FRYS SUPER HOT DEAL 120 gig WD 8MB $99!

[MonteLDS]

yes its 179 with a 80 dollar rebate. 1 per customer of course VERY HOT!

got me one so now i have 1 80 and 3 120's (the 120's will be my raid)

[cfitz]

Great deal.

Somewhat off topic, how do you intend to configure those three 120's in your RAID array?

cfitz

[eliminator]

Fry's Rocks!

[BuddhaTB]

MonteLDS, Thanks for the heads up.

However, do you really use all those hard drives?
80GB + (120GBx3) = 440GB!
Now that is a lot of space

[MonteLDS]

Great deal.

Somewhat off topic, how do you intend to configure those three 120's in your RAID array?

cfitz

i will only be RAIDing the 2 8MB buffer ones.

I use my space for Video editing (thus why i bought the Sony DRU-500A). but my works pulls me away from doing much video editing.

I need to test out this new DVD burner i got with this clients Sony lap top. that is going be fun!

[cfitz]

i will only be RAIDing the 2 8MB buffer ones.

Oh, okay. I was afraid you were thinking of setting up the three of them in a RAID 5 configuration. Will you be doing RAID 0 striping for speed or RAID 1 mirroring for redundancy?

cfitz

[dodecahedron]

i will only be RAIDing the 2 8MB buffer ones.

Oh, okay. I was afraid you were thinking of setting up the three of them in a RAID 5 configuration. Will you be doing RAID 0 striping for speed or RAID 1 mirroring for redundancy?

cfitz

why afraid?
what's wrong with RAID5?
what exactly is RAID5?

[cfitz]

what exactly is RAID5?

RAID 5 stripes data and parity across three or more discs. It allows for failure of a disc without data loss and automatic reconstruction of the failed disc when it is replaced by a new one. It is a good compromise between performance, data integrity, and disc usage.

why afraid?
what's wrong with RAID5?

There isn't anything inherently "wrong" with RAID 5. However, many people tend to automatically think "RAID" means faster performance relative to a single, plain-old disc. After all, a RAID array includes multiple discs that can be read from and written to in parallel. It must be faster, because the disc activity can be spread across multiple discs simultaneously, right? Wrong. It isn't that simple.

For example, a RAID 5 setup can improve read performance, but write performance is actually slower than for a single plain-old disc. Before a write can be made, depending on stripe sizes, the size of the write and the alignment of the write, the existing data must be read from multiple discs in the array, the parity recalculated, and the data written back to multiple discs.

The write penalty can be reduced by properly tuning the array to the application, configuring the stripe sizes, and adjusting cache settings. But this can be time consuming and difficult to do, and thus often isn't done right. The ops people at my job set up a server for a project of mine, and performance was terrible. My stinky little IDE desktop was running rings around a rack-mount server with SCSI drives. As it turned out, ops had configured a RAID 5 array and not tuned it properly. Once I pointed that out and they did their magic, the server beat my desktop (as it should have). This isn't a snip at ops, since they do a good job, but just points out that RAID 5 is not a slam-dunk.

Anyway, that was the basis for my "fear". I didn't want MonteLDS thinking RAID (in the guise of RAID 5) was the no-brainer answer to the problem of getting high performance for his video storage needs.

cfitz

[dodecahedron]

thanks for the explanations, cfitz.
i know RAID0 and RAID1. didn't know what RAID5 was.

what's the advantage then of RAID5 over RAID0?
as i understand RAID0 also gives you data integrity without a performance hit.

[cfitz]

what's the advantage then of RAID5 over RAID0?
as i understand RAID0 also gives you data integrity without a performance hit.

Oh no, definitely not. RAID 0 offers no data integrity improvement at all. On the contrary, RAID 0 actually increases the chance of data loss when compared to a single drive. RAID 0 is purely for boosting performance. It stripes data across multiple drives in parallel writes and reads to increase throughput, but doesn't include any parity information. Thus, a failure on any one drive in the array will result in loss of all data on the array.

I think what you meant was RAID 1. RAID 1 is mirroring - an identical copy of the data is written to two drives in parallel. It provides redundancy with no loss in performance, but requires two bytes of drive capacity for every byte of data you want to store, and thus isn't very efficient. That is where RAID 5 has some advantage. It provides redundancy and failure recovery without requiring a complete duplicate copy of the data, and thus is more efficient in terms of the ratio of drives required to available data storage space.

By the way, RAID 0 isn't actually an official "RAID" designation but the term is commonly used.

cfitz

[MonteLDS]

i am runing a RAID 0. I am about to test it right now to see how it is doing for speed

[cfitz]

i am runing a RAID 0. I am about to test it right now to see how it is doing for speed

Please let us know what kind of performance improvement you see. I am interested in hearing about some real-world experiences of people who implement RAID.

cfitz

[Tolyngee]

In my real-world experiences, RAID-5 has always beat the performance of RAID-1 and RAID-10, and definitely of a single drive.

But not RAID-0. BUT, as Cfitz already stated, RAID-0 isn't RAID anyway.

But the RAID-0/RAID-5 discrepancies were nothing to hoot and holler about either.

I do video editing, and I can tell you now that running RAID-5 is not the bottleneck on my system for the editing function.

Cfitz, as you said, it's not RAID-5 that's slow, but how it is implemented.

And I can guarantee you that my RAID-5 configuration will win in any performance benchmark versus pulling a single drive off of the array and testing it alone. (and, yes, I have done it already so I could determine what configuration I wanted to run...)

All up to what editing you are doing too. (he did mention this is why he bought a DVD burner)

If doing high-quality DVD authoring, any RAID will do. (I've done DVD authoring that would allow me to use floppies without a peformance hit...)


BUT you said:

"Before a write can be made, depending on stripe sizes, the size of the write and the alignment of the write, the existing data must be read from multiple discs in the array, the parity recalculated, and the data written back to multiple discs. "

Huh? if you want to write data to RAID-5, all you have to do is take the data, stripe it, calculate the parity (very simple XOR process), and write it. Where is this phantom read coming from?

But, then, at the end, you mention video "storage". For certain typed of video editing perhaps RAID-5 won't win (not by my experience though), but for video storage, RAID-5 is the clean winner based on the amount of storage you achieve versus data lost to the parity storage.

But, to end it: I was running a RAID-5 array (~1TB+), realized I needed PEFORMANCE (short-term) over storage capacity, and backed everything up and put the array in a RAID-10 ~600GB array.

And now I have a slower array.

I went against my own experience with arrays, cuz RAID-10 has gotta be quicker, and now I am kickin' myself every second until I can be bothered with taking the time to get it back into RAID-5...

In my experience, I must admit, RAID-5 has NEVER LOST in any real world situation I have ever thrown at it...

True, I can throw data at my RAID-0 at (guesses here) 90MB/s, RAID-10 at 85(?)MB/s, and RAID-5 at 70(?)MB/s.

But, if I am not processing any data at a rate of even 70MB/s, what does any of the above matter?

RAID-5 does appear to be more prone to data corruption though (though I believe this is the controller's fault of its implementation of RAID-5...)

But, yeah, any RAID is only as good as its controller.

[Bhairav]

But, yeah, any RAID is only as good as its controller

Ok. But aren't all the popular RAID equipped motherboards just implementing a software controlled RAID controller? I mean, stuff like the Promise 20376 ( or 23076.. can't remember!), and the Highpoint HPT-374 which seem to be really popular with mobo manufacturers, are just software RAID solutions,aren't they?
From what I have heard, heavy duty HARDWARE raid cards (stuff by Adaptec et al) which can implement RAID 5, are really expensive. But hey, they just follow the dictum that you get what you pay for!

[cfitz]

BUT you said:

"Before a write can be made, depending on stripe sizes, the size of the write and the alignment of the write, the existing data must be read from multiple discs in the array, the parity recalculated, and the data written back to multiple discs. "

Huh? if you want to write data to RAID-5, all you have to do is take the data, stripe it, calculate the parity (very simple XOR process), and write it. Where is this phantom read coming from?

This is a fundamental characteristic of RAID-5 that many people are unaware of (hence the common mistaken belief that RAID-5, with its multiple discs, is always faster than simpler disc configurations). It comes into effect when changing a portion of a stripe:

http://www.rare.com/products/raid_whitepaper.html
http://www.winnetmag.com/Articles/Index ... cleID=8255
http://www.ixora.com.au/tips/creation/random_writes.htm
http://www.pcguide.com/ref/hdd/perf/rai ... ite-c.html

As I mentioned, it can be mitigated by properly tuning the stripe size and cache usage for one's application. But, it is a fundamental characteristic of RAID-5, and is the reason RAID-5 is not the slam-dunk that many people think it is. RAID-5 can definitely have slower write performance than a single, plain-old-disc, even if the individual RAID-5 discs have better performance characteristics than the single plain-old-disc. I know, I've been there.

Reading, of course, is a different story. RAID-5 has a fundamental theoretical advantage in read performance over a single disc because the reads can be spread out in parallel. Still, even with reading the extent of the advantage will be influenced by the nature of the application and the design and sophistication of the controller.

cfitz

[MonteLDS]

I do video editing, and I can tell you now that running RAID-5 is not the bottleneck on my system for the editing function.

.................

All up to what editing you are doing too. (he did mention this is why he bought a DVD burner)

If doing high-quality DVD authoring, any RAID will do. (I've done DVD authoring that would allow me to use floppies without a peformance hit...)

.............

But, yeah, any RAID is only as good as its controller.

RAID 5 is nice but i can't afford to pick up another 120 WD w/ 8MB cache. I am now broke i spent some 800+ dollars on myself recently. But i have like 600+ $ that clients owe me. which reminds me i better make this fast.

RAID 0 is the way i went only for the video editing, and i knew it be fine (thus another reason not go RAID-5). Also, i suspect the next time i build a new box i am going do a SCSI RAID.

I have an onboard RAID controler (Asus board w/ promise controller)