Actually, now that I remember more about it, you are right. I was corrected in the "testing" thread that USB 2.0 would be the bottleneck no matter what. I ended up pulling an IDE drive out of my server, which at the time was a 250 GB WD SE 8 Mb cache drive, and doing some testing moving some CD image files (.isos) to the IDE drive from 1) a single Raptor, and 2) the RAID0 array of Raptors. I know the testing data was relatively similar in terms of time, but for the sake if giving all the details, I do not remember what stripe size I used...but it was the default choice on an IC7(non-G).PaHick said:Am I off in my thinking or what? I just dont think that would be a proper test.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
after formatting w/ raid 0
- Thread starter TrueGamer
- Start date
what is this everyone pick on djnes in this thread... yes I have to admit I have had my differences with him over imaging software in the past but thats all it was...differences in opinions and it was left at that.. now its a proving whos right or wrong thread to me... I respect djnes for hes very intellegent individual and I respect all other members on this board for their intelegence as well but lets give the OP who ever that may be.. ( lost track in this long thread lolol) the reason why he or she should or shouldnt use raid 0 and then give him an awnser to whats hes asking that way he has at least two ideas to ponder on... Not everyone throwing out why they do this or do that and then two individuals totally steer off path on a debate of the situation all together.. were adults here cant we all just get along...
just my rant.....
just my rant.....
Do you remember what I said about the added system complexity incurred from the decision to use RAID-0? The complexity that you dismissed as not that big of a deal?PaHick said:
Do you remember the benchmarks I showed, with RAID-0 giving a performance boost, of at best, 10% in real world applications? Never mind that you dismissed that saying "well, it IS faster"... Never mind the reduced game level loading performance...never mind double the cost for at best 10% faster...never mind the additional risk to data...never mind the additional driver and system complexity...never mind the complciated disaster recovery...never mind the reduced flexibility in physical disk resource allocation. RAID-0 is overblown hype. True, there are people that will accept all of RAID-0's drawbacks for the modest (at best) performance gains. However, my position on this all along has been that:
A) It's not worth double the cost. A value judgment admittedly, but double cost for 10% increased performance is a poor investment IMO, with gear like dual core CPUs, 2GB of RAM, and faster video cards out there.
B) Two slower drives in RAID-0 !=, !> one larger faster drive of equal or lesser cost to the pair of drives needed for RAID-0. This is well documented.
C) There are risks and pitfalls associated with RAID-0. These aren't well documented (not mentioned at all in "reviews" on RAID-0 setups), but as we can see with the OP's situation, a potential deal breaker and or time pit.
CREATE DATABASE in SQL Server; copying large text files; running simpler aggregations over flat-file data. The 35% number comes specifically from CREATE DATABASE on my test server, compared to running it on a single spindle. (I haven't compared it to creating two extents on different spindles.)Phoenix86 said:
In data warehousing, copying data from its remote source to a staging area, then building a partition, then copying the partition into the partition set (online, where it lives) is very common. It's recommended by experts like Ralf Kimball. It's a lot of copying, but if anything fails, you can just go back to the previous step. That is, it's certainly a real-world scenario.
Since the data is being copied, backup is inherent -- just copy it again. When it's loaded into the target server, it's off RAID0 and the backup ends up being the problem of the database administrator. Maybe they've got some online backup, maybe they siphon off their own copy of the partition. It's entirely possible that, for some subsets of the data, they don't care and let it fail -- it'll just be populated with more current day in the next refresh, anyway.
I figured as much, databases...
You know you should qualify statements like your 35%. You are not seeing a 35% increase in disk performance (which would apply to most all disk I/O), not just a special application.
The same logic you are using in your 35% post is the same logic that says RAID-0 is dead (using a single app to demonstrate a point). In fact, in the real-world DBs are not created on desktops and you know that, as you mention in you post.
I really don't know why you are bringing up server applications in a desktop discussion. What's the point?
Almost every RAID-0 discussion on this board, heck 99% of all threads are desktop related.
You know you should qualify statements like your 35%. You are not seeing a 35% increase in disk performance (which would apply to most all disk I/O), not just a special application.
The same logic you are using in your 35% post is the same logic that says RAID-0 is dead (using a single app to demonstrate a point). In fact, in the real-world DBs are not created on desktops and you know that, as you mention in you post.
I really don't know why you are bringing up server applications in a desktop discussion. What's the point?
Almost every RAID-0 discussion on this board, heck 99% of all threads are desktop related.
Really? What am I seeing, then? The operation is 35% faster and the CPU hasn't changed but the disk did, so why would you assert it's not a measure of disk performance?Phoenix86 said:
CREATE DATABASE does two things: formats pages in memory, and then write them to a file. Formatting pages means that the page is filled with empty data and added to a linked list, stored in the file. That's very little CPU time. I've never seen the operation not run I/O bound; you'd need more than a gigabyte a second to be CPU-bound.
I'm not making a decision; I'm just reporting an observation. I'm suggesting two different classes of applications -- data warehousing and database management sysems -- may both benefit from RAID0 in real applications.Phoenix86 said:
The assertion I am responding to wasn't limited to desktop use; it was an aggressive and unbounded indictment of RAID0's performance and applicability across the board.Phoenix86 said:
Why should not the indictment be the statement that's qualified, Phoenix? It says nothing about desktop-only.
As you stated, a 35% increase in CREATE DATABASE process.mikeblas said:
I believe you when you say you get a 35% increase. I believe it's directly related to RAID-0. I do not believe you are getting system-wide 35% increase in I/O. Otherwise it would affect many more applications.
I don't see servers listed either, nor do I see it listed in your "35%" post... See where I'm going? You are VERY quick to note when people don't qualify their statements, sorry you aren't seeing this in yours. That 35% statement is a blanket statement, and since it's not correct as a blanket statement, it needs to be qualified.
You DO know this thread relates to desktops, don't you? I mean, in general, posts are assumed to be desktop related unless otherwise stated. Coming in here and saying "ah-ha! look at RAID-0 on this server doing server tasks" is kinda pointless (especially when you don't mention it's a server task until questioned).
I don't want to discount servers, as it's an important topic too. However the normal discussions here are not server related. Lots of people saying "don't do RAID-0" would switch thier statements if you wanted to talk about DBs and server tasks.
It was asserted that RAID0 offers"at best, 10% in real world applications". I'm sharing the obervation that my particular real-word application gets a 35% benefit. I think that other real-world applications get benefits, too. Servers are a subset of the things that exist in the real-world.Phoenix86 said:
Would they? It would be very easy for you to do so, yourself, but instead you're changing the subject to the style of my post instead of discussing its meaning. Why not?Phoenix86 said:
Certainly, I don't read every note posted -- but I sure don't see anyone providing caveats when they parrot the "RAID0 is bad" line. Read the first two notes in this thread for an example: without knowing what application the original poster was referring to, the advice is to get rid of RAID0.
Right. So why isn't it valid to consider both in play? The reason you don't see a limit in my statement is that I'm providing a counter example to disprove an unbounded assertion. (It's here that I'm using logic.)Phoenix86 said:
Nope, I don't. TrueGamer never told us the role of the system he's trying to fix.Phoenix86 said:
Dude, reality calling, hello, are you there? Do you really think I'm arguing that servers are not a subset of all real things? OK, lets stop this one here.mikeblas said:
You didn't mention server or the application, that's ALL I'M SAYING.
Again, all I'm saying is you should have qualified the statement. You'd agree, no? I'll make my point farther down about the assumption this is desktops, not servers.
It is valid to consider both, I already said as much. You are making a mountian out of a mole hill. It's also smart to qualify when you are switching from desktop realm to servers.
Here's a hint. Around here, unless the discussion is servers, it's assumed desktop.
All desktop threads
exception (I think).
defuseme2k
[H]ard|Gawd
- Joined
- Oct 7, 2004
- Messages
- 1,074
Hello, my name is dr philwannabenstien. I don't run raid in my desktops, its worthless, you shouldn't either. I do run raid in my desktops its absolutely worth it if you think so or not. You're wrong, you're wronger more wrongly wrong. I'm mad, I'm madder, I'm getting so upset, watch me snort steam. SNORT.
^^lolz. I don't know why people care what other people do. I actually do have a raid 0 in my pc, but to be honest I couldn't tell the difference if there was one without it. The main reason I do it is to keep both drives as one volume without having to dynamic disks and all. Will my data get corrupted, maybe so, but then again 1 drive still has all the data on it and is still a single point of failure. A raid 0 both drives do, and you have two potential points of failure. I suppose one is more risky, shrug, but in either case if you don't your data backed up you're an idiot.
As far as the performance goes, as long as we're talking about these pathetic controllers that are built onto a motherboard I highly doubt we'll see consistant results showing the performance difference. I doubt it has anything to do with raid 0 in and of itself, so the zealots aren't wrong. Just wrong with a shitty controller. Buy a high end raid controller with less than 1% cpu utilzation. I've seen reviews for an acrea card on the internet that hauls ass.. the bad part is it COSTS OVER 500 dollars. Perhaps everyone is right, and just sees the data to confirm their own belief, making it evident to ones self.. who'dofthunkitwaspossible.
^^lolz. I don't know why people care what other people do. I actually do have a raid 0 in my pc, but to be honest I couldn't tell the difference if there was one without it. The main reason I do it is to keep both drives as one volume without having to dynamic disks and all. Will my data get corrupted, maybe so, but then again 1 drive still has all the data on it and is still a single point of failure. A raid 0 both drives do, and you have two potential points of failure. I suppose one is more risky, shrug, but in either case if you don't your data backed up you're an idiot.
As far as the performance goes, as long as we're talking about these pathetic controllers that are built onto a motherboard I highly doubt we'll see consistant results showing the performance difference. I doubt it has anything to do with raid 0 in and of itself, so the zealots aren't wrong. Just wrong with a shitty controller
. Buy a high end raid controller with less than 1% cpu utilzation. I've seen reviews for an acrea card on the internet that hauls ass.. the bad part is it COSTS OVER 500 dollars. Perhaps everyone is right, and just sees the data to confirm their own belief, making it evident to ones self.. who'dofthunkitwaspossible.
- Joined
- Feb 3, 2004
- Messages
- 27,001
Crosshairs said:
Yep...still right...how long can this go on....