gundersausage

The Death Of Raid

Recommended Posts

And (and this is just a niggling annoyance that won't go away) should we give credence to hand-timed benchmarks that display results to the nearest tenth of a second?

Should we care?

Share this post


Link to post
Share on other sites
I don't give a stinker what anyone says.

I have 2 Raptors in a Raid 0 array on a P4 3ghz.

I _definately_ notice a difference on my box.

The files are defragged monthly on my machine and I only keep 20gb of data on the 74gb partition - I (beleive) that the WINNT dir is towards the start of the disk also.

My 800mb swap file is on D: (PATA controller, not ICH5 SATA controller) and that's a 200gb WD Caviar JB.

My Raptors have a "problem" where I can only get about 25mb sustained real world throughput from them in ANY application, including apps which simply read but do nothing with the data (CRC checking apps -etc)

It's a shame about the problem - but thanks to the raid I can now write files quite quickly to C:.

Very scientific and persuasive, AbRASiON. If it weren't for the niggling spelling error (misspelling of the word definitely), this post would "definately" be worthy of a Ph.D. thesis.

:rolleyes::D

Share this post


Link to post
Share on other sites
I don't give a stinker what anyone says.

I have 2 Raptors in a Raid 0 array on a P4 3ghz.

I _definately_ notice a difference on my box.

The files are defragged monthly on my machine and I only keep 20gb of data on the 74gb partition - I (beleive) that the WINNT dir is towards the start of the disk also.

My 800mb swap file is on D: (PATA controller, not ICH5 SATA controller) and that's a 200gb WD Caviar JB.

My Raptors have a "problem" where I can only get about 25mb sustained real world throughput from them in ANY application, including apps which simply read but do nothing with the data (CRC checking apps -etc)

It's a shame about the problem - but thanks to the raid I can now write files quite quickly to C:.

Very scientific and persuasive, AbRASiON. If it weren't for the niggling spelling error (misspelling of the word definitely), this post would "definately" be worthy of a Ph.D. thesis.

:rolleyes::D

You can say what you like but basically some of you people assume "single users" only load windows and one app at a time.

I happen to move data around constantly - in 4.7gb lots to be precise - when reading from a WD Cav 250gb and writing to my Raptor raid setup I find it's very fast.

I find Windows load times to be faster.

Opening Word is faster.

I didn't say twice as fast - I said faster.

You people can dismiss it all you like - I've found superior performance with a RAID 0 setup of Raptors - period.

Share this post


Link to post
Share on other sites

I find this whole argument completely fascinating. There is one thing that nobody ever talks about. How is my system going to perform (RAID0 vs. one disk) after I have been using it for a period of time. I think we can all agree that a freshly loaded system always performs faster then one that has been around for awhile. Once the disks get fragmented does RAID0 perform faster?

By the way, I also think this is why people "think" that RAID0 helps them out so much. In order to create a RAID0 partition, they blow away all of their old stuff and start from a newly loaded system. I bet if, after they used the RAID0 system for a year, they blew away the RAID0 partition and put in a brand new 2mb 5200RPM drive they would be blown away by the speed increase.

Just some ramblings.

Share this post


Link to post
Share on other sites
I find this whole argument completely fascinating.  There is one thing that nobody ever talks about.  How is my system going to perform (RAID0 vs. one disk) after I have been using it for a period of time.  I think we can all agree that a freshly loaded system always performs faster then one that has been around for awhile.  Once the disks get fragmented does RAID0 perform faster?

By the way, I also think this is why people "think" that RAID0 helps them out so much.  In order to create a RAID0 partition, they blow away all of their old stuff and start from a newly loaded system.  I bet if, after they used the RAID0 system for a year, they blew away the RAID0 partition and put in a brand new 2mb 5200RPM drive they would be blown away by the speed increase.

Just some ramblings.

I agree with this.

Raid to my knowledge will not help at all with random access or track to track speeds, it's a limitation of the disk - and 2 or 5 of them simply won't help.

Hence DEFRAG OFTEN.

However for each and every file you read on the disk it will be anywhere from 0 to 100% faster - I find it's at least 50% in most cases - but it won't make much difference on smaller files (opening a 600kb zip file is practically instantaneous on any machine now)

Due to the fact I only keep about 25gb of data on my C: partition (72gb, 2x36gb raptors) plus it's not only defragged but I don't add or remove data from it (or rarely) I find mine is quite fast.

Share this post


Link to post
Share on other sites
Are you a RAID-0 sympathizer?  How many card carying members are there for gosh sakes?

I just find it interesting how passionate people get over having or not having a RAID 0 array. Personally, after seeing the lack of benefits I would not waste my money on RAID 0 again. Been there...done that...got the Raptor. :D

I think that the current on-going mentality, mirrors that of small-scale automobile enthusiasts. A computer without a RAID-0 array.. is like a ricer without a monsterous wing on the back. In either case, there may be zero or in fact negative benefit, but that fact doesn't change the minds of those that own them. The real benefit may be the image it conjures, not any real sort of benefit for those that care about getting actual work done with their PCs, but a definately benefit for those that use their PCs more to "show off" than anything else. ("Look at me! I've got a wing on my car!" - "Look at me! I'm running RAID-0 on my PC!")

'nuff said. :)

Share this post


Link to post
Share on other sites
Are you a RAID-0 sympathizer?  How many card carying members are there for gosh sakes?

I just find it interesting how passionate people get over having or not having a RAID 0 array. Personally, after seeing the lack of benefits I would not waste my money on RAID 0 again. Been there...done that...got the Raptor. :D

I think that the current on-going mentality, mirrors that of small-scale automobile enthusiasts. A computer without a RAID-0 array.. is like a ricer without a monsterous wing on the back. In either case, there may be zero or in fact negative benefit, but that fact doesn't change the minds of those that own them. The real benefit may be the image it conjures, not any real sort of benefit for those that care about getting actual work done with their PCs, but a definately benefit for those that use their PCs more to "show off" than anything else. ("Look at me! I've got a wing on my car!" - "Look at me! I'm running RAID-0 on my PC!")

'nuff said. :)

What's with people consistently whining about people having Raid here?

It seems like the abortion debate, both the pro and the con people whine just as much!

I don't have some fancy 54 line sig advertising my rig, nor do I spout of my system specs at every chance - I run raid because in MY case as a SINGLE user I _STILL_ find it improves performance, I do it for me, not for anyone else.

Just because it's TWICE the cost yet it's not twice the speed, does NOT mean it is not cost effective for "performance freaks" (i've hated the performance of hard disks since I was about 15 and got into hardware - they are so slow compared to the rest of the machine :( ....... )

My Windows boot times from bootstrap to login is about 7 or 8 seconds on my system - this is over 6 months after installing and my box being full of "junk"

It's a _FACT_ it was slower with a single raptor and it's a _FACT_ it was slower when I used a single 200gb WD caviar,...... it's a _FACT_ it was slower when I had a 60gb ibm (and so on)

My machine _IS_ 100% without a doubt faster with raid than without in many tasks.

If it reduces the load time of an application from 11 seconds to 9 seconds - it may not be twice as quick but I may be happy in feeling my machine is more responsive. PERIOD.

Share this post


Link to post
Share on other sites
It may feel faster, which is a subjective evaluation but the facts just dont support it

No it doesn't "feel" faster, it _IS_ faster - I'm happy to do your stupid stopwatch checks if you like but it's simply a fact that it's snappier.

Why do you people have to be so moronic, because Eugene claims something it must be absoloute fact?

Based on the purely illogical rubbish you people keep spouting, I guess a 15,000 rpm disk isn't any faster than a 10,000rpm disk, nor is a 7,200 faster than 5400 and so on?

We know for a fact that track to track and random seek times are more important than STR.

Does this therefore make STR useless? - why don't I just go back to my 20mb mfm WD, which pulled a whopping 640k a second because "str doesn't matter in a desktop environment"

Besides I've already pointed out in my posts that I keep very little data on the disk - to ensure the heads have to move very little at all

AND it's defragged

And when I DO use the disk it's primarily Sequential writes and reads.

Sigh.

Share this post


Link to post
Share on other sites
okey dokey, all the professional stopwatch tests are wrong and yours is the "magical" exception

I see you clearly pointed out a flaw in all of my points........

hmmmmmm no?

and I was not implying the stopwatch test is incorrect - it is an effective measurement if you factor in mistakes / averages - it's just silly to have to resort to that to prove a point.

I also sincerely doubt the changes I've seen are less than 15%.

(and no, I'm not claiming EVERYTHING is 15% faster or more, I'm claiming when I notice a difference it's easily 15% or higher)

Share this post


Link to post
Share on other sites

I ran a couple of 75GXP's in RAID 0 back then and it didn't feel one bit faster* than a single disk. I guess I must be wrong somewhere.

*before they both failed of course

Share this post


Link to post
Share on other sites

Face it folks, those who like striping will continue to stripe. Those who do not will not. No amount of testing and / or facts will persuade either group. Why not let this thread die?

Free

Share this post


Link to post
Share on other sites
Face it folks, those who like striping will continue to stripe. Those who do not will not. No amount of testing and / or facts will persuade either group. Why not let this thread die?

Free

Your analysis is a little off. I'll gladly consider striping a desktop system when doing so shows some repeatable, measurable benefits under expected usage patterns.

People commit the fallacy of special pleading when they suggest things are different under their particular circumstances. That may be so, but the evidence tends to cause me to doubt them, especially when they're only basing this conclusion off their perceptions. Perceptions are readily mistaken.

Raid doesn't seem like something that will be of any benefit for most desktop users. That doesn't mean there is no usage pattern that could conceivably benefit from it, only that most people are better served by not bothering with it.

Share this post


Link to post
Share on other sites

I've set up terabytes of RAID 0,1 and 5 storage in my professional life, and used to be considered pretty good at making Oracle and other large apps perform well. The key - as pointed out so eloquently by davidbradley - is to know when to use which, and when to use none.

Eventually, we will all run the next version of something like iSCSI, and connect to NAS boxes that will implement what I call "Invisible RAID" - they will analyze your data and usage patterns, and have a large farm of disks apportioned in RAIDs 0,1, 5 and none, and make the appropriate decisions as to what data should go. Professional storage vendors are doing the research on this now - the hard part is having a good enough metadata model to know the value of a given set of data - should it be RAID 1, or no RAID at all? And what about data that is almost never accessed, but when it IS you want it to fly like lightening (say in response to a critical situation)? That's the hard part...and if I had the answer I wouldn't be posting it here, I'd be submitting my business proposal to EMC.

In the meantime, we will flounder around trying to apply RAID manually when and where we each see fit. We all have some need for some RAID - the problem is that we usually don't need a whole disk of it, and few know which to use. Intel's Matrix RAID is a huge step forward in this regard, and will probably lead to MORE people experimenting with RAID 0 and 1, regardless of Anand's or SR's point of view. Consider Intel a pusher then...

Future Shock

Share this post


Link to post
Share on other sites

Well, there is one thing nobody talks about. Let's take a 10 000 rpm 74 gig Raptor, 8MB cache and a RAID 0 array made out of two 7200 rpm SATA drives with 8 MB cache each.

Raptor = $200 = 74 gig

RAID0 = $200 = 240 gig (add 30 bucks and 320 gig)

If they perform the same /someone said here that the RAID0 is IN_fact faster, but let's don't go there/ I still have 3x, 4x as much space for the same dough, right?

My dime...

Share this post


Link to post
Share on other sites

Raptors aren't purchased for their cost/MB ratio. If you want the most space per dollar, you can get several 260-250GB hard drives for about 50 cents per MB. Raptors are usually purchased by enthusiasts for performance reasons, and possible less so because of their warranty and possibly greater reliability.

Share this post


Link to post
Share on other sites
Why do you people have to be so moronic, because Eugene claims something it must be absoloute fact?

Probably because the bulk of research/benchmarks I have seen, combined with my personal experience, lead me to believe it. And I see no large number of proponents of no RAID at all; just those who think in the majority of desktop situations that it isn't useful. Would you care to argue that point?

Based on the purely illogical rubbish you people keep spouting, I guess a 15,000 rpm disk isn't any faster than a 10,000rpm disk, nor is a 7,200 faster than 5400 and so on?

I'm having trouble understanding how your analogy applies here. Could you explain it in more detail? Currently, I don't understand how the logic and evidence we are using to make the point that RAID0 is generally not useful in desktop situations would draw a logical conclusion that RPM had similarly little effect.

We know for a fact that track to track and random seek times are more important than STR.

Does this therefore make STR useless? - why don't I just go back to my 20mb mfm WD, which pulled a whopping 640k a second because "str doesn't matter in a desktop environment"

No one said STR isn't important at all; we have just made the statement that the difference between 60 MB/s and 120 MB/s isn't significant.

Besides I've already pointed out in my posts that I keep very little data on the disk - to ensure the heads have to move very little at all

AND it's defragged

And when I DO use the disk it's primarily Sequential writes and reads.

And Eugene does in fact have an image he uses which is, if I am not mistaken, a fairly fresh install, defragged. The fact of the matter is that defragging does not necessarily make the majority of requests sequential. Those requests are determined by application. And it has been proven that even tasks where one would think most reads are sequential, they are not (e.g., game map loading), and that doubling transfer rates doesn't even come close to doubling load times...in most tests, in fact, does not significantly affect it at all.

You asked that gundersausage address your points. While it's unlikely that he will bother, I hope you're just as willing to respond to my more reasoned disagreement.

Share this post


Link to post
Share on other sites

I have a LSI Logic MegaRAID 320-2 and am replacing it with a 320-2X PCI-X SCSI RAID controller with 512MB cache.

SCSI RAID 1: OS, Win2K Pro and XP Pro

SCSI RAiD 10 (6 drives): Apps and games

SCSI RAID 10 (4 drives): temp, paging, vacation pictures and movies

Two hot spares

Drives on each array split on each channel. All Seagate 10.6 U320 36GB drives in Supermicro SCA enclosures on two chassis.

Tyan Thunder K8W

Dual Opteron 240s

3GB PC3200 ECC Reg, 2 GB CPU0 and 1 GB CPU1 in 128 bit interleaved mode.

Performance is great on the 320-2 with 256MB battery cache. I cannot wait to see a 64/133 PCI-X SCSI RAID controlelr with 512MB battery cache. I would not use RAID 0. No fault tolerance.

I read the article TCQ, RAID, SCSI, and SATA. I found the SCSI representation to be weak at best. Why not use a U320 SCSI RAID 64/66 or PCI-X controller instead of U160. Sorry, and Acceleraid controller is not a great choice. Where can I get your drivemark program to run it on my workstation?

Share this post


Link to post
Share on other sites
Guest Eugene
Where can I get your drivemark program to run it on my workstation?

As illustrated in the article, the improved throughput delivered by U320 / PCI-X would not significantly change the results on their own.

Similarly, the effects of a hardware processor and onboard cache (should some consider the Acceleraid's processor/cache weak) only exert themselves when serious parity calculations (RAID5ish) come into play.

The Acceleraid itself, independent of objections that it's only U160 and that its processor/cache are not up to snuff, may indeed not be the best choice to absolutely maximize the SCSI array's performance. That said, however, you seem to believe that a better controller would deliver a truly massive difference. This is not likely- what we'd instead see would be an incremental improvement. As a result, the article's conclusions would stand.

The DriveMarks are proprietary.

Share this post


Link to post
Share on other sites

Too bad about DriveMarks. It would be an interesting test on my workstation. Storagereview needs to release a common benchmark that users can run to compare their hardware. IO Meter does not give a realistic result for single user SCSI RAID.

I found that setting my 320-2 to read ahead, write back, direct IO did help performance on gaming, EE analysis software (PSPice), and other high disk utilization apps MyDVD, Photoshop, Autocad, etc. Changing to PCI-X SCSI RAID with my configuration is to increase throughput. I will have 6 active + 1 spare per channel. I was surprised with the results. If they considered U320 SCSI RAID, why not include it in the graphs?

Share this post


Link to post
Share on other sites

Here's and interesting article, not 100% scientific, but a hell of a lot better then "my system feels faster" anecdotal evidence some feel the need to regurgitate on command ;)

http://www.overclockers.com/articles1063/

I've know this for years and performed many test that have shown these results, the ONLY time we build RAID0 arrays is for analogue video capture (40-45 MB per second sustained over 100-200 GB)

Cheers

Cezar

Share this post


Link to post
Share on other sites

I am thinking about building a RAID 1 set for my next PC.

I am thinking about getting 2 x Western Digital 250GB S-ATA drives and using the on motherboard controller (most likely the nforce chipset)

I am not too concerned about performance, but I would like to know for sure what type of advantages/disadvantages there are. Has anyone else configured there systems like this?

Share this post


Link to post
Share on other sites

RAID 1 is fault tolerant. Write performance does take a hit. You can minimize it by placing one drive on each channel, if applicable. Does the board have SATA RAID or just SATA?

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