Raid 0 Sata 2 (x2 SSDs) vs. Single SSD Sata 3

My Sony vaio SA has room for 2 SSDs but, one of the slots only supports SATA 2 which is half the speed of SATA 3.

So, I could only do a Raid 0 setup @ Sata 2 speeds.

Would 2 SATA 2 drives in Raid 0 be faster than 1 in SATA 3

 

I don't see why, when an SSD is fast enough as it is, you would want to RAID them together and double the risk that you'll lose all your data (we all know that SSDs are sooooo reliable right?). The marginal gains from RAID'ing a SSD are minimal at best and probably zero when you factor in the higher cost; you still have the same low latencies and even though you in theory double your throughput it's still already 'fast enough' that most people are not going to notice or care if it's any faster.

 

Thanks for the quick response, Greg!

I guess I'll just throw a big 7200rpm HD in that second slot then!

 

Really Greg, you know if it exists somebody's going to go for it if for no other reason than because it's there or that they can. And I've already seen some of the demonstrations on the web and they are very impressive.

Nevertheless, unless money is no object for those intrepid few, practicality has to pull us back to reality. Even with the lightning speed and instant gratification, the cost ratio is just not worth it.

 

You could also say you should not buy 2 laptops because there is twice the chance one of them will fail. lol

Would you also suggest because one drive only has half the chance to fail as two that it is not necessary to do backups.

Cost ratio? my 2 60GB drives cost $10 more than the 120GB drive.


Performance wise there certainly is a very noticeable difference in using 2 drives over 1 of the same type. How much difference there is between 2x60GB and one 120GB I can not say.

 

LapbookNotetops,

Besides the almost doubling the likelihood of losing your data spread over two drives, you should also understand you will eventually lose performance over time due to write amplification issues as today's RAID drivers do not *yet* support TRIM.

Also to clarify (yes, I know it posted as a jest), the chances of failure are not because you have multiple drives, laptops, etc. You could put two exact same drives in your system, and the chances of losing a single drive would remain the same. The increased failure rate occurs because of the *dependency* of multiple drives - a loss of any SINGLE drive in a RAID-0 volume causes the entire volume to fail. - http://forum.notebookreview.com/har...99-setting-up-raid-0-oneself.html#post7370656


Also, some other posts to consider:

http://forum.notebookreview.com/sol...5112-why-ssds-raid0-make-almost-no-sense.html

http://forum.notebookreview.com/sol...torage/610593-clevo-x7200-intel-320-ssds.html

 

Pardon me for being a complete contrarian, but after reading that whole thread on why RAID0 for SSDs make no sense, I have to say that it is at least 50% hogwash.

1) The risk of losing all your data in a crash with 2 drives in RAID0 is exactly the same as 1 drive, which is what the majority of laptops have. (I presume the OP is removing his/her optical drive and putting in a second HD.) This is why we have - BACKUPS!

2) You can't tell anyone who has owned a laptop with one SATAII SSD and also a Sony Z1 with SATAII SSDs in RAID0 that it doesn't make a difference! Think in terms of DOUBLE the speed!

3) People have owned Sony Z1s for almost 2 years now and there are threads tracking SSD performance and the conclusion is overwhelming: the garbage collection (GC) on the Samsung and Toshiba SSDs in the RAID arrays in the Zs have performed pretty much as TRIM does: there has been no noticeable degradation of performance. Also, Intel has announced TRIM for RAID is coming in a future RST updated

That said, however, in the OP's case, I happen to agree that taking two SATAII SSDs and RAID-ing them when there is an option of a SATA III SSD does NOT make sense. I believe a good SATAIII SSD with a SATAIII connection will run almost twice as fast as two SATAII SSDs in RAID O. Either way, you're in nosebleed territory as to speed, far faster than the original SATAII single SSD speed that got everyone hooked on the technology. But definitely go for the SATAIII and use the other bay for a bargain priced SATA II SSD or a high-cap HDD for media.

 

I think in this case a single fast SSD + mech HDD data backup is a better idea.

 

I agree 100% with Greg. Get a single SSD for OS, apps, games + single mechanical HDD for bulk media storage.

If you ignore the whole discussion about RAID-0 increasing chance of data loss, you're still left with the fact that 2x SSDs in RAID-0 gives you a negligible real-world performance boost. Since you're limited to a laptop form factor with only 2 drive bays, you'd find it much more useful to put a large inexpensive mechanical HDD in that second drive bay for use as bulk storage.

I would also recommend you consider getting a large 5400rpm drive, instead of a 7200rpm drive. In general, a 5400rpm drive is cheaper, produce less heat, and draw less power than a 7200rpm model. Furthermore, the kind of content you're storing on that drive (music, video, pictures, documents, etc) isn't affected by drive performance. A 1080p BluRay rip is going to play equally well off of a cheap low-temperature 5400rpm drive as it will off of an expensive fast higher-temperature 7200rpm drive.

 

Agreed. Unless you're in an situation where money in no object. that the most efficient and most economical arrangement for the present and near future.

 

This one is nonsense, of course: Hardware failure probability, and with it the probability of an associated loss of all data, is exactly twice as high for the dual-drive RAID0 configuration as it would be in a single-drive configuration.

Otherwise I agree with you.

 

not to mention that data stripping is an issue itself, and the op aint taking that into account.

RAID is for specific solutions, specially software RAID.

 

Let's not get too carried away here.

No it's not. There's no disputing the math, the rate of failure for a RAID-0 is (1.0-x)^N where x is the failure rate of a single drive, and N is the number of drives. Probability is almost 2x for a 2 drive RAID-0 array than a single disk SSD. Next point.

Agreed. I don't believe the numbers are double as there is going to be some overhead w/ the RAID, but [EDIT]device controller aside[/EDIT] it should be really, really close [EDIT] with large sequential IO operations[/EDIT].

Too many factors in play here. What people? How do you know the OP is going w/ Samsung or Toshiba SSDs? How do you know how the user will use their computer? Will they be doing massive writes/deletes on the array? There can be no dispute that TRIM plays an important part in the performance of an SSD (or a RAID volume on SSDs for that matter). The mere fact that Win7, OS X, and Linux have all added it to the core OS is a mere testament to that.

Yep. - note the emphasis on *YET* in my previous post. For those interested - more info regarding TRIM in RAID-0 and performance - http://www.anandtech.com/show/5136/intel-to-add-trim-support-for-raid-0


Finally, in regards to Tiller's post on why SSD makes no sense - if you disagree, go ahead on post to that thread. Tiller just loves a good debate!

 

Keep in mind, RAID-0 only really boosts Sequential Read and Write Speeds. And we all know that sequential read/write patterns are prety meaningless when it comes to real-world performance. A large boost in sequential read / write speeds is only going to be useful in a very specific niche of use cases (e.g. large RAW photos or video editing).

The actual metric that matters for real-world performance when it comes to multitasking, high NCQ queue depth reads, OS, apps, games is random read speeds. And RAID0 has practically no effect on random read throughput.

A RAID-0 array of mechanical disks is bottlenecked by the seek times of the drive. And a RAID-0 array of SSDs is bottlenecked by the disk I/O controller. PC Perspective did a bunch of tests in late 2010 to measure the impact of RAID-0 on consumer level hardware RAID. The conclusion was that a single Intel X-25M G2 SSD maxed out at ~40,000 IOPS under the most extreme circumatances (4KB reads @ QD=32). Using 2x Intel X-25M G2 SSDs in onboard hardware RAID-0 maxed out at ~50,000 IOPS, because the system was bottlenecking at the Intel onboard RAID controller. They had to go to multi-$100 server level RAID cards with dedicated onboard processors to exceed 50,000 IOPS.

Don't get me wrong... 2x SSDs in RAID-0 is of course better-performing than 1x standalone SSD. If you had unlimited resources (money, drive bays, heat, power, SATA ports, etc), then of course go for it. But unless you're a professional video / photo editor, RAID-0 on a laptop just isn't worth the money and drive bay for the marginal and circumstancial benefit.

 

In my experience:

-Failure rate for RAID0 is indeed higher than a single drive. There are more opportunities to fail in the same time frame.

However, if you are't backing up in Win7... you are making a mistake.
If for some reason you cannot fit that SSD (or SSD RAID) on your 2GB USB3 external, then I think you need to do some house cleaning.

Backup/restore in Win7 has never been easier, and those of you with Ultimate (or Enterprise) can actually boot from the VM on the USB drive it is backed up to!

Failure of a disk really shouldn't be an issue for most people creating RAID volumes. Yes, its a downside... but one that can be dealt with fairly easily.
I think I have had three drives from my personal computers fail on me completely in the last 20 years. (Servers are different)

-Performance... I am sorry but the performance is increased across all categories. Yes, the sequential is the highest, and no, most people wouldn't notice, but its still there. RAID0 SSD most definitely DOES help certain games as well, although its mostly loading areas/levels. I have to say in older games like L2, it is absolutely amazing the difference even over a single SSD drive. (loading any new zone is instant on RAID0, on one drive it takes 3-4 seconds.. this can be the difference between winning or losing!)

-Degradation... yep write performance degrades over time as TRIM doesn't function for now. Not really an issue for those willing to do manual maintenance, but yep its more trouble until Intel releases RST 11.5.

-One thing not mentioned... you can buy two smaller drives, get better read performance, and get plenty of really fast space on the cheap. The truth is, most personal computers need read performance more than write.

-All the above considered, I really think the RAID0 SSD config is not really a laptop thing. It works MUCH better in the desktop where you have room for at least one mechanical drive so you can lessen the effect of the degradation and have somewhere to put data files.

 

Edited my post. Thanks.

 

To answer his question though, 2 sata II drives in RAID is faster than 1 sata III drive, right?

 

Dunno. There are many variables in play here:

+ what drives
+ what kind of controller is in the sony?
+ what will the user's usage pattern look like?

The overall consensus I'm seeing on forum/benches is that it is a bit of a wash.

Also, one of the thing not mentioned in this thread here is the % of time your system waits on data from storage. Remember, retrieving/storing data from HDD/SSD only accounts for a small percentage of time an application uses the disk. By speeding up/eliminating that bottleneck, the *overall* results of using a piece of software may only increase by a small percentage. Real-world performance gains would depend on how IO bound the operations are for any given piece of software.

Finally, some quick searching revealed these links -

Technical Discussion SATA III RAID 1 vs SATA II RAID 0

Vertex 2 RAID 0 SATA II vs Crucial M4 AHCI SATA III - YouTube

 

Look at the pic...Raid 0 128gig Crucial M4's raised all numbers significantly...not just sequentials.

The only good reason i can think of for not going Raid 0 is the higher probability of losing all your data if one of the drives crash...but its not because of performance or lack of increase in performance. The 512 and 4k numbers were boosted significantly going Raid 0

 

lol...

You don't think the almost 5GHz clock speed on the cpu had anything to do with raising those numbers, huh?

 

I've had enough single drives fail that I don't worry about this. What happens when your single drive fails, you get it replaced and recover your backup or reinstall. What happens when a RAID0 drive fails, you get it replaced and recover your backup or reinstall. Doesn't seem much different to me.

Next, people will be advocating running RAM dual channel for extra performance isn't worth the risk due to doubling the chance of failure, lol. j/k

Here's a performance comparison from StorageReview running 2 160GB Intel 320's in RAID0 against 1x 300GB Intel 320. Price difference per GB in storage between two 160's and one 300 is about 7% going by todays newegg prices.

2x 160GB Intel 320 RAID0 vs 1x 300GB Intel 320 <caption></caption> <tbody valign='top'>

StorageMark 2010	2x 160GB Intel 320	1x 300GB Intel 320	RAID0 % increase
HTPC MB/s	304	152	100
HTPC IOps	6566	3271	100
HTPC Latency	1.15	2.37	106
Productivity MB/s	448	150	199
Productivity IOps	15206	5137	196
Productivity Latency	0.66	1.54	133
Gaming MB/s	380	243	56
Gaming IOps	7201	4558	58
Gaming Latency	1.08	1.7	57

It would be nice to see how the lower capacities fair against a single drive of double capacity as there can be a bigger performance gap between them, but I'm not that interested to go digging or pay for another drive to check.

 

Just for clarification, this question should read, "What happens when a RAID-0 drive fails..." as this is not a problem when one drive fails in a RAID 1, 5, or 10 volume.

OK, but this does not address the original question? RAID-0 SSDs (SATA II) vs. single SSD (SATA III)

Also, this brings up another question. A 2 drive RAID (using a Fake-RAID BIOS) over two different SATA speeds. What is the effective throughput? Are speeds determined by implementation of the RAID in BIOS? Are the speeds 3Gbps (SATA II for both drives), or is it slightly better due to one of the drives still running at SATA III?

 

This is past the point of diminishing returns. Yes, there is a difference. But I would argue that the difference is so small that it really isn't worth the tradeoffs (added cost, complexity, risk of dataloss, and reduced storage capacity).

You're not going to see much of a real-world performance difference between single SSD vs 2xRAID-0 SSDs. You'll get higher benchmark scores, or slightly reduced Windows boot times. But it isn't going to be anywhere close to the monumental performance shift from mechanical HDD --> SSD. You really need to put an artificially high workload on the drives before the benefit of SSD RAID-0 kicks in. And 99% of the people out there don't have usage patterns that can choke down a single SSD (with the exception being professional video and photo editors).




As for SATA-2 vs SATA-3... the SATA-3 bandwidth is going to benefit sequential read/write speeds. But real-world performance is better measured by Random Read speeds. And Random Read speeds doesn't even come close to saturating SATA-2 bandwidth, let alone SATA-3 bandwidth. So if you're looking for the top-performing SSD, look at Random Read speeds.

Now, the best performing drive may coincidentally be SATA-3. However, that is going to be due to the higher Random Read speeds of the SSD controller, and not because the drive happens to support SATA-3.




I believe that once you get above a certain level of performance that is "good enough," then these differences in performance don't really matter. Other factors such as price, capacity, and reliability are going to be higher priorities once you reach that "good enough" level of performance. I say, buy a single SSD at a good value, and use the 2nd drive bay for a slow, inexpensive 5400rpm mechanical HDD.

 

Yes, that was a typo, it was supposed to be RAID0, now corrected thanks.

The first part shows what performance may possibly be had from using one drive vs two smaller ones to give comparable size.

The last part I think might be pertinent to the OP's question as I would think the OP would either use one bigger drive or two smaller ones at a similar overall cost and combined size. This could mean more possible losses due to reduced performance of the smaller capacity drive. Something IMO to bear in mind when making a decision.

 

Raising the numbers like that? no. The bclk is still only 104 which means the SATA3 bus is constant. Plus the drives can only operate so fast, regardless of CPU speed

here let me post the two side by side for you to compare

Heres without Raid 0...single 128gig Crucial M4



heres with Raid 0 2 x 128gig Crucial M4's



A 5 ghz CPU will not account for increases like that.

 

Those numbers pretty much confirm what i've seen with my own eyes. SSD's in Raid 0 are significantly faster than a single SSD

 

This is off topic, but the data here blows away Tiller's 25% increase statement from his other thread. If we're talking straight disk I/O, then that statement is just plain wrong.

 

No, drives don't operate in a vacuum. The system/platform/cpu they're being driven by make a difference (always).

You really believe those synthetic benchmarks run at different settings mean anything (including StorageMark on two different types of drives!!!...)?

Also, you really believe that a RAID0 setup can have scores of more than double a single drive's 'scores' and this fact doesn't raise the validity of the synthetic benchmarks for you? (I have some ocean front property in Montana I can sell you 'cheap'...).

Not to mention that both drives were set up obviously differently (18GB/119GB vs. 71GB/238GB).

As to the 'data' that is presented here blowing away the 'data' that Tomshardware presented that shows much less performance differences - well, you can take/analyze/twist it and interpret it any way you want.

All I know is that in my actual real world use from a decade+ ago (where RAID0 made the most performance difference, btw...) I stopped using RAID0 because the performance increases were not worth the additional hassles and headaches when the array went down (for one reason or another).

If someone were to do this comparision right (as I did in circa 1999) for themselves, I don't have any doubts that they will come to the same conclusions too.


How to 'do it right':
Given that they test a single drive vs. the exact same drive times 2 in RAID0.

Given that they test only real world result/workloads. Synthetic tests are so boring not to mention inaccurate vs. what actual % difference results would be seen in a real work load.

Given that in a multiple HDD bay platform, they use multiple drives 'intelligently' (for example: moving the Users folder to a different physical drive; using a different physical drive (short-stroked) as a temp/scratch disk; and/or using a different physical drive (than the O/S drive) to store data on).

Yeah, not saying I don't believe the 'data' presented, but... I don't.

 

Tiller, I didn't think that Raid0 made that much of a difference either. But the evidence is growing and the numbers don't lie.

 

Yeah numbers don't lie - but you do have to look at what/who is generating those numbers.

Different drives compared??? What!?! How is that isolating RAID0 performance?

StorageMark 2010 - a trace based benchmark that is allowed to run at maximum speed??? How is that imitating how we actually use our storage sub-systems?

Single SSD vs. same SSD x2 but with different setups and different benchmark settings??? How is that 'evidence' that is growing?

Make no mistake about it - a RAID0 setup will 'feel' faster than a single drive solution. But so will a two drive NON-RAID0 setup too when setup optimally.

The difference with a two drive Non RAID0 setup is that when a drive fails - you'll still have your 'other' data on the other drive.

When a RAID0 array fails - you have no data at all.

See why RAID0 'almost always' makes no sense now?

 

When I said "straight disk I/O", I mean all we're talking about is about measuring disk I/O. No other variables are at play.

So, I took the statement from the other post to mean a 25% improvement in just plain disk operations. However, from this latest post, I *think* you're saying that there will be a 25% overall improvement because disk I/O is a mere portion of the entire process. For example, loading a file, performing some encoding, and saving to a new may see an overall 25% improvement due to just a change in disk. Is that accurate?

Can you clarify the original statement?