noob raid question - simple

[General_Disturbance]

I have seen people talk as if raid0 gives you a greater chance of losing all your data...why is that? What makes raid0 hard on the disks?

If the chance of failure for 2 disks in raid0 is still the same as one of the disks failing individually, that is FINE by me! I've used single disk setups all my life and have never had one fail, ever.

So is the simple idea that if you have 2 disks instead of one, you have twice the chance of a disk failure?? Is that it? If so I'm not woried and I'm gonna raid0, because I do a lot of HDD read-time limited work. I backup data on portable HDD anyway.
Thanks!

 

[kshipper]

Your last paragraph is correct. RAID 0 splits the data over two drives. You lose all your data if either hard drive fails. That's why it's more risky because you have increased your chances of hard drive failure. The benefit of a RAID 0 is speed. It really increases the read and write speed by a big margin, expecially if you use WD Raptors.

I have had a RAID 0 fail on my main machine and although I do backup it was still a pain to get back up and running. So many programs/games to re-install <ugh>.

 

[mrmez]

After reading this: http://www.tomshardware.com/2007/10/17/parallel_processing/

I dont think id bother with raid
:/

 

[General_Disturbance]

After reading this: http://www.tomshardware.com/2007/10/17/parallel_processing/

I dont think id bother with raid
:/

Actually it was that article that MADE me want to go raid. pg 7 second graph down shows HDD performance. HUGE improvement if you do lots of HDD intensive work, which I do.

 

[General_Disturbance]

So a 4 disk raid10 with 4 250GB HDD's would give me 500GB total space, with the benefits of both raid0 and raid1, but perhaps not so much extra speed when writing, only when reading?

I'll go use the search if noone replies to this...but just looking for a quick "yae thats the general idea" response if you got it.
Thanks

edit: nevermind...http://www.tomshardware.com/2007/07/02/raid_scaling_charts/page8.html#experiences

 

[4745454b]

There is another "weak link" that people almost never talk about, driver failure. When people come to the forum with an array problem, I've never seen a dead harddrive as the cause of it. Its usually because they were messing around with the settings and they did something stupid, OR because the computer had something happen to it and the driver for the array got corrupted. If the computer gets a power surge/brownout, its possible for the data that holds the array info to get corrupted. From what I've seen on the net, this usually requires a format and reinstall.

As for page seven, are we looking at the same thing? Sure, the synthetic benchmark of PCMark 05 harddrive shows an extra ~3500 "marks". But what does that get you? The Cinebench scores are nearly identical. Look at the scores for real programs. Page six shows games, notice the first result, the one for Quake 4. The single disk provides the fastest FPS, NOT AID0! If you look at the three game results, none of them show AID0 to be significantly faster then a single disk. Page 8 also shows there to be NO difference when using winrar or 3D Studio Max 8. Page 9 also shows basically no difference when encoding things. There is a small 3 second difference if you encode with Divx/xvid, but Lame and the H.264 encoder are all tied. Synthetic apps showed a "large" 3500point advantage for AID0, but real world programs showed NONE! Rather then spend your money on extra harddrives, you are better off spending it on faster CPUs and more ram, where the speed would be better felt.

 

[jt001]

There is another "weak link" that people almost never talk about, driver failure. When people come to the forum with an array problem, I've never seen a dead harddrive as the cause of it. Its usually because they were messing around with the settings and they did something stupid, OR because the computer had something happen to it and the driver for the array got corrupted. If the computer gets a power surge/brownout, its possible for the data that holds the array info to get corrupted. From what I've seen on the net, this usually requires a format and reinstall.

As for page seven, are we looking at the same thing? Sure, the synthetic benchmark of PCMark 05 harddrive shows an extra ~3500 "marks". But what does that get you? The Cinebench scores are nearly identical. Look at the scores for real programs. Page six shows games, notice the first result, the one for Quake 4. The single disk provides the fastest FPS, NOT AID0! If you look at the three game results, none of them show AID0 to be significantly faster then a single disk. Page 8 also shows there to be NO difference when using winrar or 3D Studio Max 8. Page 9 also shows basically no difference when encoding things. There is a small 3 second difference if you encode with Divx/xvid, but Lame and the H.264 encoder are all tied. Synthetic apps showed a "large" 3500point advantage for AID0, but real world programs showed NONE! Rather then spend your money on extra harddrives, you are better off spending it on faster CPUs and more ram, where the speed would be better felt.

Quoted for truth. I've been using RAID0(and it's "RAID0" not "AID0", even if AID makes more sense that's not the correct terminology) for years, and any data loss has been because of the controller messing up. I've lost data 3 times now that I can remember, and never from a drive failure. So many things besides a drive failure can upset a RAID0 array. Basically, the only time you should ever use RAID0 is if you have a backup, that's it.

As far as it being worth it or not, I've had raid on my secondary "general usage" machine for about a year. I eventually figured it wasn't really benefiting me and only making more noise in my quiet system, so I went back to a single drive. I can say without having RAID0 now my system definitely feels much slower, even just for general usage. I would however recommend a fast drive before RAID0, because low seek times help a lot more IMO than STR, which a Raptor for example would also give more of.

 

[mrmez]

Actually it was that article that MADE me want to go raid. pg 7 second graph down shows HDD performance. HUGE improvement if you do lots of HDD intensive work, which I do.

Problem i have with that... its a SYNTHETIC benchmark.
Its all theoretical. In theory, communism works. In theory.
THG could/did not provide any realworld evidence to demonstrate that raid 0 is faster than a single drive.
Over all game/app/video/synth benches the only one proving faster was PCMark '05 HDD bench.

That article is the least convincing evidence ive seen for raid 0.

 

[General_Disturbance]

Problem i have with that... its a SYNTHETIC benchmark.
Its all theoretical. In theory, communism works. In theory.
THG could/did not provide any realworld evidence to demonstrate that raid 0 is faster than a single drive.
Over all game/app/video/synth benches the only one proving faster was PCMark '05 HDD bench.

That article is the least convincing evidence ive seen for raid 0.

No no no...you guys are missing the point totally and utterly. There is a difference between RAM limited work (games, encoding, etc) and HDD limited work (reading data and writing data). It is the reading of data which is sped up by raid0, not fps or cpu performance etc.
I do work where all I do is read data off the HDD. If that can be sped up by raid0 by a factor of 1.7 or so, that is a huge time saver. I am not expecting any software to RUN faster, I am simply expecting data to READ faster. This is what raid0 does. So my statements stand. This is a general property of raid0, faster read times, that doesn't need to be argued.

PS: read the comments in that article on raid. Many people are pointing out that the review and benches waere flawed because noone expects raid0 to improve anything other than HDDread times, but they're applying it to game fps's and stuff.

 

[onestar]

Raid 0 has never seemed an appealing option for me. Raid 5, on the other hand, has always seemed to be like an automatic backup that you rarely have to think about. The read performance is almost the same as Raid 0, however wites are slower. A failed disk....no problem.

 

[General_Disturbance]

Raid 0 has never seemed an appealing option for me. Raid 5, on the other hand, has always seemed to be like an automatic backup that you rarely have to think about. The read performance is almost the same as Raid 0, however wites are slower. A failed disk....no problem.

Writes are slower than raid0 but faster than single disk, or slower than both?

 

[rubix_1011]

I have had a RAID 0 fail on my main machine and although I do backup it was still a pain to get back up and running. So many programs/games to re-install <ugh>.

Maybe you should consider an imaging solution such as Acronis or Ghost. Create a fresh install with updates and software installs, create an image and store it away. When you need to restore, you can have a *fresh* install in about 20-30 mins, depending on machine specs and image size. Just a suggestion instead of taking 2-3 hours of installs and updates...

 

[4745454b]

No no no...you guys are missing the point totally and utterly. There is a difference between RAM limited work (games, encoding, etc) and HDD limited work (reading data and writing data). It is the reading of data which is sped up by raid0, not fps or cpu performance etc.

The problem is none of the benchmarks that they showed in that article showed ANY improvement with AID0. None. 3seconds faster if you encode with Dvix or Xvid, but I don't consider that worth the extra expense of the additional drives, or increased risk of data loss.

I also laugh a bit because you came here asking a question, but haven't let us in on what you need the array to do. All you keep saying is that you have a task that requires a lot of reading from drives. What do you need the array for? AID0 does have its uses, most people don't need it.

 

[General_Disturbance]

you came here asking a question, but haven't let us in on what you need the array to do. All you keep saying is that you have a task that requires a lot of reading from drives.

Uuummmm, I think you answered the question.
I have tasks that require a lot of reading of data from HDD. My processor waits around for data to get to it from the HDD, if I can speed that up with raid0, which I can, great!

 

[General_Disturbance]

The problem is none of the benchmarks that they showed in that article showed ANY improvement with AID0. None. 3seconds faster if you encode with Dvix or Xvid, but I don't consider that worth the extra expense of the additional drives, or increased risk of data loss.

Again, those are CPU and RAM limited, and have nothing to do with HDD read times. The 3 seconds came from the VERY beginning of the encode because the data gets read into RAM first, THEN the encode starts. But if all you do is read off data and your computations are faster than the read time, you are HDD limited and it is good to RAID, and this is what the HDD performance chart on page 7 of that article showed.

For anyone who understands what raid is and what it does, all those charts make perfect sense and are a nice confirmation of theory. raid0 can speed up read time by almost a factor of 2, but doesn't affect CPU or RAM limited processes.

I have a feeling you're trolling...you have a lot more posts than me and should know this stuff.
Cheers!

 

[4745454b]

You still avoided answering the question. What task are you trying to do? All you say is that you have these "tasks" that require reads from harddrives. Lets try this one more time. If you post a response, just answer this ONE LITTLE QUESTION, what are you trying to do/what is this "task"? I get the feeling that you are either an AID0 diehard, or you already have it and are trying to make yourself feel better about having it.

Reading the one link that I've already responded to, where do you see increases of "almost a factor of 2"? Real applications showed no decrease in times. Are you running only synthetic apps?

As for post counts, they pretty much mean nothing. I could make a new name and have 5000 posts in a week if I wanted to. What is far more important is that I'm not the only one saying no to AID0. But seeing as you brought it up, seeing as I do have more posts then you, why are you so sure that your right and I'm just a troll? (and I do know my stuff, I'm been preaching against AID0 for a long time.)

 

[General_Disturbance]

Well the point is, I read data from the HDD a LOT. If I can make HDD read time faster, that is great. Almost a factor of 2 comes from the chart, more likley around 1.6-7 which is "almost" 2. Data is all backed up on portable Lacey terrabyte drives and is only put on my drives when I need to analyze it - I'm not worried about disk failure in the slightest.

I'm running my own software which reads hundreds of thousands of files of image data totaling into the terrabyte range. Each file is 512kB and needs to have some algorithms applied to it to extract a small amount of data from each. I pull only a few bytes of data from each 512kB file, so the results are stored in memory until the entire is batch is done processing - I don't do any writing during the process.

This is all about HDD read times...you need to get that straight. My software analyzes each 512kB file in RAM much faster than they are read off the HDD, so if I can speed up the read time I will have an overall reduction in processing time, which adds up to 2 or 3 more batches reduced per day.

 

[4745454b]

My signature will take over, this is my last post (to you anyways....). When you wrote backed up on Lacey Terrabyte drives, and into the terrabyte range, I have an idea (finally!) of what you are trying to do. If you choose be an @$$ about everything, then I'm not going to help you. Do you have any idea how many people come onto this forum saying they need AID0 to load a 1GB map in a game faster? Probably not seeing as your post count is under 60. Rather then go off on me (a person who might have some good ideas...), you could have just answered the question.

Seeing as you think I know nothing, if your file sizes are 512kB, you probably should be looking at drives with very low seek times, unless the data is sequential. (but what do I know, I'm a troll.)

 

[General_Disturbance]

Yah cheers man didn't mean to offend. Didn't realize you were such AID0 h8r Just teasing. I thought it was enough to say I di HDD limited work, but I appreicate that you needed more info.
Peace!