15

I hear lots about SATA being slower than SAS but I've only found 1 decent benchmark. http://blog.whitesites.com/SAS-RAID-0-VS-SCSI-RAID-0-VS-SATA-Benchmarks__633689244191943184_blog.htm

In this, SAS comes out on top until you read that the SAS drives are 10K and the SATA drives are 7200 RPM.

Has anyone seen any good benchmarks/comparisons with equivalent speed disks? Has anyone seen any real-world loads (DB loads/searches, media streaming, etc?

  • 2
    Ive talked to a HP guy about his opinion about this. He reasoned that SATA are only 7200 rpm, too. Well - he was wrong. I easily found 10K drives as well. So lets look if the bounty will provide hard evidence. I have my doubts... – Nils Sep 14 '11 at 21:24
  • SAS comes into play more noticably when you have multiple simultaneous access to a drive. Also it is easy to compare the two if you use nearline SAS and compare with the same base SATA drive. I don't have the benchmarks to hand anymore but in our tests they perform the same under some loads with SAS upto 20% faster under our simulated typical SAN usage. – JamesRyan Sep 16 '11 at 9:18
  • @JamesRyan: Have you a explanation for this? "Typical" is random access with how many simulated clients? – Nils Sep 16 '11 at 20:54
  • And don't forget, if speed is what you're after, an SSD trumps anything else regardless of interface. – aroth May 15 '13 at 6:05
10

The "SATA = 7.2K RPM, SAS = 10/15K RPM" mind-set is strong, and (in my opinion anyway) where most of the "SAS is faster than SATA" thinking comes from. There are some slight differences between SAS and SATA drives, notably in their on-board caching algorithms (NCQ vs. TCQ). However, the performance difference of equivalently specced hard-drives will be fractional percentage points in most use-cases.

10
+50

IOPs is the difference your looking for in the "speed".

The simple way to explain the difference is that SATA is half duplex and SAS is full duplex. SATA drives are dumb and have to communicate with the controller for operations. SAS drives are smart and only requests and returns use the bus.

Depending on your usage case, spending more may not gain much..

  • So you are saying the SAS controllers essentially perform like the old SCSI controllers? I found when testing those, the thru-put was nearly the same (within 10-20%) but the CPU usage for SCSI was under 5% and for ATA was over 90%. Of course that could have been the controller in use at the time, but still showed me how SCSI beat the pants off ATA. (This was back when the CPUs where single core and less than 1GHz.) – boatcoder Sep 15 '11 at 17:50
  • I think this is the first answer hitting the mark. So the benchmark has to include multiple short transfers - then you should see the difference - IMHO. – Nils Sep 16 '11 at 20:59
  • @Mark0978: Yes, but instead of using a parallel bus where all the drives shared the same bandwidth, each drive is serially attached.. – wrmine Sep 18 '11 at 5:53
3

This is only a single anecdote, but I did some performance comparisons a while back for sequential read/write speeds on Seagate Barracuda ES.2 SATA and SAS disks, and found the SAS disks were significantly (low-double-digit percentages) worse than the SATA disks. These disks were not only the same speed rotational (7200rpm), but the same brand and model, just with a SAS interface instead of SATA (although I believe the SAS drives had smaller cache as well - 16MB vs 32MB for the SATA disks)

I didn't test random-access however, as I didn't care.

Just point anecdote, for a specific use case. YMMV :)

  • Useful, since we are looking at Seagate Barracuda drives in this case... – boatcoder Aug 5 '11 at 2:59
  • But what about the controller? That's likely the limiting factor. – anastrophe Aug 5 '11 at 4:46
  • 1
    @anastrophe: In the tests I ran, we were using LSI8888 SAS controllers, with either Barracuda ES.2 SATA or Barracuda ES.2 SAS disks, with identical RAID and filesystem settings - the only variable was the SATA/SAS model of disk (and therefore the on-disk cache size as well, however that is moot for sequential disk access). We tested both RAID5 and RAID0, and while with RAID5 the controller is a bottleneck, with RAID0 it is not. Finally, the SATA disks performed better in both cases, even in the RAID5 case. As I said though, YMMV :) – Daniel Lawson Aug 7 '11 at 21:04
2

So much misinformation here!

I can't imagine a meaningful benchmark between SAS and SATA and I wouldn't bother looking for one. This is like benchmarking a six passenger minivan versus high-speed bullet train.

1) At any given point in time, the SAS interface is always twice as fast as SATA -- and 4x faster if you consider that SAS is dual-ported and full duplex. Today, SAS is 12gbits, dual-ported and full duplex while SATA is 6gbits, single-ported and single duplex. There has never been a time when SAS and SATA were both 6gbits.

2) The main reason for using a SAS vs SATA drive is when many disks or ssd devices are sharing a SAS domain. Now as many here already know, your AVAGO (LSI) or equivalent RAID adapter can talk to either SAS or SATA devices via 8x12gbit SAS ports. For the most demanding workloads, I have seen comparative testing in our labs showing a "fully loaded" 12gbit SAS RAID adapter connected via SAS expanders with 16x SAS SSDs and this is up to 8x faster than the exact same adapter with 16x SATA SSDs, and in this case we are not even leveraging the dual-ports on the SAS drives!

3) There are numerous features built into SAS drives that (in the presence of lots of bus activity) help them reduce contention and allow these drives to get on and off the bus much, MUCH faster than a SATA drive can on the same SAS bus with the same workload.

Bottom line is SAS vs. SATA is all about the use-case. The comparison of a single SATA drive vs a single SAS drive is a meaningless exercise.

  • "The comparison of a single SATA drive vs a single SAS drive is a meaningless exercise." Great statement. Its like comparing how a 1/2 ton truck tows 500lb vs a 1 ton truck over a 100 mile distance. The 1/2 ton could be better in all regards. Then try to tow 12,000 lb for 100k miles and the 1/2 ton shows its true colors. SAS was designed for enterprise loads under enterprise duration. – Damon Feb 13 '18 at 4:49
1

Let`s see it from the server perspective. SATA is the successor of ATA which is the successor of IDE. Built for standard PCs. SAS ist the sucessor of SCSI which has the better (i.e. more stable, more standardized) protocol in hard and software. SCSI were the expensive disks used in servers.

Also note that if you buy a Dell-Server (I take that brand as an example) with 5 years support the support will only be 2 years for a SATA drive but you will get the full 5 years for SAS. This is a good hint about reliability.

  • What I'm finding from the drive manufacturers is identical latency, 6Gb/s speeds, * SAS - 3year warranty, SATA - 5year, * SATA cost $, SAS pretty much 3 times more. The longer warranty on the SATA drives surprised me, the identical speed specifications further confused me, so I'm really wanting to see some hard and fast benchmarks, not anecdotal evidence. We descended from apes, but most of us don't swing in trees anymore..... – boatcoder Aug 4 '11 at 21:06
  • 2
    All of the information regarding reliability, maturity, etc. is true; however, none of it addresses the question at hand. Don't get me wrong, I run SAS drives in my db server, for all the reasons you cite. But those reasons aren't relevant here. – anastrophe Aug 4 '11 at 21:06
  • It's all about costs - mostly. But I think the main point is how commands are queued - so I guess that on DBs with heavy loads the difference should be measurable. Let`s see if someone comes up with real numbers. – Nils Aug 5 '11 at 19:42
  • @anathrophe if SAS takes off load from your CPUs during massive random access (like SCSI) this might be the point here. – Nils Sep 17 '11 at 20:43
-1

While some people tend to focus on interface speed, I'm thinking this is what the hardware makers want us to focus on. When talking about SSD’s I understand that interface speed matters. But are there actually any mechanical drives reaching its interface speed at any point, other maybe when something is read from its cache? For general file storage, especially handling large files I don’t think the interface speed will make any difference. The fastest consumer SATA harddrive I know about for the moment are 12TB models which can read/write about 250MB/s, which is just a tad above 2Gbit, way below the interface speed. Raid is the option to go faster, but interface speed is not the limitation for that to happen.

WD Gold drives has 5 years warranty. I see no reason to use SAS unless you aim for 10k rpm or more. By buying branded HP or Dell you are forced accept what warranty they offer. It's worth the risk (for HP/Dell) giving you a longer warranty if you pay more, if it holds up they made a lot of money. If it fails they have still made some money. SAS is no guarantee the drive will last any longer. I've seen plenty of failed SAS drives over the years.

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.