RAID0 overhaul

[mojo8472]

Hey guys,

My PC is finally in a place where the slowest link in the chain is the HDD (a WD Caviar Black 500GB) and I'm thinking about doing a full on storage overhaul.
My plan is this:
2x 60GB Vertex 2Es in RAID0 for boot, apps and a few games
3x 500GB Caviar Blacks in RAID0 for my Steam folder, media etc

Now, my motherboard (Asus P7P55D-EVO) has both an ICH10R controller as well as 2 ports for a JMicron JMB322 RAID controller. My questions to you guys are:

Would the JMicron controller be better suited to RAID the SSDs than Intel's?
Does Intel's controller let you set up multiple arrays?
And would the whole thing be bottlenecked by 2 SSDs and 3 HDDs running together?

 

[taltamir]

RAID0 causes you to lose TRIM, which lowers your performance a good bit. Personally I wouldn't do it unless its at least 3 or 4 SSDs.

I would use the intel's over the jmicron.

And would the whole thing be bottlenecked by 2 SSDs and 3 HDDs running together?

I am not sure I am getting what you mean by that.
if you mean to ask if the overall planned bandwidth (and the locations of programs and data) is enough, then probably it will be enough. Bottlenecks always depend on the specific task being performed though.

If you mean to suggest that having 2 separate RAID0 arrays would somehow harm your performance, then no it will not.

 

[MagnusTheBrewer]

Do you benchmark? If not, RAID 0 for SSD's is completely worthless.

 

[Chapbass]

Do you benchmark? If not, RAID 0 for SSD's is completely worthless.

Eh? how so? from all the reports I've seen, RAID on SSD's appears to be orders of magnitude better than RAID for HDD's, and performance seems to be almost linearly increased from a single SSD.

EDIT: Heres an article from AT about it, featuring the x25-v 40GB drives.
http://www.anandtech.com/show/3618/intel-x25v-in-raid0-faster-than-x25m-g2-for-250/1

From the showings, it looks like everything is noticeably improved over the single x25-v. Obviously the numbers on some of the charts arent groundbreaking, but given that theyre using pretty much the cheapest SSD around from intel, its more for just comparison. Pair it with two better drives (x25m, vertex/agility2, etc) and you'll get huge performance benefits, imo.

Now, the one part you could be right is if the OP doesn't use his storage very much. If hes doing decent multitasking or anything like that though, it definitely removes a bottleneck. My guess is that hes doing a decent amount of it, given the fact that hes looking to upgrade his storage in the first place, but maybe thats a poor assumption on my part.

 

[taltamir]

Eh? how so? from all the reports I've seen, RAID on SSD's appears to be orders of magnitude better than RAID for HDD's, and performance seems to be almost linearly increased from a single SSD.

Other way around.
RAID0 has linear increase in speed, 2 HDD get 2x the speed. Since HDD are so much slower in the first place, you notice RAID0 a lot more than you do on a blazing fast SSD.

SSDs however, lose out on TRIM, which lowers the individual performance of each drive. They are still faster though.
They are definitely NOT orders of magnitude faster. 10x is an order of magnitude... so it takes 10 drives in RAID0 to get 10x aka 1 order of magnitude increase in speed.
With SSD thats closer to 15 drives needed for 1 order of magnitude due to loss of trim.

TRIM issue mentioned in the same article you linked: http://www.anandtech.com/show/3618/intel-x25v-in-raid0-faster-than-x25m-g2-for-250/6

Also: http://www.anandtech.com/show/3618/intel-x25v-in-raid0-faster-than-x25m-g2-for-250/3
notice that random read on 2 of the X25-V 40GB in RAID0 actually get LOWER performance than one single X25-M 80GB (albeit by a very small insignificant amount) and are only faster then a single X25-V 40GB by a small insignificant amount, essentially they perform about equally.
Sequential writes, sequential reads, random writes, and random reads all need to be looked at individually.

I am not saying RAID0 of SSD is necessarily BAD, if I could afford it I would use 4 SSDs in RAID... although you could just buy the OCZ collosus or one of their several other models that does exactly that (4 way raid0 that is)

PS. no TRIM also raises the concern of rampant write amplification eating up your drive's lifespan much quicker. I need to further investigate it though.

 

[MagnusTheBrewer]

Eh? how so? from all the reports I've seen, RAID on SSD's appears to be orders of magnitude better than RAID for HDD's, and performance seems to be almost linearly increased from a single SSD.

EDIT: Heres an article from AT about it, featuring the x25-v 40GB drives.
http://www.anandtech.com/show/3618/intel-x25v-in-raid0-faster-than-x25m-g2-for-250/1

From the showings, it looks like everything is noticeably improved over the single x25-v. Obviously the numbers on some of the charts arent groundbreaking, but given that theyre using pretty much the cheapest SSD around from intel, its more for just comparison. Pair it with two better drives (x25m, vertex/agility2, etc) and you'll get huge performance benefits, imo.

Now, the one part you could be right is if the OP doesn't use his storage very much. If hes doing decent multitasking or anything like that though, it definitely removes a bottleneck. My guess is that hes doing a decent amount of it, given the fact that hes looking to upgrade his storage in the first place, but maybe thats a poor assumption on my part.

You completely missed the "do you benchmark" comment. All of Anands results showed "benchmark" improvements. Show me one appreciable real world improvement in putting SSD's into RAID 0.

 

[HendrixFan]

You completely missed the "do you benchmark" comment. All of Anands results showed "benchmark" improvements. Show me one appreciable real world improvement in putting SSD's into RAID 0.

http://www.anandtech.com/show/3618/intel-x25v-in-raid0-faster-than-x25m-g2-for-250/5

 

[MagnusTheBrewer]

http://www.anandtech.com/show/3618/intel-x25v-in-raid0-faster-than-x25m-g2-for-250/5

So? Look, I have nothing against folks who like to benchmark but, I have seen nothing that shows appreciable real world improvement. does it encode a 5 Gb file faster than a couple of seconds? Does it improve load times for games by more than a second? Does it aid rendering by more than a couple seconds? Everyone has shown nothing but benchmarks. Where is the real world advantage?

 

[HendrixFan]

So? Look, I have nothing against folks who like to benchmark but, I have seen nothing that shows appreciable real world improvement. does it encode a 5 Gb file faster than a couple of seconds? Does it improve load times for games by more than a second? Does it aid rendering by more than a couple seconds? Everyone has shown nothing but benchmarks. Where is the real world advantage?

The Anand Storage bench uses Firefox, Office 2007, Adobe Reader, Outlook, Windows Media Player 11, Microsoft Security Essentials, Photoshop CS4, 7-zip, uTorrent, and World of Warcraft. Please read the link I sent you if you care to actually learn about the technology. Those are real world programs in real world environments running real world tests.

Encoding is CPU bound unless you are writing a file faster than 100 MB/s or so while encoding. I don't think any storage upgrade will affect encoding too much, due the the nature of the job. That is like asking if new car stereo will make my car faster. That isn't its job.

 

[MagnusTheBrewer]

The Anand Storage bench uses Firefox, Office 2007, Adobe Reader, Outlook, Windows Media Player 11, Microsoft Security Essentials, Photoshop CS4, 7-zip, uTorrent, and World of Warcraft. Please read the link I sent you if you care to actually learn about the technology. Those are real world programs in real world environments running real world tests.

Encoding is CPU bound unless you are writing a file faster than 100 MB/s or so while encoding. I don't think any storage upgrade will affect encoding too much, due the the nature of the job. That is like asking if new car stereo will make my car faster. That isn't its job.

The 'benchmark' does include all those real world apps in amalgamation. I say again, show me the real world appreciable improvements in each individual application. A second here a second there. The fact is there is no real world appreciable difference to be achieved by putting SSD's in RAID 0. Your e-peen is a different story. If you continue to post 'benchmarks' then I'm happy for you if that's your thing.

 

[HendrixFan]

The 'benchmark' does include all those real world apps in amalgamation. I say again, show me the real world appreciable improvements in each individual application. A second here a second there. The fact is there is no real world appreciable difference to be achieved by putting SSD's in RAID 0. Your e-peen is a different story. If you continue to post 'benchmarks' then I'm happy for you if that's your thing.

*sigh*

You still refuse to read the link. Here is more info on the Storage Bench, don't be confused by the word "Bench".

http://www.anandtech.com/show/2865/8

Finding good, real world, storage benchmarks is next to impossible. The synthetic tests work up to a certain point but you need real world examples. Measuring performance in individual applications often ends up with charts like these where all of the drives perform identically. Benchmark suites like PCMark Vantage are the best we can do, but they are a bit too easy on these drives in my opinion and while representative of a real world environment, they aren’t necessarily representative of all real world environments.

Recently we’ve been able to get our hands on a piece of software that allows us to record all disk activity on a machine and then play it back on any other machine. The point is that we can now model a real world usage scenario without waiting for BAPCo or Futuremark to do it for us.

...

The benchmark is 22 minutes long and it consists of 128,895 read operations and 72,411 write operations. Roughly 44% of all IOs were sequential. Approximately 30% of all accesses were 4KB in size, 12% were 16KB in size, 14% were 32KB and 20% were 64KB. Average queue depth was 3.59.

It isn't just sequential read/writes that benefit from RAID, which is useful for loading games, but the random read/writes scale upward. It is in the random read/writes that there is the most noticeable difference from a standard HDD to a SSD, and the gap is only widened when SSDs are put in a RAID-0 array.

The Anand Storage Bench is not synthetic, it is just a method to have a 100% accurate recreation of the same computer usage, using programs that will stress your storage system. If you are trying to argue that random read/write numbers don't have a real world performance difference, or that running programs don't have a real world performance difference you will find few who buy into your ideas.

 

[MagnusTheBrewer]

*sigh*

You still refuse to read the link. Here is more info on the Storage Bench, don't be confused by the word "Bench".

http://www.anandtech.com/show/2865/8



It isn't just sequential read/writes that benefit from RAID, which is useful for loading games, but the random read/writes scale upward. It is in the random read/writes that there is the most noticeable difference from a standard HDD to a SSD, and the gap is only widened when SSDs are put in a RAID-0 array.

The Anand Storage Bench is not synthetic, it is just a method to have a 100% accurate recreation of the same computer usage, using programs that will stress your storage system. If you are trying to argue that random read/write numbers don't have a real world performance difference, or that running programs don't have a real world performance difference you will find few who buy into your ideas.

I believe you just proved my point. No one is arguing that there aren't measurable advantages to putting SSD's into RAID 0. What I am arguing is that the real world advantages are imperceptible to most users. If you are not into benchmarking putting SSD's into RAID 0 is just silly, not to mention, a waste of money.

 

[HendrixFan]

I believe you just proved my point. No one is arguing that there aren't measurable advantages to putting SSD's into RAID 0. What I am arguing is that the real world advantages are imperceptible to most users. If you are not into benchmarking putting SSD's into RAID 0 is just silly, not to mention, a waste of money.

Again, they are real world actions, not just benchmarks. If you are a heavy multitasker and you use disk bound programs like the ones listed in the storage bench, the difference between one setup and another is pretty significant. Since the ICH10R RAID controller scales very well with SSD drives, and the price/size ratio of SSD drives works out in the favor of smaller drives; getting two 120GB drives and running it in RAID-0 versus a single 240GB drive will be both cost effective and provide a significant boost in performance.

http://www.techpowerup.com/116575/Patriot_Shows_off_So_Called_Fastest_System_on_The_Planet.html


To give you an idea that performance isn't just synthetic, this link shows a very clear performance boost in real world performance on an extreme RAID array. Of course users won't be running a 40 drive array, but the performance does scale linearly and most people wouldn't be too concerned with ripping a Bluray in 0.9 seconds just yet. For the most part, running a 2 drive RAID-0 array will half the time it takes to do something.

 

[mojo8472]

If I could interject for a moment; don't the new Intel RST drivers coupled with Sandforce's garbage collection mean there wouldn't be any degradation in performance from a lack of TRIM in the RAID array?

 

[MagnusTheBrewer]

Again, they are real world actions, not just benchmarks. If you are a heavy multitasker and you use disk bound programs like the ones listed in the storage bench, the difference between one setup and another is pretty significant. Since the ICH10R RAID controller scales very well with SSD drives, and the price/size ratio of SSD drives works out in the favor of smaller drives; getting two 120GB drives and running it in RAID-0 versus a single 240GB drive will be both cost effective and provide a significant boost in performance.

http://www.techpowerup.com/116575/Patriot_Shows_off_So_Called_Fastest_System_on_The_Planet.html


To give you an idea that performance isn't just synthetic, this link shows a very clear performance boost in real world performance on an extreme RAID array. Of course users won't be running a 40 drive array, but the performance does scale linearly and most people wouldn't be too concerned with ripping a Bluray in 0.9 seconds just yet. For the most part, running a 2 drive RAID-0 array will half the time it takes to do something.

Lets see, he talks about Mb/s, pcmarks and, IOPS. Nowhere does he state, using RAID 0, it ran this operation x seconds faster. Are they indicative of faster operations? Yes, they are but, no one to date (to the best of my knowledge) has shown real world data giving specific improvements in specific applications that make SSD's in RAID 0 anything more than nerd bragging. Ask yourself this, if you let a computer knowledgeable friend use your computer not knowing it had SSD's in RAID 0 vs. a SSD boot disk, would he be able to tell the difference? I think not.

 

[HendrixFan]

Lets see, he talks about Mb/s, pcmarks and, IOPS. Nowhere does he state, using RAID 0, it ran this operation x seconds faster. Are they indicative of faster operations? Yes, they are but, no one to date (to the best of my knowledge) has shown real world data giving specific improvements in specific applications that make SSD's in RAID 0 anything more than nerd bragging. Ask yourself this, if you let a computer knowledgeable friend use your computer not knowing it had SSD's in RAID 0 vs. a SSD boot disk, would he be able to tell the difference? I think not.

Yes. My mother can tell the difference when I put my older 30 GB Vertex in her computer going from a HDD. She runs tons of apps at once on her desktop, working on genealogy files, photoshop, quite a few firefox tabs, acrobat reader, word and excel. Benchmarks showed there was an appreciable difference and her real world experience reinforced that.

The Storage bench runs at a set time of 22 minutes, so it won't run faster. That's not how it works, instead it measures the performance of the programs in that time. I think you have a fundamental misunderstanding of storage technology, or at the very least a huge misunderstanding of the storage bench Anand uses. MB/s (especially random) and IOPS are the key to faster storage speeds in real world use. The sequential speeds of an SSD aren't too much faster than a HDD, and can be slower than a couple HDDs in RAID-0. But it is in the IOPS and random MB/s that SSDs show their worth.

http://techgage.com/article/ocz_vertex_turbo_120gb/8

When running a virus scan or malware scan on a normal HDD, you completely bottleneck the drive and kill the performance of not only the scan but any other program you have running trying to complete an operation. Try unzipping a file and running a scan on a HDD at time it. Do the same on an SSD and see that you see an exponential increase. This link clearly shows that the IOPS and random read/write speeds of the Vertex drive give it a speed advantage over other SSDs, in the form of lower times. Those are "specific improvements in specific applications" that show better times.

What about RAID?

The Revo drive is essentially a RAID-0 SSD on a PCI-e card. Will running that show in improvement in those times?

http://techgage.com/article/ocz_revo_120gb_pci_express_ssd_performance_preview/

Yes it does, especially in the "heavy loads" where IOPS and random read/write really start to make a difference from one SSD to another. As you can see, going from nearly 8 minutes to 5 and a half minutes with RAID is a significant boost.

 

[MagnusTheBrewer]

I said, "if you let a computer knowledgeable friend use your computer not knowing it had SSD's in RAID 0 vs. a SSD boot disk."I did not say if they could notice the difference between an SSD and a HDD.

You STILL have not shown me real world "appreciable" advantages to putting SSD's into RAID 0. You can drool over IOPS all you like but, can you provide a link or article showing real life noticeable improvements in specific applications?

Engineers and techno geeks love to report on what they know but, will jump through hoops and turn themselves inside out before saying, I don't know or, I don't know how to measure that. Worse yet, are those who convince themselves that those things which either cannot be measured or, are difficult to measure, are unimportant.

 

[HendrixFan]

You STILL have not shown me real world "appreciable" advantages to putting SSD's into RAID 0. You can drool over IOPS all you like but, can you provide a link or article showing real life noticeable improvements in specific applications?

Engineers and techno geeks love to report on what they know but, will jump through hoops and turn themselves inside out before saying, I don't know or, I don't know how to measure that. Worse yet, are those who convince themselves that those things which either cannot be measured or, are difficult to measure, are unimportant.

Are you trolling? I just gave you a link showing a time decrease from 8 minutes to 5 and a half with RAID-0. That isn't appreciable? What is your standard? You asked if running RAID did anything "more than a couple seconds". I showed you minutes. It won't make a sandwich for you though.

You asked for measurements in time (since you are unwilling to see the link between performance in IOPS and random read/write and the time it takes to carry out a task) and now you have it and you respond by saying that somehow I'm stating it is difficult to measure or is unimportant? Again, in the link I sent you, the time it took to:

Playback of a 56MB FLAC music file in Winamp.
50 ~8MB images queued to open in Photoshop CS4.
Opening of three Excel, three Word, and one PowerPoint files (various large sizes, for example one Excel file consists of an actual 72MB database).
Browsing to four different websites in Firefox.
Extraction of a 1GB RAR containing numerous "program file" folders.
Extraction of an 893MB ZIP containing 100 RAW images.
Transfer of a 7.16GB file to a second partition on the same drive
Viewing of two PDF documents.
Viewing of two small RAR utility archives
Execution of four small utilities

All while running a full "Anti-Virus scan running concurrently in the background with the start of the test. The AV scan uses a static, unchanging 5.1GB test folder that contains 19,748 files and 2,414 sub-folders created from the Program Files directory."

The test went from 447 seconds to 321 seconds for the RAID drive.

I found the links containing the information you asked for, please take the time to read them before responding.

 

[HendrixFan]

If I could interject for a moment; don't the new Intel RST drivers coupled with Sandforce's garbage collection mean there wouldn't be any degradation in performance from a lack of TRIM in the RAID array?

I wouldn't say there won't be any degradation, but the GC does do a great job of compensating for a lack of TRIM.

 

[taltamir]

Benchmarks are perfectly legitimate for estimating performance increases, but it is important to understand exactly what those benchmarks are saying and not draw incorrect conclusions.

I wouldn't say there won't be any degradation, but the GC does do a great job of compensating for a lack of TRIM.

one test I saw showed TRIM getting 99.something% of new drive performance, GC getting ~90%, and neither getting about 65%

 

[MagnusTheBrewer]

Are you trolling? I just gave you a link showing a time decrease from 8 minutes to 5 and a half with RAID-0. That isn't appreciable? What is your standard? You asked if running RAID did anything "more than a couple seconds". I showed you minutes. It won't make a sandwich for you though.

You asked for measurements in time (since you are unwilling to see the link between performance in IOPS and random read/write and the time it takes to carry out a task) and now you have it and you respond by saying that somehow I'm stating it is difficult to measure or is unimportant? Again, in the link I sent you, the time it took to:

Playback of a 56MB FLAC music file in Winamp.
50 ~8MB images queued to open in Photoshop CS4.
Opening of three Excel, three Word, and one PowerPoint files (various large sizes, for example one Excel file consists of an actual 72MB database).
Browsing to four different websites in Firefox.
Extraction of a 1GB RAR containing numerous "program file" folders.
Extraction of an 893MB ZIP containing 100 RAW images.
Transfer of a 7.16GB file to a second partition on the same drive
Viewing of two PDF documents.
Viewing of two small RAR utility archives
Execution of four small utilities

All while running a full "Anti-Virus scan running concurrently in the background with the start of the test. The AV scan uses a static, unchanging 5.1GB test folder that contains 19,748 files and 2,414 sub-folders created from the Program Files directory."

The test went from 447 seconds to 321 seconds for the RAID drive.

I found the links containing the information you asked for, please take the time to read them before responding.

Good morning! You once again have attempted to show SSD's vs. HDD's. We are talking about SSD's in RAID 0 vs. an SSD by itself. You are quite fond of IOPS. Tell me, what increase in IOPS on a system with two SSD's in RAID 0 will allow me to run a specific program one second faster? Oh, that's right, you can't because it's a subsystem that depends greatly on MB and CPU. Once again (3rd or 4th time?) unless you are benchmarking, there is currently no real world advantage to running SSD's in RAID 0 versus an SSD by itself.

 

[rgallant]

I said, "if you let a computer knowledgeable friend use your computer not knowing it had SSD's in RAID 0 vs. a SSD boot disk."I did not say if they could notice the difference between an SSD and a HDD.

You STILL have not shown me real world "appreciable" advantages to putting SSD's into RAID 0. You can drool over IOPS all you like but, can you provide a link or article showing real life noticeable improvements in specific applications?

Engineers and techno geeks love to report on what they know but, will jump through hoops and turn themselves inside out before saying, I don't know or, I don't know how to measure that. Worse yet, are those who convince themselves that those things which either cannot be measured or, are difficult to measure, are unimportant.

You STILL have not shown me real world "appreciable" advantages to putting SSD's into RAID 0.

-lets see ,I can use 200gb on my 2x120 g2's now,before raid I could only use 100gb for the os so is that real world or in my mind ?

 

[HendrixFan]

Good morning! You once again have attempted to show SSD's vs. HDD's. We are talking about SSD's in RAID 0 vs. an SSD by itself. You are quite fond of IOPS. Tell me, what increase in IOPS on a system with two SSD's in RAID 0 will allow me to run a specific program one second faster? Oh, that's right, you can't because it's a subsystem that depends greatly on MB and CPU. Once again (3rd or 4th time?) unless you are benchmarking, there is currently no real world advantage to running SSD's in RAID 0 versus an SSD by itself.

The link showed a speedup comparing SSD in RAID-0 on a PCI-e card versus the same SSD non RAID. It is exactly what you asked for, and it does exactly what you claimed it does not do, which is shave quite a bit of time off of storage bound tasks.

 

[mojo8472]

Decided to go for a single 120GB Vertex 2E and 3x 500GB WD Blacks in RAID0

For anyone who's interested, here's what the HDD bench looks like