At work today, during some down time, I was rotting my brain on youtube. I ran across a "test" pitting a 128 gig SSD against a pair of 1 TB 7200 RPM drives in raid 0. The SSD destroyed the 2 TB raid array. My question is since raid 0 is all about speed, how may spinning drives would it take to close the performance gap with an SSD. There is no way I'm going to spend that much money on spinning drives, but it is an interesting question anyway.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
strange RAID question
- Thread starter rdsii64
- Start date
- Sort by reaction score
In theory, you could add up the speeds of all hard drives in a RAID 0 array.
So lets say you have an SSD with 300 MB/s (2012 MBA) read/write speed and you have HDDs with 60MB/s. In this case you would need 5 of that HDD in a RAID 0 array.
Many SSD can achieve speeds over 500MB/s, and the new PCIe SSD should get near 1GB/s. So you would need quite a lot of hard drives to match the speed of an SSD.
There are SAS hard drives spinning at 15000 RPM used mainly in servers. These are faster than consumer HDD, but SSD still destroys it in terms of speed.
So lets say you have an SSD with 300 MB/s (2012 MBA) read/write speed and you have HDDs with 60MB/s. In this case you would need 5 of that HDD in a RAID 0 array.
Many SSD can achieve speeds over 500MB/s, and the new PCIe SSD should get near 1GB/s. So you would need quite a lot of hard drives to match the speed of an SSD.
There are SAS hard drives spinning at 15000 RPM used mainly in servers. These are faster than consumer HDD, but SSD still destroys it in terms of speed.
A 7200 RPM SAS drive does not have a sequential write/read speed any higher than a 7200 RPM SATA drive. SAS drives usually offer a higher MTBF (Mean Time Between Failures) than SATA drives. Only with 10K to 15K RPM SAS drives will you start gaining speed advantages over SATA.
Faster spinning SAS drives have lower seek times compared to lower RPM consumer SATA drives, and even 7200 RPM SAS drives can usually handle a lot higher queue depths than their SATA counterparts. Sequential transfer speeds aren't much higher with faster RPMs. You should see the difference with 4K (or smaller) random writes/reads and with high queue depths, this is where SAS has an advantage over SATA HDDs, and sometimes even over SATA SSDs.
Depending on the controller used however, RAID 0 seek times and latency might quickly get higher the more drives you add, so reaching for higher sequential speeds with RAID 0 might cause you to lose the advantage SAS would otherwise offer.
But like yliu said, sequential read speeds scale well in RAID 0, so doubling the amount of drives doubles the sequential speed (while possibly wrecking seek times). So throwing together eight drives with 100MB/s sequential write speed in RAID 0 should tive you close to 800MB/s, provided the controller can handle it.
Last edited:
While RAID0 might get you better throughput, it also doesn't address the seek times, where SSDs have a huge advantage.
Ok, your first post led me to believe otherwise:
Like said, if you are looking for high sequential read/write speeds it does not matter much if the disks are 7.2K or 15K RPM.
I see where my post was confusing. truth be told there were a few of these "tests" on youtube. One of the test actually did use 15k SAS drives. Some others used 7.2k drives. The end result was the same.
Sequentially, the blocks are passing under the heads twice as fast, while the head seeks are minimal. In my experience a 15k drive is faster, though not twice as fast as a 7.2k drive.
This is true in raid 1, raid 5 or raid 6 (I hate that term. I prefer raid 5 double parity). This is for both servers and big storage frames that my systems use in production daily.
I do not mind the term RAID 6, what bothers me more, is that what it (and some other levels) is called depends on the manufacturer...
Yes, 15K disks are faster than 7.2K disks, in sequential and random I/O, as well as seek times. What I meant, is that both are still slower compared to a fairly new SSD, even in two-disk RAID 0.
Sorry, I should have been more clear about that.