raid 0 vs raid 0 + 1 whch is better.

[Dirt]

I don't see how you get any better performance at all on raid-1. You have a primary drive and a secondary (mirror) drive. How is the second drive going to help the speed of accessing data from the primary drive?

You can read a large file from both drives at once. Whether this helps would depend on the implementation, but I've read that in practice it often does.

You could also read a small file from both drives and use the one that arrives first, which will vary some.

[bob]

I don't see how you get any better performance at all on raid-1. You have a primary drive and a secondary (mirror) drive. How is the second drive going to help the speed of accessing data from the primary drive?

You can read a large file from both drives at once. Whether this helps would depend on the implementation, but I've read that in practice it often does.

You could also read a small file from both drives and use the one that arrives first, which will vary some.

And which will chew up double the bandwidth. Most raid-1 controllers I have worked with only mirrored to the second drive, they did not read from both.

[Dirt]

I don't see how you get any better performance at all on raid-1. You have a primary drive and a secondary (mirror) drive. How is the second drive going to help the speed of accessing data from the primary drive?

You can read a large file from both drives at once. Whether this helps would depend on the implementation, but I've read that in practice it often does.

You could also read a small file from both drives and use the one that arrives first, which will vary some.

And which will chew up double the bandwidth. Most raid-1 controllers I have worked with only mirrored to the second drive, they did not read from both.

The small file example was meant to be theoretical, not practical.

For large files using more bandwidth is what you would want. It's not like you're just wasting it: Use it or lose it. As for whether the improvement exists in practice I have no direct knowledge. I know I have read that raid-1 can speed up reads by using both drives. If the controllers can't actually do it I'm disappointed.

[BBauer]

Yes, it's mostly the access time. For SCSI devices you may have 3.6 ms,
for IDE 8.5 ms, but I have seen also 5 ms. For memory sticks you will have about 1 ms which is even better than SCSI.
Using tablebase together with a slow IDE disk will only hurt your program,
but I think that is what many do.
Kind regards
Bernhard

[M ANSARI]

I did some tests some years back with different RAID arrays and setups, and if I remember correctly I had a performance increase in RAID 1 over simple HDD. It could be that this was irrelevant since the increase was very small and maybe not statistically relevant. But it seems like if you have the same data on 2 drives you should be able to retrieve it a little quicker than a single drive. I lost interest in RAID for desktop computers when I realized that a WD Raptor drive running at 10,000 rpm can outperform a 2 HDD RAID 0 setup using 7200 rpm drives.

I think what people have to realize that not all RAID performance is the same ... different RAID controllers can differ dramatically in performance. Most onboard RAID controllers (available in most motherboards today) cannot touch expensive specialized RAID cards performance wise.

Here is a full description of different RAID setups

http://www.integratedsolutions.org/raid_ov.htm

[bob]

I don't see how you get any better performance at all on raid-1. You have a primary drive and a secondary (mirror) drive. How is the second drive going to help the speed of accessing data from the primary drive?

You can read a large file from both drives at once. Whether this helps would depend on the implementation, but I've read that in practice it often does.

You could also read a small file from both drives and use the one that arrives first, which will vary some.

And which will chew up double the bandwidth. Most raid-1 controllers I have worked with only mirrored to the second drive, they did not read from both.

The small file example was meant to be theoretical, not practical.

For large files using more bandwidth is what you would want. It's not like you're just wasting it: Use it or lose it. As for whether the improvement exists in practice I have no direct knowledge. I know I have read that raid-1 can speed up reads by using both drives. If the controllers can't actually do it I'm disappointed.

I'm talking about _wasted_ bandwidth, which reading the same data twice would do (the comment about read both and use the one that arrives first).

[Dirt]

I don't see how you get any better performance at all on raid-1. You have a primary drive and a secondary (mirror) drive. How is the second drive going to help the speed of accessing data from the primary drive?

You can read a large file from both drives at once. Whether this helps would depend on the implementation, but I've read that in practice it often does.

You could also read a small file from both drives and use the one that arrives first, which will vary some.

And which will chew up double the bandwidth. Most raid-1 controllers I have worked with only mirrored to the second drive, they did not read from both.

The small file example was meant to be theoretical, not practical.

For large files using more bandwidth is what you would want. It's not like you're just wasting it: Use it or lose it. As for whether the improvement exists in practice I have no direct knowledge. I know I have read that raid-1 can speed up reads by using both drives. If the controllers can't actually do it I'm disappointed.

I'm talking about _wasted_ bandwidth, which reading the same data twice would do (the comment about read both and use the one that arrives first).

As I said, that wasn't meant to be practical. I put it that way because it makes the idea easy to understand. A useful implementation would probably have to know where the heads and data were and decide ahead of time which drive would be able to get the data first. I've never heard of this being done but I see no reason why it couldn't. Which doesn't mean that it is necessarily worth doing, of course.

Checking Wikipedia, it explains that raid-1 can speed up reading, but that not all implementations do.