Interesting RAID 0 claims/accusations. Validity at question?

[GeForceX]

My friend and I recently argued over RAID 0. My stance is that RAID 0 isn't worth it. His stance is that RAID 0 is worth it.

He made a couple of accusations (or reasons) of why RAID 0 might fail:

• RAID 0 crashes because it doesn't work well with non-dual core CPU's because of the stress on the CPU.
• RAID 0 does not work well with Pentium 4's Hyper-Threading technology.
• Hyper-Threading wasn't good in the first place.

I told him, "My friend used RAID 0 while editing his Quake 3 movie. From what I could remember, he used a capable system: Pentium 4 3.2 GHz, 1 GB RAM, some Asus motherboard, 2x 250 GB Western Digital HDD's in RAID 0, GeForce 6800 GT running on Windows XP Pro. His RAID 0 array failed and he lost just about everything on it. The performance was not anything special as he told me." He replies saying that it crashed becaused he used a single core CPU which can't handle multi-tasking well.

I said to him, "RAID 0 is pointless. If you really wanted RAID 0, then go for RAID 5 - at least that protects your data." He says in return, "RAID 5 is pointless for everyone."

I told him that I disagreed on many of the claims he made. I asked him where he gets such information. He said, "Use your common sense - Google!" I said, "So you believe everything you read on the intarweb?"

He also claims that RAID 1 slows down things. Not sure if that's true.

Is there any validity behind his absurd (in my belief) claims?

-J.

 

[Lazn_Work]

RAID 0 and RAID 1 both have minimal impact on the CPU and single vs dual CPU would not make any difference. RAID 1 does not significantly slow things down vs single drive.

Hyperthreading is usefull for desktop multitasking on single CPU P4 machines. It is not good for server aps, and dual core / dual CPU machines don't need it. But it has no affect on RAID one way or the other.

As for if RAID 0 usefullness, it depends on what you are doing. Video editing etc. can get a very good boost from it. (anything with large linear-sequential writes/reads) For most people however, the added risk of data loss and the minimal performance increase in non-linear use patterns is not worth it.

RAID 5 has poor write performance, so you have to take that into account when deciding if you want to use it.

http://arstechnica.com/paedia/r/raid-1.html

==>Lazn

 

[GeForceX]

Isn't RAID 5 literally a conjunction of RAID 0 and 1? Best of both worlds as they call it. how come it has suffering write performance?

What's the alternative RAID for best performance and fault protection?

-J.

 

[mwarps]

My friend and I recently argued over RAID 0. My stance is that RAID 0 isn't worth it. His stance is that RAID 0 is worth it.

He made a couple of accusations (or reasons) of why RAID 0 might fail:

• RAID 0 crashes because it doesn't work well with non-dual core CPU's because of the stress on the CPU.
• RAID 0 does not work well with Pentium 4's Hyper-Threading technology.
• Hyper-Threading wasn't good in the first place.

I told him, "My friend used RAID 0 while editing his Quake 3 movie. From what I could remember, he used a capable system: Pentium 4 3.2 GHz, 1 GB RAM, some Asus motherboard, 2x 250 GB Western Digital HDD's in RAID 0, GeForce 6800 GT running on Windows XP Pro. His RAID 0 array failed and he lost just about everything on it. The performance was not anything special as he told me." He replies saying that it crashed becaused he used a single core CPU which can't handle multi-tasking well.

I said to him, "RAID 0 is pointless. If you really wanted RAID 0, then go for RAID 5 - at least that protects your data." He says in return, "RAID 5 is pointless for everyone."

I told him that I disagreed on many of the claims he made. I asked him where he gets such information. He said, "Use your common sense - Google!" I said, "So you believe everything you read on the intarweb?"

He also claims that RAID 1 slows down things. Not sure if that's true.

Is there any validity behind his absurd (in my belief) claims?

-J.

Your friend is a walking, talking troll. The pure amount of misinformation in those three talking points was enough to make me laugh outloud.

 

[DougLite]

Where do I start?

Well, here is some generic stuff: http://www.hardforum.com/showthread.php?t=952998

As for the reliability of RAID-0 (in terms of mechanical disk reliability) it isn't as much of an issue as I and others would like to make of it. RAID-0, RAID-1, or SLED, if you are nto backing up, you will be sorry. However, it introduces additional drivers, puts the data at the mercy of a single file system, and cannot be readily moved to a different system in the case of a catastrophic failure (PSU, etc).

As for the relationship between the CPU that the system has and RAID reliability...that argument should be dismissed as making no sense at all. CPU resources and physical disk resources are entirely different. Different applications make different use of various resources. A 1M SuperPi run will max out your CPU and use essentially no disk time, while a disk to disk file transfer will max out the drive and the CPU will be chilling out at <10% utilization. Other tasks may make use of both, like video encoding (max out CPU and moderate load on disk) or game level loading (max out disk, moderate CPU load).

RAID-1, presuming that there are no driver or BIOS level bottlenecks, delivers the exact same speed as SLED under single I/O loads. Concurrent read speeds can be doubled, as each disk can serivce read requests independently.

As for RAID-5, it has more in common with RAID-0 than RAID-1 in terms of mechanics. It is like RAID-1 in that it allows the data on the array to remain intact after a disk failure, but that is a conceptual similarity, not a mechanical similarity.

 

[BillLeeLee]

Isn't RAID 5 literally a conjunction of RAID 0 and 1? Best of both worlds as they call it. how come it has suffering write performance?

What's the alternative RAID for best performance and fault protection?

-J.

Well, first off, your friend is ill-informed. You can find information, such as via storagereview.com's FAQ database, that will dispel the RAID 0 performance myth.

True enough, RAID 0 does have its uses - it has very good sequential reading/writing capabilities, and as such is useful in select applications as Lazn_Work has described, such as video editing where you frequently work on very large files. It is also used for applications such as scratch disks. However, as Lazn_Work has also said, it is not very useful at all for random reads and writes, and normal desktop usage patterns are much more oriented on random disk access rather than sequential access.

It is up to each person to decide whether they want RAID0 or not, but in most cases, it's probably unnecessary and you will probably not notice any significant gains in performance.

RAID 5 is not a combination of RAID 1 and 0, that's what RAID 0+1 and RAID 1+0 are. RAID 5 has bad random write performance because of the overhead of writing parity information to all drives in the array. As such, RAID5 is not supposed to be used for random write-intensive applications, and is more suitable for file servers and more business-oriented use.

Now if you want RAID for regular desktop usage, RAID 1+0 is probably one of the better choices. RAID 10 (shorthand for 1+0) provides good fault tolerance, and provides overall good performance for reading/writing. However, it does have a higher initial cost because RAID 1+0 does require at least 4 hard drives, though if you are not interested in having a lot of space (unlike some people...like me ), you could build a relatively cheap RAID 1+0 array with good budget drives like the smaller Hitachi 7Ks.

 

[GeForceX]

I truly appreciate all the clear information to all that posted: BillLeeLee, DougLite, Mwarps (for the funny comment), and Lazn_work. I have read the ArsTechnia article. A lot of technical explanations of how it all works out. So, RAID 69 - is that even real?

Thanks, guys.

-J.