Almost 20 TB (Or $50,000) Of SSD DC S3700 Drives, Benchmarked

[mras]

Twenty-four total bays are enabled by a trio of eight-drive bays grafted to our server's exterior. Each possesses two SAS 8087 ports (one for every four drives), and they're just not able to get data through unmolested. Whether any backplane would work in this situation is uncertain, and bypassing ours simply wasn't an option for today's experiment

So, 3 x of backplane that have 2x8087 connectors each.
Each 8087 should do 4x600MB/s (x2 connectors)=> 4800MB/s times 3 this should give 14400MB/s
Where is the bottleneck?
Are you saying that the backplanes got chips onboard that limits the transfers?

 

[warmon6]

I'm having a flash back to this video.....


http://www.youtube.com/watch?v=96dWOEa4Djs

 

[cryan]

[citation][nom]yialanliu[/nom]Very cool to see the performance but I would love to see a test of RAID 5/6 as a much more practical usage of multiple SSDs[/citation]

Due to the tight time constraints, putting another sixty hours of testing in on RAID5/6 wouldn't have been practical. I did lots of preliminary RAID5 testing with some smaller arrays and the results were promising. So promising, they deserve their own article.



Regards,
Christopher Ryan

 

[jeffunit]

"As we know, the SSD DC S3700 ships in capacities as low as 10 GB. Two dozen of those smaller drives could do some real damage in the right hands. After all, you'd be looking at 2.4 TB in RAID 0."

I suspect there is a typo here. 10gb*24= 240gb, not 2.4TB.
Perhaps what was meant was 100 GB?
Perhaps someone needs to proof read articles before posting?

 

[merikafyeah]

The numbers aren't very impressive considering it takes such a huge and expensive array just to match the throughput and speed of basic RAM drives, which easily reach 765,000 IOPS (13GB/s) on modest consumer hardware:
http://thessdreview.com/our-reviews/romex-fancycache-review-ssd-performance-at-13gbs-and-765000-iops-in-60-seconds-flat/

16TB also isn't very impressive considering the cost compared to the amount you'd get from HDDs at an equivalent total cost. $50,000 would get you 1 PB of HDD storage, to put things in perspective.

So at the end of the day, this article paints the same picture of SSDs at the consumer level, only scaled upwards. SSDs have always been, and will be for the foreseeable future, the middle-ground between truly face-melting performance (RAM drives), and truly gratuitous storage space (HDDs).

 

[warezme]

That's disgusting. (in a good way)

 

[internetlad]

Kevin Parrish got drunk and ordered a bunch of drives on Newegg again, didn't he?

 

[cangelini]

[citation][nom]jeffunit[/nom]"As we know, the SSD DC S3700 ships in capacities as low as 10 GB. Two dozen of those smaller drives could do some real damage in the right hands. After all, you'd be looking at 2.4 TB in RAID 0."I suspect there is a typo here. 10gb*24= 240gb, not 2.4TB.Perhaps what was meant was 100 GB?Perhaps someone needs to proof read articles before posting?[/citation]
Fixed!

 

[g-unit1111]

24 800GB SSDs... that's over 19.2TB... wow... *head explodes*

 

[mindbreaker]

What no system giveaway?

 

[swifty_morgan]

yeah, but can it play Crysis......??

did I read somewhere that IBM ( i think ) was going to invest One Billion Dollars into research/development in SSD's for server use? retiring the platter.... they had it figured out, even though the SSD was more expensive up front, it was actually cheaper to do so in the long run.....

 

[frontliner]

Please perf test using Spaces before simply bashing it & Windows. It is not like the disk management RAID setup of old. I bet it could stand up fairly well against Linux in this testing. I use Spaces today with SSDs for MSSQL processing and it seriously outclasses RAID controllers, because the 9207-8i + Spaces can pass raw data much faster and lower latency than a HW RAID Controller.

 

[ipwn3r456]

Should I get this or a car? lol

 

[warmon6]

Should I get this or a car? lol

Of course you should get this. Your on a Hardware site. : P

 

[redeye]

when will you be reviewing the IBM 14TB of flash storage for 331,000 dollars?... when IBM donates it to for a test ROFL...
seriously, if that did happen TOMS Hardware would be in the upper echelons of testing websites!!!...
but of course this is the stuff dreams are made of!!!

april fools!... oh wait... that is nor right...

 

[remosito]

Am I blind, but can't seem to find numbers for how much CPU the software RAID uses (Linux). Any chance you could redo these tests with raid 10?

 

[remosito]

Am I blind, but can't seem to find numbers for how much CPU the software RAID uses (Linux).

Any chance you could redo these tests with raid 10?

 

[WyomingKnott]

Am I blind, but can't seem to find numbers for how much CPU the software RAID uses (Linux).

Any chance you could redo these tests with raid 10?

On the last page, it says

All of this could have gone horribly pear-shaped without a proper server to drop the disks in. For what we wanted to do, we needed all of the performance our Xeon E5s could give us. We successfully found the limit of what the SSDs and our test platform could do together, and the result was on the (very) high side. The RAID 0 calculations performed by the host are fairly lightweight, but creating enough of a workload to stress the storage subsystem is fairly taxing.

Admittedly, it states that most of that load was the testing threads.

 

[Evolution2001]

I'm fairly certain I get the idea here that this article / testing was simply material for the techie spank-bank. And with that, I can only guess that any person or company that would invest in something like this would have parallel systems or a comparable backup system.
In a situation that would require(?) this type of capacity, whether it's programs or data being stored on these drives, it would be 'mission critical' and there's no way it could be considered responsible to build it on a single RAID0 where one single failed drive or controller could bring the entire setup to an immediate and unrecoverable stopping point.

All that reality aside, happy wanking! :-þ

 

[dontknownotsure]

nice 3d graph

 

[ron baker]

We double that with eight drives and quadruple it with twenty-four.

DOH!

 

[Aegean BM]

You gotta love Tom's Hardware. What benchmarking site knows more that Tom's? When they write something, you know, they know what they are talking about.

If you don't like Tom's, then that just means you don't like the results, which means you don't like facts.

So if I don't like Tom's choice of tests, or Tom's web layout, or their crappy Kontera ad links that screw up my iPad reading, or their sophomoric pictures of girls holding hardware (look at me! I have boobs under this tank top and I'm holding a wireless A/C router, too), or how they exaggerate a 3% difference between two products as a "notable lead" to pick a winner, then I don't like facts? There's a lot of crappiness about Tom's site. They just don't care enough to take out the garbage and put in the polish.

However, I totally agree with you that Tom's writers know what they are talking about in hardware.