edited:
duervo said:
Yes, it all depends on the workload. This is why my standard reply to customers when they ask me these types of questions is, "It depends." ... followed by much tortured groaning (in jest, of course) by them, before I go on to explaining the details.

In my line of work IOPS is king though, as it involves a lot of architecture for virtualized servers. In those environments, you size for IOPS first, as storage workload patterns for virtualization hosts are generally random in nature, where high IOPS will present much greater benefits. For consumer products? It's not something to really worry about, but it can be interesting fodder for technical conversations sometimes.

I'm not a fan of that Wikipedia entry, BTW. It targets the consumer industry, and does nothing to consider the enterprise/corporate virtualization market. One of the reasons I take Wikipedia articles with a huge grain of salt.
Click to expand...

I have done some work in cloud servers (Azure and AWS) and on-prem stuff and understand the bias toward IOPS when you are taking about SANs and other storage options in those types of environments. But this thread is about SSDs in an laptop. Quite different. That is why I thought the relevancy of IOPS in this context should be discussed.

And I kind of like the Wikipedia's posting analogy, if for nothing else than the humor.
 
Well I got the crucial m4's and they work great as intended. Weird how prior to using crucial, the 840 and 850 evos ran terrible on my MBP, I thought dropping the laptop several times was finally catching up to it. I even tried enabling TRIM on those and I was getting terrible spin wheels and hangs on all OS's.

One last question, I may want to upgrade to 1tb in the future, nothing fancy just for reading and no writes, is it possible to find the IOP specs on this?

http://accessories.dell.com/sna/pro...tid}:82727171957:901mtv7630:c&ven2=:#Overview

or is this just a crap shoot since it's not officially supported?
 
sevenhells said:
Well I got the crucial m4's and they work great as intended. Weird how prior to using crucial, the 840 and 850 evos ran terrible on my MBP, I thought dropping the laptop several times was finally catching up to it. I even tried enabling TRIM on those and I was getting terrible spin wheels and hangs on all OS's.
Click to expand...

Strange you had issues with the Samsung 850. I have used them with great success on MacBooks. Usually saw over 500 MB/sec read and 300 MB/sec on BlackMagic. And maybe one bounce of the beachball on first launch.

But, my experience has been with on 2011 and 2012 models that have the faster Sata 3 interfaces. So perhaps that explains the difference.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back