RAID 1 Faster > Single Disk?

[djmob7]

I've recently read the following from wikipedia?

'Mirrored set without parity' or 'Mirroring'. Provides fault tolerance from disk errors and failure of all but one of the drives. Increased read performance occurs when using a multi-threaded operating system that supports split seeks, very small performance reduction when writing. Array continues to operate so long as at least one drive is functioning. Using RAID 1 with a separate controller for each disk is sometimes called duplexing.

In particular it suggests that RAID 1 is faster than a single disk at reads and maybe a slight penalty for writes. Does anyone know if this is actually true for Vista? Does the raid controller matter in this scenario?

This is what I just read from Anandtech forums, which seems to say the opposite.

Unless you need redundancy, it is much better to simply have two unRAIDed hard drives and split data between them (OS on one, games on the other) as benchmarks show it to be much faster.

So now I'm a little confused. I like the idea of raid 1 to protect against a single hard drive failure. If RAID 1 also descreases read times, that seems really nice too!

 

[RebateMonger]

That's an interesting question that I've always wondered about.

This guy, who apparently runs a lot of host servers in RAID 1 mode, says he couldn't find anybody at 3Ware that had even HEARD of "split seeks", and couldn't find any indication that the 3Ware controllers were providing it.

A search of other RAID-related sites didn't provide an answer for me. The general answer seems to be: RAID 1 can "potentially" speed up read operations if split seeks are enabled. Otherwise, RAID 1 Reads are the same as for a single disk. And RAID 1 Writes are considered to be a "bit" slower than a single disk, but not nearly as slow as RAID 5.

I can't claim to have ever tested those assumptions. I don't have any super-busy database servers, and file servers across a network are pretty much network-limited anyway, at least for 10/100/1000 Mbps connections.

 

[Comdrpopnfresh]

I've read in the past that RAID-1 offers faster reading performance. I'd think performance gain would vary on what type of controller is used, but if one support simultaneous reads, there should be some performance gain there. So from what I've come across + read, what you present seems to be backed up some.

Are you asking in general, or is there some sort of storage size you're contemplating mirroring? If the size isn't too large, and you're seeking performance and redundancy.. you could use a raid-5 array

 

[Nothinman]

If you're using hardware then it depends on if the controller is smart enough to do the split seeks, if it's a cheapo onboard controller then it'll be the driver's job and I wouldn't get my hopes up. I'm pretty sure Linux software RAID is smart enough but I have no idea about Windows Server software RAID.

 

[RebateMonger]

Originally posted by: Nothinman
...I have no idea about Windows Server software RAID.

Hard to tell, since a search of Microsoft's entire site gave near-zero results for "split seek(s)". :frown:

Areca's and 3Ware's sites have no "Search" function, and Adaptec's site returns zero results for a search for "Split Seek".

 

[Nothinman]

That's actually the first time I've heard that term, I believe the Linux dm RAID1 code calls it balancing reads.

 

[djmob7]

Are you asking in general, or is there some sort of storage size you're contemplating mirroring? If the size isn't too large, and you're seeking performance and redundancy.. you could use a raid-5 array

I'm just asking in general. Currently I have a raid 1 setup for my personal computer, but it's only for personal use (nothing important). Once upon a time I though I was going to store alot of media files on it, but I haven't really done that.

I'm in process of researching my new system. I will be building it and since hard drives are so cheap I could easily buy multiple drives. I don't want raid 0, as the performance gains don't outweigh the 50% increased chance of data loss IMO.