gundersausage

The Death Of Raid

Recommended Posts

Never assume that because you have a RAID1 (mirroring) you can skip disk/CDROM/DVD/tape backup. A virus, software error, controller problem, user action can still render RAID1 useless for recovery.

RAID1 only protects from disk drive failure (typicaly one drive in array).

A software/controller error causing corruption of disk data will be written to both disks : RAID1 replicates the file system corruption.

Important data and historical (version, 1month, 3m,6m,12m,2yr,...) are still necessary for better long term protection. These tertiary backups will allow recovery of a file destroyed by a virus/worm or user error. If a user deletes a file from a RAID1 array it is deleted from both drives. In terms of data recovery limitations discussed hare, RAID5 is similar to RAID1.

Separate backup (held offsite, in fireproof safe) can be used in case of main system being stolen, or damaged by fire etc.

There is nothing wrong with RAID1 but just be aware of what it can and can't do.

Share this post


Link to post
Share on other sites
I'm eventually going to do RAID1 (mirror) on my "big drive", so that I don't have to backup 250GB.  Anyone that does RAID0 on more than two drives must like playing with fire.  :)  (Go RAID5 instead.)

You do realize that RAID is not a substitute for backups?

Yes, tygrus summarized it nicely that any RAID is not infallable. :) But it should make your system more reliable instead of less reliable, unless you're using RAID0 (striping).

Share this post


Link to post
Share on other sites

Yeah I totally agree RAID-0 is for morons and retards only. Massive reduction in system reliability for questionable performance gain.

All of my RAID arrays are RAID-5 currently. On the whole, this has worked OK. But its clear to me that there are performance bottlenecks in my current setup (LSI Ultra320-2 card in a PCI 33MHz/32-bit slot with 6 x Cheetah 15k.3 73 MB disks, Asus P4C800E-Deluxe MB, Pentium 3.0 Northwood).

My plan for the fall is to get the Asus NCCH-DL motherboard, with a pair of Nocona 3.6 GHz/1MB L2/800 FSB processors. And to get a LSI Ultra302-2x board with PCI-X and a much faster RAID XOR processor. Hopefully that will improve performance, has anybody else got more suggestions? I know that going completely RAID-1 would be faster, but then I'd need 10 disks instead of 6 and the cost and heat problems would mount up.

Share this post


Link to post
Share on other sites

Isn't one of the only few real life workstation benefit of raid 0 striping of disks solely used for Photoshop's scratch disk?

Share this post


Link to post
Share on other sites
Personally, if I happen to get one of the new 925 or 915 boards, I'll make use of Intel's "Matrix RAID".

I do alot of video editing, so I have a RAID for 'temporary storage'.  Any increase in STR is a good thing for video editing.  What I'd like is a Raptor for OS/Apps, and a pair of Hitachi 400GB drives, and have one 'partition' RAID-1 for user space, and one 'partition' RAID-0 for the scratch drive.  (Then the last SATA port could be for the new Plextor SATA DVD burner.)

But, at this point I'm looking toward a dual 2.5GHz G5.  Haven't decided yet if I'll replace the stock drive with a Raptor for OS/Apps; but I'll have an external drive or three for scratch space.  (I already have three FireWire hard drives, one 160GB, and two 80GB.)  Probably RAID-1 the two 80GB drives for user, and get one of LaCie's 'Big Disk Extreme's for scratch space.

(All of my video editing is personal, not professional, so I'm not going to spend thousands on just the video editing subsystem.)

Looks more like you'll be sending it back to the vendor for a replacement.

http://forums.storagereview.net/index.php?showtopic=15921

Share this post


Link to post
Share on other sites
Yeah I totally agree RAID-0 is for morons and retards only. Massive reduction in system reliability for questionable performance gain.

I'd like to think that the group of ppl on SR who still chose the go RAID-0 suffer from the "placebo" effect.

The only difference from the clinical trial patients is that they know the "drug" they're taking is a fake - from SR FAQs and forum feedbacks, they KNOW the pros are minimal at the cost of increased risk of data loss.

:P

Share this post


Link to post
Share on other sites
Guest Eugene

http://www.anandtech.com/storage/showdoc.html?i=2101

Whoa, this is some revolutionary crap, eh?

The -real- question is whether Anand will be better at convincing his own writers of RAID0's uselessness than, say, SR has been at convincing its own readers.

The article right below, for example, recommends Raptors in a RAID0 array. :rolleyes:

http://www.anandtech.com/guides/showdoc.html?i=2100&p=9

Share this post


Link to post
Share on other sites

From Anand's articles:

If you haven't gotten the hint by now, we'll spell it out for you: there is no place, and no need for a RAID-0 array on a desktop computer. The real world performance increases are negligible at best and the reduction in reliability, thanks to a halving of the mean time between failure, makes RAID-0 far from worth it on the desktop.
There are genuine performance advantages to a SATA RAID array, and a RAID 0 Array with two of the fastest SATA drives that you can buy makes for a blistering high end system. The storage capacity for two 74GB WD Raptor drives is a generous, but rational, 148GB of fast access storage. Compared to SCSI RAID solutions with similar performance, the Raptor RAID is a virtual bargain.

Hey - sometimes you gotta walk the fence I guess. Everybody loves Anandtech.com! No matter what your viewpoint, you can get some of it there!

Share this post


Link to post
Share on other sites
That's "non RAID 0 goonies," not Anti-RAID thugs;)

Yeah, I still prefer the drive performance beneifts of RAID 0.  I buy better processors and more memory for application performance.  Still amazes me that people test application performance to benchmark RAID.  Ok, so let see how many people we can cram into that new Corvette and still drive it.  If we can get more poeple in there, it's a better faster car.  lol ;)

That's fine - you continue to enjoy the non-existent drive performance benefits. Of course, if your determining that the raid is faster simply by your perception - this has already been covered, a users perception is massively compromised by his/her expectations.

The part about you being amazed that people test application performance to benchmark a raid 0 volume explains why your so reluctant to see the truth.

Application performance is the bottom line - nothing else matters. If it doesn't make your applications run faster, then its pointless junk.

Share this post


Link to post
Share on other sites

Interesting article that was just posted:

Western Digital's Raptors in RAID-0: Are two drives better than one?

"Bottom line: RAID-0 arrays will win you just about any benchmark, but they'll deliver virtually nothing more than that for real world desktop performance. That's just the cold hard truth."

I guess others are starting to see the real-world effects also. :huh:

Share this post


Link to post
Share on other sites
Guest Eugene

Those who want a real treat should peruse through the article's associated comments from readers.

It is absolutely staggering how many folks refuse to give up their beliefs, as if belief that RAID0 MUST HELP MY GAME LOAD FASTER is a cherished childhood tenet.

Anand's readers deserve to be elsewhere'd... and they will be, tomorrow :P. I'm taking nominations for the choicest quotes.

Share this post


Link to post
Share on other sites

I assume of course you mean the comments on this article:

Anandtech's Recent Raid Article

I brought this article below up once before, but in the middle of a smoking thread where everyone was fighting. It's the base cause of all of this nonsense though:

We perceive what we expect to perceive

Anand's readers' comments for that article, as well as the SR boards, are flooded with posts/comments such as: My windows just feels faster, the windows experience is alot smoother, my stuff loads faster, it feels snappier, etc.

If you think your RAID is fast, it feels fast. But it's not. The same argument can be applied to the IDE/SCSI war. The article above states that your perception is heavily influenced by your senses, including hearing. I have experienced this myself - a silent drive just 'feels' slower to me then a loud grating 10k SCSI. Drives in RAID 0, unless the silent type, are going to make more noise, and feel faster. (Never mind the fact that you are 'expecting' it to be faster)

This is why we must trash every single argument that puts the "Screw the benchmarks, I know my system feels snappier" viewpoint forth.

Time to realize the truth, RAID is for servers - putting it in your workstation makes more noise, heat, costs more $$, introduces more failure points, double the risk of your drives failing, etc.

But dont worry, this isnt the death of RAID for everyone - just those that know better. There are plenty of users that combined with mass marketing will keep RAID alive and well.

Share this post


Link to post
Share on other sites
If you think your RAID is fast, it feels fast.

Well, before I get jumped on, I'll clarify - if you think it will be faster then a single drive of the same manufactorer, same bios, etc, then it will feel faster.

Share this post


Link to post
Share on other sites
I'm going to sit on the fence and get the flak from both sides  :P

I think RAID has its use depending on the situation, RAID 1, 5 etc that have some type of redudency is a great thing especially as has already been stated;

You can tell someone to make regular backups all day, and the first day they listen is when their drive dies

What I would like to see is a RAID 1 or 5 that can use all different sizes of HDs at their full capacities, to me that makes much more sense than the raid 0

That's really the one thing that I could see being useful, in a "home user RAID" scenario (RAID-1). Considering the declining cost of RAID (most mobos have it as a built-in option now), and both the declining cost and reliability (at least warranty periods) of consumer-level PATA (and now SATA) disks, then it makes a certain amount of sense for home users to, by default, run a RAID-1 setup, if only to protect their data when a HD goes belly-up.

(Note that I am not advocating this *instead* of backups, heck no. Ghost is your friend. Get to know it well.)

Touching on your second point, I was having a thought the other day, on "strange" RAID setups. What about running 2 x 160GB drives in RAID-0, and then running that set as a logical drive, as part of a RAID-1 setup with another physical disk? (Say a 250GB, and only use 250GB of the RAID stripe logical disk.)

The idea would be, if the RAID subsystem could have an independent-reader optimization algorithm, then some read-only accesses could be supplied at the higher STR of the RAID-0 stripe, and yet, the data stored on that faster stripe would also be safely and redundantly mirrored on another drive.

So you get your high STR, and your failsafe redundancy, with only 3 physical disks instead of 4. Does anyone think that would be a feasable setup?

Share this post


Link to post
Share on other sites
I brought this article below up once before, but in the middle of a smoking thread where everyone was fighting.  It's the base cause of all of this nonsense though:

We perceive what we expect to perceive

Anand's readers' comments for that article, as well as the SR boards, are flooded with posts/comments such as: My windows just feels faster, the windows experience is alot smoother, my stuff loads faster, it feels snappier, etc.

If you think your RAID is fast, it feels fast.  But it's not.

That is an awesome article. Thanks much.

I think that Figure 2 goes a long way toward explaining Picard and his ilk. Perceptions form quicky and resist change. People got the idea that RAID-0 helps performance in all cases. People got the idea that STR matters. Unfortunately, people have a hard time getting rid of those ideas.

Share this post


Link to post
Share on other sites
Guest Eugene
People got the idea that RAID-0 helps performance in all cases.  People got the idea that STR matters.  Unfortunately, people have a hard time getting rid of those ideas.

And... don't forget to add SCSI to this list.

Share this post


Link to post
Share on other sites
So you get your high STR, and your failsafe redundancy, with only 3 physical disks instead of 4. Does anyone think that would be a feasable setup?

No. STR doesn't matter except in a few specialized applications. The entire point of this is not that RAID 0 is more fragile - we have always known that. The point is that there is no benefit to it.

Share this post


Link to post
Share on other sites
No.  STR doesn't matter except in a few specialized applications.  The entire point of this is not that RAID 0 is more fragile - we have always known that.  The point is that there is no benefit to it.

Well, there is indeed still the one benefit it was designed to have: making smalller disks "into" one larger disk for the "convenience" of the user / OS.

"tmd"

Share this post


Link to post
Share on other sites

There is something to be said for that - although a larger disk is not really useful unless someones preference just lies that way. Also with modern OS's, you can typically mount drives in folders rather then a new drive letter. So unless you want double the size with all files in the root, it becomes an uninteresting feature.

Share this post


Link to post
Share on other sites
There is something to be said for that - although a larger disk is not really useful unless someones preference just lies that way.  Also with modern OS's, you can typically mount drives in folders rather then a new drive letter.  So unless you want double the size with all files in the root, it becomes an uninteresting feature.

agreed.

tmd

Share this post


Link to post
Share on other sites
So you get your high STR, and your failsafe redundancy, with only 3 physical disks instead of 4. Does anyone think that would be a feasable setup?

No. STR doesn't matter except in a few specialized applications. The entire point of this is not that RAID 0 is more fragile - we have always known that. The point is that there is no benefit to it.

Point taken. Possibly I got taken up by attempting to be too "clever", by implementing a (hypothetical) RAID 0+1 sort of setup with only three disks instead of 4, and missed the bigger point.

I'm still more of a fan of independent disks myself. OS + apps + registry + swap, and then seperate drives/partitions for bulk data, and things like CD images, etc. I might have attempted to try RAID at some point in the past, if only I had two drives of the same size. I tend to always buy a slightly-bigger drive when I upgrade, so I never have a matched-pair.

I wonder if RAID-0 would see more benefit under Longhorn, with it's new "SuperRead" caching strategy.

The way I look at this, it shouldn't really be an issue of dogma either way, pro or con. It's kind of like dual-channel RAM, Socket-A can't really take advantage of it, because it doesn't need it, but yet the P4 platform can. Perhaps our current OSes are akin to a short-pipeline AMD, whereas Longhorn's design will be more similar to the P4, with big caches and larger cache lines. Just a thought.

Share this post


Link to post
Share on other sites
...demonstrated that the gains in striping them were minimal especially when one considered  that a large portion of what gain there was due to the doubling of capacity and the resulting smaller distances that the actuator must travel to execute the same chain of disk accesses.

Ahem, can you define "minimal"? 1%? 5%? 0.5%? <_<

Well, I have always been in this category of RAID-believers, especially if the controller is integrated on the mobo and comes in a right price. Even if those advantages above are due to doubling the capacity and decreasing actuator distances, they are always RAID-0 advantages to my eyes.

Other than doubling the possibility of drive failure, are there any really bad disadvantages running two HDDs in stripping array?

Share this post


Link to post
Share on other sites

I just can't believe Storage Review is making such bold claims on such limited data. The SR results only prove there is a limited performance improvent from striping in very light desktop workloads on a legacy PCI bus. Using a four years old SCSI RAID adapter does not tell anything about the performance of modern SCSI RAID setups.

Striping is insane? Depends on your definition of 'desktop performance' and your target audience. Grandma who occasionally emails her friends and has a hard timing handling just a single application won't benefit from striping. Grandma doesn't care about RAID. She doesn't care about storage or computers either.

The average computer enthusiast can create some pretty heavy disk I/O. Many people are heavy into p2p sharing, digital photography or video-editing. They move large amounts of data around their hard drives. These people will greatly benefit from striping (in combination with mirroring or parity). Your hard drive won't stall when you're making a system backup over a gigabit network or when the weekly full system virusscan is running. File copy times will be greatly reduced.

Why do people buy cars with six seats when most of the times only one or two seats are occupied? Because sometimes they take their family on a trip. Why use striping? Not to improve MS Office load times but to improve performance in scenarios with heavy I/O. This may happen frequently or rarely, but when it happens you will be happy to have multiple drives working in parallel.

I use a RAID 5 array with five Maxtor Atlas 15K drives on a LSI MegaRAID SCSI 320-2X with 512MB cache. Performance: 240MB/s sequential read, 180MB/s sequantial write, desktop I/O performance three times higher than a single Raptor WD360GD. It is virtually impossible to get this array to stall. It doesn't matter if you're pumping 35MB/s over the network or have a virus scanner or file defragmentation running in the background. The only way you will notice it is because of the blinking LEDs on the front panel or the crispy seek noises making its way out of the case.

Share this post


Link to post
Share on other sites
I use a RAID 5 array with five Maxtor Atlas 15K drives on a LSI MegaRAID SCSI 320-2X with 512MB cache. Performance: 240MB/s sequential read, 180MB/s sequantial write, desktop I/O performance three times higher than a single Raptor WD360GD. It is virtually impossible to get this array to stall. It doesn't matter if you're pumping 35MB/s over the network or have a virus scanner or file defragmentation running in the background. The only way you will notice it is because of the blinking LEDs on the front panel or the crispy seek noises making its way out of the case.

STR is only one aspect of disk I/O performance. The arguement over raid isn't that it can't achieve very high STR rates, but rather the importance of STR in desktop (or server) performance. Two of the activities you mention (video editing and photoediting) do benefit from improved STR. Most everyday activies that add to the "percieved" feel of a computer...don't. If you've got enough memory in the machine....ideally you should rarely be touching the disk.

File copying is a a tricky area....keep in mind that to copy a file on a stripe set each block of data must be copied into memory, the drive heads repositioned and then writen to the array. Even if the benched STR of your array is ~80MB/sec, your Effective STR is the same as a single drive, ie ~40MB/sec. Now obviously with more disks in the array your STR wil scale....but 4 drive stripe sets are not indicitive of what most people are trying to run (nor do stripe sets scale linearly). Now, if we consider the case of a file being coppied to an array from somewhere else (network, dvdrom, another hard drive...) then the effective STR is almost assuredly going to be limited by that device and not the array. Under ideal circumstances even GigE has a maximum real throughoutput of about 700Mbps....or about 85Mb/sec......right around the performance level of a single modern fast disk.

Oh, and STR for peer file sharing? That has to be a joke, right?

-Chris

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now