Why SSD's in RAID0 make almost no sense.

Almost no sense? Sure, there are some specialized cases where it may even be a requirement to use the software fully, but in a notebook system without a Video editing/creation requirement, RAID0 is simply for bragging rights.

See:
Benchmark Results: PCMark Vantage : SSDs In RAID: A Performance Scaling Analysis


I know, I'm using a benchmark (when I don't like benchmarks) to make my point; but that is the best 'real world' indication the above article can provide us.

SSD or HDD; RAID0 is only beneficial (overall) in very few scenarios and the additional cost and potential headaches more than outweigh an ~25% performance increase over simply using a single drive by itself.

If you do have the option to use two drives, setting them up optimally, yet independantly, will make that ~25% delta even smaller and without any associated risks of running your data (stripped) across two (or more) drives.

As an aside; I have come to this conclusion independantly of the article above on my desktop systems - almost a decade ago.

So, for those thinking the 'ultimate' is RAID0 on your notebook - sure, it can be. But don't spend your money thinking that a 2x SSD RAID0 setup is twice as productive (or anywhere close to that) as the single SSD solution.

 

That and TRIM isn't passed along to the SSDs so you are degrading them. Unfortunately people see larger numbers and assume they are better.

 

In terms of "short term" performance, wouldn't you want to look at the *overall* picture here? While PCMark does measure some HDD statistics, the CPU and RAM also are measured within the test, leading to a more weighted score.

If you look at the previous pages for IOMeter tests strictly a IO based bench, a single disk vs. RAID-0 (2 disks) does offer a significant performance (looks like 80-110%) increase. But I think that is your point... other parts of the system will be in play. A user WILL need to account for CPU and RAM as all parts together run the software we use.

I still say to look at the long term picture. I personally wouldn't recommend RAID 0 myself as the point of failure is greater and you will eventually hit the WRITE degradation issue. But hey, if someone is adamant about backups, has extra drives in case of drive failure, and doesn't mind restoring in the case of a catastrophe or periodically wiping their drives when write performance is worse than a single drive, then live and let live.

Also note, to muddy the waters... some laptops are capable of running RAID 0+1. But I guess that is a whole other thread.

 

Tsunade_Hime,

Good points!


jclausius,

I know of no notebooks that have 4 bays to run a RAID 10 config?

Also, I don't know of any test that will give 110% better with 2 RAID0 drives vs. a single drive in the same test.

Keep in mind that I am looking at overall performance here - a ~25% increase in synthetic benchmarks is meaningless when they are artifically inflated (biased as they are by a high 4K R r/w in Vantage). Especially if your work is more cpu/gpu dependant - it doesn't matter if your storage can perform faster than it is being fed finished data from the processor/ram.

 

While the headaches definitely outweigh the potential speed from setting up an SSD RAID0 array for the average user, some people might have a legitimate need for the extra capacity. I can think of a number of workstation-class scenarios where having 1TB or more of SSD storage is important.

 

If you have the drive bays available, then you'll have the same capacity.

Why throw a wrench called RAID0 into it?

 

With a desktop and old disk drives...yes there was a clear benefit but a lapper with ssd's .... no need IMO.Specially with there failure rate ATM.I couldnt recommend it even in a desktop but some people love that kind of stuff...I used to be one of them too

 

The Clevo x7200 / Sager NP7280 can do it. Some resellers are providing an option for the x7200 to use a 4th disk drive instead of the optical drive -> total of 4 bays. So it's just a matter of ordering that configuration and setup.

Look at the previous pages' ioMeter results on the link you provided. Unless I read the graphs incorrectly, single disk vs. 2 in RAID 0 for "pure IO" for all the different IO based measurements look like ~80% to ~110% increase over single disk.

For instance, 4k Random Reads - somewhere above 30K IO/s for single SSD and above 60K IO/s for RAID-0 (2 disks). Random writes is not as good -> looks like 30K vs. 50K. Streaming Reads 117 IO/s vs. 219 IO/s, etc.

Also in the conclusion of the article they mentioned the more drives you throw in RAID-0, the near linear increase of the results (although I wouldn't think it is a 1:1 slope of drives/increase in performance).

Note, I'm still would not recommend SSD in RAID.

 

Oh, I have read about 3/4 Bay notebooks (of course), just none that could do RAID10 (still not sure that the Clevo/Sager can?).

Yeah, you're right about greater than 100% with 2 drives in RAID0 - however, this is simply an anomoly in the controller used (it doesn't allow a single drive to reach it's full potential, obviously).

Add to that the caching, and other driver interactions a RAID card can have on these scores (not to mention that for use in notebooks, a 'server' setup is highly unlikely and would not scale at anything close to this dedicated RAID card) and the fact that high I/O 'scores' on queue depths greater than 5 are almost impossible on single user workstations...

and you can see why I totally ignored those benchmark 'scores'.

 

The x7200 is an awesome machine, capable of multiple configurations. And it's RAID 10 capable.


I think this might help crystallize your point. Now, I *do* believe in the beginning RAID 0 SSDs may possibly provide up to 100% performance improvement over a single SSD. HOWEVER, in the general case, disk IO is a mere fraction of how your computer operates. The other components required to run software (CPU, northbridge, southbridge if applicable, video cards, RAM, etc.) are not sped up by using RAID. That is the point of the PCMark benches. The disk IO is amazingly fast, but in the overall scheme of things it does not shine through due to the other components. Thus the modest gains in that bench.

So, if you spend a lot of money (double SSDs) to speed up an entire process, is that money well spent? If you were a general user or occasional gamer on a budget and had to choose between a single SSD and high end GPU (or SLI/CF) or RAID-0 SSDs w/ an average GPU, what would be the better choice? Speeding up reading the info from disk which may happen only a fraction of your use, or speeding up the GPU(s) which would be used a majority of the time while running any of today's current games?

What if you're someone who needs to do some heavy, heavy IO, video encoding for example, make back-ups every 6 hours, and know about wiping drives every so often to avoid write degradation in RAID-0? What would be the better choice in that scenario?

In the end, only the person spending the money can really be the judge. I would prefer to get the information out in the open and let the person who knows what they need and their technological limitations make an informed decision.

 

jclausius,

Thanks for the additional information.

I agree the more real info we have, the better decisions we can make based on our specific needs and conditions.

 

one major point of ssds is latency. the low latency is one of the reasons why they feel so fast. raid0 might increase it measurable, even noticable (esp on crap onboard chips). never verified, but an idea.

an average hdd having 10ms latency is not affected by 0.1ms really. but a 0.065ms intel ssd would be.

no clue what latency increase a raid controller brings, though.

 

dual drive RAID 0 is just like dual channel memory as far as performance gain is concerned.

 

While taken slightly out of context, this link shows how little storage plays in actual productivity (as I've been trying to point out for over 18 months here...):

Notice that the RAID0 SSD 'storage' score in the link below is over 9x the HDD score (comparing $500 to $2000 systems) - BUT, the three other productivity scores are all well below this highly inflated number (183%-208%, or just over double the performance on these still mostly synthetic scores - for 4x the money).

See:
Performance Scaling And Efficiency : System Builder Marathon, March 2011: Value Compared



Note also that we are not keeping the system the same either though: the $500 build is using last century technology (AMD) while the $2,000 build, in addition to 2x RAID0 SSD's is using one of the best processors currently available. A lot of that 2x performance gain is also realized through the cpu - not solely on SSD performance (even RAID0 SSD performance).

This dovetails nicely with what I've seen and have hinted at with my posts lately:

See:
http://forum.notebookreview.com/7296747-post96.html

See:
http://forum.notebookreview.com/7297853-post100.html

See:
http://forum.notebookreview.com/7299760-post104.html



This info above also nicely contains (in one post) why productivity and percieved performance ('snap') is so far off with SSD's. 'Far off' between me (focusing on productivity) and almost all others (focusing on 'snap').

A system can have no snap, yet be able to do a great deal of work (eg. slow, lumbering, 'green' 5400 RPM drives with an i7 990X, 24GB 'work engine') or it can have great snap and be worthless (relatively! ) in a 'production' setting (eg. almost any SSD in an Atom based netbook).


As I've stated before; actual 'work' only happens in the RAM and CPU - if you want to increase your output, upgrade one or both of those to the maximum your platform and wallet can stand. Storage, while it will make a system respond smoother, will not increase the performance/productivity/output of a system as much as those two will.

Even in a RAID0 setup.

 

Yes. We covered this earlier...

 

jclausius,

sure, but some people still don't get it.


Even with the 'proof' above.

 

What people? What are they saying?

As I mentioned earlier, RAID0 may or may not make sense. It really all depends as there's no black and white here.

Also, if someone reads the data in this thread (my elderly neighbor for instance), and she insists on RAID-0 over SSDs, even though I know she won't need it (needs a machine to email grandkids and type letters), I'll explain to her the pros/cons of that decision. If she still insists on it, then I'm not going to lose sleep. I did my job of informing her. When she says she can boot in 8 seconds, I can hold out the extra $300 or so, and say that I'd rather have the cash. And if she loses her drive and forgot about backups, then all I can say is "told-ya."

In a way, this same argument would hold for someone insisting on SLI/CF when it is not needed. If someone doesn't do any gaming (me, for instance), but insist on spending the money on double GPUs (No, I didn't go down that route). Just spending money foolishly on power you won't (or can't) use. But to each his own.