RAID-0 proved ineffective at boosting desktop application/game performance

[Lord of Shadows]

Could some one perform the same test, but with varying stripe sizes as well?

Also, what stripe size was used in his benchmarks?

Edit: Also when you think about it the test is flawed, a game loads large textures and various small files (models, scripts, etc). Raid0 cant win both at the same time, so the performance benefit is canceled by the performance hit.

 

[masher]

Also when you think about it the test is flawed...

This is just too funny. Running a real-world game tested to measure real-world game performance is "flawed"?

a game loads large textures and various small files (models, scripts, etc). Raid0 cant win both at the same time...

I don't know many games that try to load more than one file at a time. That takes either multithreaded or asychronous i/o. Nearly always, they're loaded in serial fashion....one right after the other.

 

[Lord of Shadows]

This is just too funny. Running a real-world game tested to measure real-world game performance is "flawed"?

I don't know many games that try to load more than one file at a time. That takes either multithreaded or asychronous i/o. Nearly always, they're loaded in serial fashion....one right after the other.

I said a game loads both small files and large files, I didnt mean at the same time, raid0 performs best with either a large stripe size to move large blocks of data around or with a small stripe size to move smaller files around. A middle ground tries to minimize overhead, but doesnt perform either task a well as it could.

A game is not exactly the best way to benchmark drive performance, unless the goal of your benchmarking is to make game X load faster. I can understand the need to get "real world" results, but with a game you dont have a solid idea of whats being done when. For all I know it could be waiting on information from another device. Whereas if you time a file transfer you know exactly whats going on.

And to elaborate on my previous edit, a game has both small and large files, which will only perform well (assuming equal distribution) on the middleground scenario.

 

[DougLite]

The access patterns of games are extremely difficult to analyze. MikeBlas noted extremely poorly coded disk access, with odd offsets and actuator repositionings between reads within the same file, when at least at first glance it would make sense to simply do a linear read of the entire file and then work with it in much faster physical memory.

WoW is even wierder...it stores its data in 8 large files (one of them is right at 1GB), yet SR's results suggest that it is seek performance bound, as the MAU remains out of reach fform even the swiftest ATA drives. Obviously, there are some unnecessary seeks involved there.

Part of this comes from a time when physical memory was a precious commodity. Many systems still only have 256MB or 512MB of RAM, so simply dumping everything to memory is not an option. Even if it is an option on a modern top flight gaming rig with 2GB of RAM, game developers simply can't develop a second code base for a fairly small portion of the game market.

Despite my study of desktop access patterns, I still am far from understanding them.

 

[unhappy_mage]

I did an analysis of Guild Wars, a MMORPG like WoW, and found that it was doing (mostly) more intelligent reads from its single 1GB data file. When loading a map, there's a short delay while it checks that there are no updates to that map with the server, and then does 64k sequential reads to load the map. Once it's loaded, it reads small (~2k) chunks out-of-order for characters' items, clothes, etc., but it's already drawing the map at that point, the character can move around, and the game is functional. Anyone have WoW want to try?

 

[Asian Dub Foundation]

raid 0 for me is not worth it over file's integrity

 

[djnes]

raid 0 for me is not worth it over file's integrity

You sir, just hit the nail on the head with your virtual hammer.

 

[PaHick]

No offense, but this post is a bunch of senseless banter. You're hiding behind some veil of non-information (not a real world, but deal with it) instead of ponying up some reality centered results.

Yes, there are tons of variables, too many to list. Why don't you benchmark your own system with a RAID0 array, since you're so knowledgeable.

For the record, most people would actually evaluate that kind of performance with draconian methods (stopwatch, lol) rather than just come out with some blanket statement of "Oh, there's tons of variables and because neither one of us can predict all the outcome, I'm right and you'r wrong."

Please. Just stop.

Ive tested my system and thats why I run RAID0 on my machine, and our families file server. Only once ive had a failure of an array and that was when I upgraded our file server. I was transfering files from one to the other after having backed up everything to the system I use strickly for backups and an old Seagate died on me. As for your stopwatch comment, sometimes you have to resort to such methods. Maybe not the most accurate, but still damn close. On my machine access times slowed an average of 1.7ms . On my server access times slowed 4.2ms on average , both tested with HDTach. My new server has an uptime of a little over 3 months, nothing wrong yet. The old setup at its longest ran for 8 months before I rebooted. Again no problems. Boot time on my machine isnt any greater than when I ran a single drive. Overall the system runs smoother opening apps. and such. Re-doing some old VHS to DVD's really shows off the system. I dont game anymore so I havent tested games, not my thing.

You can choose to believe what you want, but facts are facts. Do some testing of your own. Look at studies and reviews. Forget what non supporters say and look at the charts. RAID0 shows speed gains in almost every one over a single drive. The only blanket statements being made are "RAID0 is dead", "placebo effect" and SR and Anandtechs statements "there is no place for RAID0 in a desktop system". When you repeat something over and over, people eventually believe it. Now its time to stop listening and actually look at the numbers.RAID0 has a place, and its not only for large file transfers.

 

[DougLite]

But at what cost?

 

[PaHick]

But at what cost?

As in price? Reliability? Im pretty certain you meant reliability. Modern HD's are more reliable today than they once were. Your always going to have a failure somewhere along the line, but you'll have to agree reliabilty has risen. HD failures or mysterious array failures shouldnt be a problem anyhow, as there should ALWAYS be a solid backup plan in use.

 

[feigned]

Ive tested my system and thats why I run RAID0 on my machine, and our families file server. Only once ive had a failure of an array and that was when I upgraded our file server. I was transfering files from one to the other after having backed up everything to the system I use strickly for backups and an old Seagate died on me. As for your stopwatch comment, sometimes you have to resort to such methods. Maybe not the most accurate, but still damn close. On my machine access times slowed an average of 1.7ms . On my server access times slowed 4.2ms on average , both tested with HDTach. My new server has an uptime of a little over 3 months, nothing wrong yet. The old setup at its longest ran for 8 months before I rebooted. Again no problems. Boot time on my machine isnt any greater than when I ran a single drive. Overall the system runs smoother opening apps. and such.

There's still the possibility of the array failing, but with backups that's not an issue.

Re-doing some old VHS to DVD's really shows off the system. I dont game anymore so I havent tested games, not my thing.

Uh, duh? Of course it's going to help with massive amounts of data that needs to be processed...it's kind of what RAID0 is "made for." As for the gaming comment, what you do with RAID0 and what marketing/advertising departments want you to believe are two totally different situations.

You can choose to believe what you want, but facts are facts. Do some testing of your own. Look at studies and reviews. Forget what non supporters say and look at the charts. RAID0 shows speed gains in almost every one over a single drive.

I did it for a year, I've earned my wings. It's not worth it for everyday use or for every gamer's wet dream.

The only blanket statements being made are "RAID0 is dead", "placebo effect" and SR and Anandtechs statements "there is no place for RAID0 in a desktop system". When you repeat something over and over, people eventually believe it. Now its time to stop listening and actually look at the numbers.RAID0 has a place, and its not only for large file transfers.

That is the point of RAID0, I don't care what you say. Desktop systems do not need RAID0, and that's a fact.

 

[DougLite]

As in price? Reliability? Im pretty certain you meant reliability. Modern HD's are more reliable today than they once were. Your always going to have a failure somewhere along the line, but you'll have to agree reliabilty has risen. HD failures or mysterious array failures shouldnt be a problem anyhow, as there should ALWAYS be a solid backup plan in use.

What about the drivers? One of the reasons I don't fool with RAID on my gaming rig is that I simply will not install the Nvidia IDE drivers on my rig - they cause too many problems. Admittedly, this isn't as big of an issue on Intel platforms.

What happens if your motherboard dies? If you're running SLED, you simply connect the drive to another machine. Not so with RAID.

What about the extra power consumption? What about the extra heat dissipated into your case? What about the added noise fro mthe extra drive and additional cooling?

What of the added cost in $$$$? For a gamer with a $1500 budget, a pair of WD740GDs will eat up a full 20% of their budget. Double the cost, with all of the above problems, for a performance boost of 10% in some access patterns.

What about scenarios where RAID-0 actually performs _worse_ than SLED? SR notes a _drop_ in performance in WoW and FarCry when running RAID-0, and it only gets worse as you throw more drives at it.

For gamers, RAID-0 is a waste of hard earned cash. Pure and simple.

 

[masher]

Overall the system runs smoother opening apps. and such. Re-doing some old VHS to DVD's really shows off the system...

You can choose to believe what you want, but facts are facts.

But you didn't state any facts. Just your own subjective perception.

 

[PaHick]

I don't care what you say.

And that right there sums up your decision.

What about the drivers? One of the reasons I don't fool with RAID on my gaming rig is that I simply will not install the Nvidia IDE drivers on my rig - they cause too many problems. Admittedly, this isn't as big of an issue on Intel platforms.

What happens if your motherboard dies? If you're running SLED, you simply connect the drive to another machine. Not so with RAID.

Ill admit there are some pitfalls, but your going to have to put up with some very small annoyances everynow and then with any hardware. AMD X2 is an example.

What about the extra power consumption? What about the extra heat dissipated into your case? What about the added noise fro mthe extra drive and additional cooling?

Personally ive never heard of power consumption or heat problems. Look how many of us are running PSU's that IMO are overkill, not to mention the number of case fans in the average system. I really dont think theres a problem.

What of the added cost in $$$$? For a gamer with a $1500 budget, a pair of WD740GDs will eat up a full 20% of their budget. Double the cost, with all of the above problems, for a performance boost of 10% in some access patterns.

Theres where we disagree again. I really dont think most people going RAID0 are running two Raptors. The most popular purchase for how long has been to buy two cheap 80G HD's and run RAID0 to get near Raptor performance. I remember seeing alot of people running Raptors(36 and 74's) in RAID0 and being very happy.

What about scenarios where RAID-0 actually performs _worse_ than SLED? SR notes a _drop_ in performance in WoW and FarCry when running RAID-0, and it only gets worse as you throw more drives at it.

For gamers, RAID-0 is a waste of hard earned cash. Pure and simple.

Two games. Thats it. Look at the Office DriveMark which consists of ...

A capture of Veritest's Business Winstone 2004 suite. Applications include Microsoft's Office XP (Word, Excel, Access, Outlook, and Project), Internet Explorer 6.0, Symantec Antivirus 2002 and Winzip 9.0 executed in a lightly-multitasked manner.

Look at your chart.WD740GDx2 is slower than WD740x4(which happened to tie WD1500ADFD). Those are everday tasks, not WoW and FarCry. Purely from a gamers standpoint, I agree, there are only a handfull that take advantage of RAID0. It goes back to generalizing everyone as a gamer again. You cant.

The truth of the matter is this. The numbers dont lie. You can try to "dispell the myth" all you want. There is no hype. No placebo. Its a fact. And anyone willing to go through the numbers will see that. The benefits may be small, but as I said before, who are you to say whats acceptable? If someone wants to pick up some cheap 7200 rpm drives and throw them in RAID0 to pick up a little speed instead of spending bigtime cash on a 10,000 rpm drive I see no problem with it.

 

[PaHick]

But you didn't state any facts. Just your own subjective perception.

Im not going to post alot of studies and articles to prove what im saying. Hit up the search button, this has really been beat to death in not just here but all over the net. Just look at some of the facts DL posted especially his chart in the first post. And if you dont believe reconstructing DVD's( large file transers) will benefit from RAID0 you need to research RAID0.

 

[masher]

And if you dont believe reconstructing DVD's( large file transers) will benefit from RAID0 you need to research RAID0.

Why put words in my mouth? I said no such thing.

Two games. Thats it. Look at the Office DriveMark which consists of ...

Anyone primarily using Office apps is going to have heavy concerns about data loss...NOT interested in shaving a percent or two off the load time of Word. Raid-0 is a mistake for such users. When I use Office apps on a notebook with a 4200rpm, the machine is still idle 99.99% of the time while I type. And the only thing I'm worried about when I do is the machine or disk crashing, and losing all my work.

Yes the facts are there. Raid-0 is a mistake for Office app users. A mistake for web browsers. A mistake for gamers. It's a slight benefit for those who work regularly with video or other extremely large files. A mistake for everyone else.

Those are the facts. The rest is hype.

 

[PaHick]

Anyone primarily using Office apps is going to have heavy concerns about data loss...NOT interested in shaving a percent or two off the load time of Word. Raid-0 is a mistake for such users. When I use Office apps on a notebook with a 4200rpm, the machine is still idle 99.99% of the time while I type. And the only thing I'm worried about when I do is the machine or disk crashing, and losing all my work.

Yes the facts are there. Raid-0 is a mistake for Office app users. A mistake for web browsers. A mistake for gamers. It's a slight benefit for those who work regularly with video or other extremely large files. A mistake for everyone else.

Those are the facts. The rest is hype.

Are you blind or just foolish? Im guessing the second, as you clearly didnt look at the chart in the first post. Again Office DriveMark consists of..

A capture of Veritest's Business Winstone 2004 suite. Applications include Microsoft's Office XP (Word, Excel, Access, Outlook, and Project), Internet Explorer 6.0, Symantec Antivirus 2002 and Winzip 9.0 executed in a lightly-multitasked manner.

Are you telling me those apps. are STRICKLY for office use? Any virus scan or compression tool will make use of RAID0. Some of the most popular apps on an average system correct?

 

[unhappy_mage]

Ill admit there are some pitfalls, but your going to have to put up with some very small annoyances everynow and then with any hardware. AMD X2 is an example.

Theres where we disagree again. I really dont think most people going RAID0 are running two Raptors. The most popular purchase for how long has been to buy two cheap 80G HD's and run RAID0 to get near Raptor performance. I remember seeing alot of people running Raptors(36 and 74's) in RAID0 and being very happy.

The truth of the matter is this. The numbers dont lie. You can try to "dispell the myth" all you want. There is no hype. No placebo. Its a fact. And anyone willing to go through the numbers will see that. The benefits may be small, but as I said before, who are you to say whats acceptable? If someone wants to pick up some cheap 7200 rpm drives and throw them in RAID0 to pick up a little speed instead of spending bigtime cash on a 10,000 rpm drive I see no problem with it.

First, the reason the amd x2 is a hassle is because it's a new idea. Two cores on one socket means that you have to deal with them in a different way. Raid had been around for 20 years now, and it's *still* adding complexity - it hasn't gotten any better. The x2 problem will be nonexistent within a few years, once the scheduler knows how to deal with it.

Second, running two small drives in raid 0 won't help. The advantage of a Raptor is that it has fast seek times, not its high STR. Seeking takes a lot of time, so if the application does any appreciable amount of seeking, it means the Raptor has the advantage. Most of the time, those people would be better off running two seperate drives, with WIndows and the swapfile on one, and their games on the other. Much of the gain raid 0 gives (see, I can admit there's some) is due to the drives not having to seek as far - since the blocks are interleaved, both drives seek half as far, leading to marginally shorter seeks. This is counteracted by the fact that if anything other than the game files needs to be loaded, both drives have to stop what they're doing and go get it.

And if you dont believe reconstructing DVD's( large file transers) will benefit from RAID0 you need to research RAID0.

Tell me again, just how fast is your DVD drive? Mine's 16x, or about 1.5MB/s*16=24 MB/s... The Maxtor maxline 3 300gb drives I bought do about 60 MB/s, no need for any raid there.

 

[DougLite]

Two games. Thats it.
...

Those two games are reflective of game loading access patterns in general. UT2004 is remarkably similar to FarCry, while games like Doom3 exhibit patterns much like WoW.

Look at your chart.WD740GDx2 is slower than WD740x4(which happened to tie WD1500ADFD). Those are everday tasks, not WoW and FarCry. Purely from a gamers standpoint, I agree, there are only a handfull that take advantage of RAID0. It goes back to generalizing everyone as a gamer again. You cant.

Look at both of the RAID-0 setups delivering significantly slower performance than a single WD740GD in both WoW and FarCry, which once again are exemplary of many other games out there.

The truth of the matter is this. The numbers dont lie. You can try to "dispell the myth" all you want. There is no hype. No placebo. Its a fact. And anyone willing to go through the numbers will see that. The benefits may be small, but as I said before, who are you to say whats acceptable? If someone wants to pick up some cheap 7200 rpm drives and throw them in RAID0 to pick up a little speed instead of spending bigtime cash on a 10,000 rpm drive I see no problem with it.

What I am here to say is that running RAID-0, whether it be with cheap 7200RPM drives, flagship 7200RPM drives, or 10,000RPM drives, is simply not cost effective, for desktop users and gamers. The Penitum Extreme Edition and/or Athlon64 FX lines are faster than regular Pentium D or Athlon64 processors, but most people will tell you that they are simply not worth the price premium. There are people that are willing to pay this premium, but it's not for everyone. The same reasoning applies to RAID-0.

Even your argument that RAID-0 significantly improves linear transfer performance over a superior single drive is flawed...check out this comparison on SR of the four Raptor iterations. The WD740GD delivers transfer performance a full 25% faster than WD360GD, a gap that RAID-0 may be able to close in purely linear transfer work, but certainly will not be able to close under loads that are even slightly random access in nature. Extending this comparison between WD1500ADFD and WD740GD, the WD1500 is 22% faster in linear transfers than the WD740GD, once again a gap that RAID-0 will only be capable of closing under the most favorable of conditions, rare for a typical desktop user/gamer.

 

[Dennis Gordon]

This has been a good thread to follow. Thanks for the ...ummm...lively discussion.

My .02. I ran RAID 0 on my old (still running) office machine. I built it in 2002 with 2 60G Maxtors (a model that came and went in about 2 weeks at Newegg... that should have told me something). Got the RAID up and running. Moved about 3-4 gigs of archives over to the new machine. One of the Maxtors failed the next day. Good bye array. Good bye data (no backup). RMA'ed the bad drive. Rebuilt the array, and it's been running fine ever since (except when the connectors get wobbly and I have to smack the mobo to get 'em to work) The performace was decent, certainly better than running single IDE in 2002. Not incredibly fast, but pretty good. Some latency issues.

My new build is X2, XP 64 bit, all 15K SCSI. Now SCSI is still relatively expensive, but I'm an Ebay hound and put in 3 drives (73, 36, 18) for about $300. Yeah, still pricey, but honestly, there is no comparison IMO to my former rig. Of course and X2 is gonna sizzle over a P4 any day, but I still maintain that fast SCSI is superior to most any ATA config, although the Raptors are pretty tempting.

I have no hard data, other than great HD Tach benches, but although RAID 0 served it's purpose for me for 3 years, I'd rather pay the premium and get better drives.

 

[PaHick]

Even your argument that RAID-0 significantly improves linear transfer performance over a superior single drive is flawed...check out this comparison on SR of the four Raptor iterations. The WD740GD delivers transfer performance a full 25% faster than WD360GD, a gap that RAID-0 may be able to close in purely linear transfer work, but certainly will not be able to close under loads that are even slightly random access in nature. Extending this comparison between WD1500ADFD and WD740GD, the WD1500 is 22% faster in linear transfers than the WD740GD, once again a gap that RAID-0 will only be capable of closing under the most favorable of conditions, rare for a typical desktop user/gamer.

DL, your talking about RAID0, not comparing for example WD740GD to WD1500ADFD, correct? Thread tilte is RAID0 is dead. I did not try to argue running older generation drives in RAID0 outperform new drives. I am arguing RAID0 performance on same drives. Go back to SR's FAQ again to compare the SAME drive single/RAID0, which is the only way to do it right. And in that case, as in all similar tests, there is an increase.

 

[PaHick]

Second, running two small drives in raid 0 won't help. The advantage of a Raptor is that it has fast seek times, not its high STR.

Seek times arent everything, disk density plays an important part also. Fill a single Raptor and put the same 74G of files on two 250G in RAID0. Seeks time will not always outperform placement of files on the disk.

Tell me again, just how fast is your DVD drive? Mine's 16x, or about 1.5MB/s*16=24 MB/s... The Maxtor maxline 3 300gb drives I bought do about 60 MB/s, no need for any raid there.

I undertsand what your saying, but you missed the part where I said VHS>DVD. I have to copy all the old VHS to disk the rework the files to DVD.

 

[masher]

Are you blind or just foolish? Are you telling me those apps. are STRICKLY for office use?

Are you able to read? I said no such thing. What I *am* saying is that users of Office apps are more interested in stability and reliabity over raw speed. Whether they use those apps at home or in an office. Doubling (at least) your chance of data loss in exchange for a trivial speed increase in a program that already spends 99% of its time waiting on you is a poor tradeoff.

Clear now?

The debate in this thread had better not turn personal - see [thread=760666]Rule[/thread] #1. If I see any flaming, disrespect, or personal attacks directed at other members, you may have a a few days to think about what you're going to post next. - DougLite

 

[PaHick]

Are you able to read? I said no such thing. What I *am* saying is that users of Office apps are more interested in stability and reliabity over raw speed. Whether they use those apps at home or in an office. Doubling (at least) your chance of data loss in exchange for a trivial speed increase in a program that already spends 99% of its time waiting on you is a poor tradeoff.

Clear now?

Failures can happen with single disks also, correct? It doesnt matter if single, RAID0, RAID1. If a file is lost prior to saving nothing will help. Once that file is saved a good backup plan takes care of any reliability problem, which IMO is blown our of proportion anyhow.

OH and yes, CLEAR NOW!

 

[unhappy_mage]

Seek times arent everything, disk density plays an important part also. Fill a single Raptor and put the same 74G of files on two 250G in RAID0. Seeks time will not always outperform placement of files on the disk.

True, with 74gb of files it makes a difference. But most (all?) games are smaller than that; say 5gb tops. So take 5gb of files, and then the wide seeks the 74gb drive would have to make with 74gb of files are gone, so the 250s lose that advantage; they both have to make small seeks.

I undertsand what your saying, but you missed the part where I said VHS>DVD. I have to copy all the old VHS to disk the rework the files to DVD.

So what bitrate are you capturing at that you need more than say 10 MB/s? You are capturing in realtime, no? And if you're capturing and then converting, the limitation isn't hard drives, it's CPU, and you'd be better off with two seperate drives - one to capture to and one to convert to. That'd knock seek times out of the picture, and improve performance.

WRT failure; this page has some information on it:

RAID Array MTBF = lowest independent disk MTBF / no. of disks used in array

That sounds good, right? Drives are commonly spec'ed at over a million hours, so 2 drive setups shouldn't fail for 500k hours (57 years!). Unfortunately, that's not what it means.

 

[PaHick]

True, with 74gb of files it makes a difference. But most (all?) games are smaller than that; say 5gb tops. So take 5gb of files, and then the wide seeks the 74gb drive would have to make with 74gb of files are gone, so the 250s lose that advantage; they both have to make small seeks.

If the only thing on your HD was the game, that may be true. But you have OS files,apps. among others.

So what bitrate are you capturing at that you need more than say 10 MB/s? You are capturing in realtime, no? And if you're capturing and then converting, the limitation isn't hard drives, it's CPU, and you'd be better off with two seperate drives - one to capture to and one to convert to. That'd knock seek times out of the picture, and improve performance.

Variable bitrate. Im not just converting, also editing(adding text, rearranging video by date..etc.)

WRT failure; this page has some information on it:

That sounds good, right? Drives are commonly spec'ed at over a million hours, so 2 drive setups shouldn't fail for 500k hours (57 years!). Unfortunately, that's not what it means.

I totally understand what your saying. IMO a good backup plan takes reliability out of the equation. It doesnt matter to me if a single disk fails or an array. Backups are a neccessity.

 

[unhappy_mage]

If the only thing on your HD was the game, that may be true. But you have OS files,apps. among others.

But (hopefully) most of what's being read will be in a small area. As long as you keep the drive defragged, this will be true. And if you have a raid 0 array, you can break it and make the only thing on one of the drives the games.

Variable bitrate. Im not just converting, also editing(adding text, rearranging video by date..etc.)

So what part of that needs raid 0? I mean, if you're capturing 1080i, maybe, but for standard VHS video you shouldn't need much in the way of transfer rates.

 

[PaHick]

So what part of that needs raid 0? I mean, if you're capturing 1080i, maybe, but for standard VHS video you shouldn't need much in the way of transfer rates.

And why not? 8 hours of just one video is alot!

 

[unhappy_mage]

That makes no sense to me.

8 hours of video, at say 6 mbits, is 2.7 GB/hr, or 21.6 GB overall. That's a lot of video, but what are you going to do with it? Even uncompressing it is fairly CPU intensive; adding text etc, and then recompressing is going to be a completely CPU driven task. Suppose you have an uber-fast system that can transcode at 8x realtime. So you need to read 48 mbit/s off disk - that's only 6 MB/s - and write about the same. Raid just doesn't factor into it. If for some reason you decided to create a copy of the video, you could do that faster with two disks seperated; that way you'd get linear reads on one disk and linear writes on the other, rather than basically random reads and writes on both disks. Then raid 0'd be a lot slower than the seperate-disk setup.

 

[PaHick]

That makes no sense to me.

8 hours of video, at say 6 mbits, is 2.7 GB/hr, or 21.6 GB overall. That's a lot of video, but what are you going to do with it? Even uncompressing it is fairly CPU intensive; adding text etc, and then recompressing is going to be a completely CPU driven task. Suppose you have an uber-fast system that can transcode at 8x realtime. So you need to read 48 mbit/s off disk - that's only 6 MB/s - and write about the same. Raid just doesn't factor into it. If for some reason you decided to create a copy of the video, you could do that faster with two disks seperated; that way you'd get linear reads on one disk and linear writes on the other, rather than basically random reads and writes on both disks. Then raid 0'd be a lot slower than the seperate-disk setup.

Believe me I know its CPU intensive, which is why I wanted to jump on Opteron but finances didnt allow. I spend alot of time at 100% CPU usage no doubt. But even small edits see a benefit. Your talking more close to 30G but your figures are pretty much right. We're not talking shaving an hour off time, but even 15 minutes is alot with the limited amout of time I have to spend on it. Conversions take time. And if I screw up thats more time lost..lol. Its not a profession, but most of my family love to record absolutely everything. And im the one chosen to do all the hard work.

 

[masher]

Failures can happen with single disks also, correct? It doesnt matter if single, RAID0, RAID1

Your argument is like saying "accidents can happen at any speed, so I should drive 150mph on rain-slicked roads.

It matters because Raid 0 doubles the chance of any single failure. More than doubles, actually, as you have the possibility of interdrive synch failure in addition to the failure of any single drive.

IMO a good backup plan takes reliability out of the equation.

No they don't. They simply reduce the impact of a failure. If you're backing up nightly and you have a failure...you lose up to a days work of data, plus the time needed to restore and rebuild the array.

And why not? 8 hours of just one video is alot!

As other posters have repeatedly pointed out, Raid-0 doesn't help you here because in most cases you're capture rate-limited or cpu-limited. In the rare case where you aren't, then two independent drives are faster than Raid-0.

 

[PaHick]

No they don't. They simply reduce the impact of a failure. If you're backing up nightly and you have a failure...you lose up to a days work of data, plus the time needed to restore and rebuild the array.

And that would happen on a single disk also. I said before once the file was saved. If it wasnt saved it doesnt matter what disk options you had.

As other posters have repeatedly pointed out, Raid-0 doesn't help you here because in most cases you're capture rate-limited or cpu-limited. In the rare case where you aren't, then two independent drives are faster than Raid-0.

I also said its very CPU limited. Your not just converting, your also copying/transfering large files. I cannot work on a video without first duplicating the file.

 

[unhappy_mage]

I also said its very CPU limited. Your not just converting, your also copying/transfering large files. I cannot work on a video without first duplicating the file.

And two single disks are good at duplicating files, raid isn't. Tell ya what, I'll try it. I have 2 120gb seagate 7200.7s that I have to get the data off of anyways, so I'll try two disks versus one raid, duplicating.

 

[PaHick]

And two single disks are good at duplicating files, raid isn't. Tell ya what, I'll try it. I have 2 120gb seagate 7200.7s that I have to get the data off of anyways, so I'll try two disks versus one raid, duplicating.

Thank you. I will check later, I have company coming.

 

[Vertigo Acid]

Theres where we disagree again. I really dont think most people going RAID0 are running two Raptors. The most popular purchase for how long has been to buy two cheap 80G HD's and run RAID0 to get near Raptor performance. I remember seeing alot of people running Raptors(36 and 74's) in RAID0 and being very happy.

I did not try to argue running older generation drives in RAID0 outperform new drives. I am arguing RAID0 performance on same drives.

Isn't that exactly what you are saying?

 

[masher]

And that [failure] would happen on a single disk also.

But only half as often as it would on a Raid-0 array. Isn't that more important than speeding up MS Word a couple percent?

I also said its very CPU limited. Your not just converting, your also copying/transfering large files...

Both of which are faster on two independent disks than they are on a Raid array. So where's the benefit again?

 

[PaHick]

But only half as often as it would on a Raid-0 array. Isn't that more important than speeding up MS Word a couple percent?

I said it before, but must repeat again apparently...A solid backup plan takes care of any reliability problems you may have. Which I really cant find any tests or studies on. Ive never seen a study testing the theory on an arrays reliability. If someone has a link to a study on array failures I would like to see it.

Both of which are faster on two independent disks than they are on a Raid array. So where's the benefit again?

Theres a huge benefit copying/transfering large file in RAID0. Hell thats what its meant for.

Isn't that exactly what you are saying?

No, what i was saying in the first post some people that could not afford a Raptor were buying two cheap 80's and putting them in RAID0 to try to get better performance, not necessarily to beat a Raptor, but to get better performance than their current drive.

 

[DougLite]

No, what i was saying in the first post some people that could not afford a Raptor were buying two cheap 80's and putting them in RAID0 to try to get better performance, not necessarily to beat a Raptor, but to get better performance than their current drive.

Or you could buy a 160GB drive that delivers better application level performance for 25% less - ~$90 for the 160GB drive vs ~$120 for a pair of 80GB units.

 

[PaHick]

Or you could buy a 160GB drive that delivers better application level performance for 25% less - ~$90 for the 160GB drive vs ~$120 for a pair of 80GB units.

Yes you could do that and be left with the same transfer rates. When going the 2x80 gained maybe 20mb/s. Most times beating the Raptors transfer rates. I didnt see any 160G drives beating Twin Hitachis couple months back. And on the right board or controller access times didnt take a hit if at all.

 

[unhappy_mage]

Still copying stuff off the 120s, sorry. I kept my kernel source trees for several things on them (a raid 1 array), so it's a *lot* of files (135k), and it's moving onto my raid 5 array, which is in PIO mode. Anyways, about half done. I'll be using Dynamic Disks in winXP for the raid.