Last Minute Advice Requested - Page 2

Ok, now to answer your question:

No, RAID will gain you nothing. RAID level 0 sucks ass, thats why the industry stopped using it decades ago. Furthermore, RAID is designed to be used with SCSI setup not with IDE. The IDE bus can't handle raid like SCSI can. With SCSI you could put 15 hard drives on a chain and access them all simultaneously, independantly, and without performance loss. Furthermore SCSI drives spin at speeds regular IDE drive can't (10,000 rpm up to 20,000 rpm). The SCSI bus channel is wider and have shorter access times. The biggest problem with IDE channels is you can only access one device per channel at a time. So if you have the two hard drives in the notebook (which are on the same channel) and the opticals on the other (note the 8890 has room for three hard drives and one optical). So, as others have posted, they found that they couldn't really use the 3rd drive and burn cd/dvds at the same time. This is because only one device per bus can be used. The IDE controller switches between the two devices to prevent one from preventing all use of the other device, but serious performance is lost. There is a reason why IDE drives have jumpers labeled "master" and "slave" since only one at a time can be used and the "master" device has priority over the "slave."

Ok, now bring this into RAID. With RAID level 0 on the notebook you have two IDE drives (lets stick with two for now). Both occupy the same IDE channel. Lets say you setup spanning. Ok, no performance increase here because all spanning does is let you create larger partitions, it doesn't split up the data in nice neat sections, it just writes until it runs out of room and continues on to the next drive as necessary after the first one is full. Joy, gee that was useless, with 60GB or 80GB drives on the market why the hell would you do that anyway.

Ok, example two. Lets use the 8890, the only notebook with options to have three hard drives. Lets setup RAID level 0 as disk stripping. Now we have three drives being trying to be accessed simultaneously across two buses. The bus with two drives on it has to slow down and constantly shift between writting to both drives while the third drive also has to slow down since it has to wait its turn for data after the other two are done since the data is STILL written in sequential order. Yaaaay, wasted time! Oh, and god forbid should you actually try to transfer data to or from a CD/DVD drive. Now BOTH buses are full and all four devices are competing for bandwidth and their turn. Just great, wonderful, the term "f**ktard" comes to mind.

In conclusion, only use IDE based RAID solutions if you absolutely need data security and redundancy. There just might be a reason why desktop IDE RAID controller cards don't sell very well and are very hard to find. I repeat unless you use RAID 5 or 6 using SCSI controller cards and drives, there is no appreciable performance increase. There just might be a decrease in performance.

Just get one or of the 60GB 7200 rpm drives and keep them separate. Store data on both if you want for redundancy, that's what I do. I even bought a DVD burner just for archival purposes.

I am not sure about needing 3 HD to use stripping, I will look in to it. I did not think it was possible to have an IDE raid ether, so I sent off an email to some raid manufacturer a few years back. He basically said, "if you disk stripping, even if both HDs can not used at the exact same time, the off HD will be storing data in its buffer waiting for its turn" The performance increase won't be the same as if you were using SCSI, but it will still be there. I never tried using an IDE Raid so i don't know if it it is true or not, i guess we will have to wait and see.

Well, I certainly love overly broad statements totally bereft of supporting evidence. In all honesty, until someone sits down and benchmarks (carefully) the performance of Raid 0 on the 8790, no one, not even those who profess knowledge can offer factual data on what performance gain there may or may not be with Raid setups on this particular notebook.

The real answer requires testing of the hardware in the environment that is intended. Any Raid performance will vary depending on what exactly you are doing. Frequent access of very small files will have a different performance impact from less frequent access of very very large files. The stripe size used will have an effect. The efficency of the motherboards implementation of Raid 0 will have an impact. .. etc. Its far too complicated a question in reality to be able to arrive at a simple black and white answer.

It could be faster. It could be slower. It could be less reliable. The reliability of the drive setup may be so high to begin with, that the lowered Raid 0 reliabilty will not be noticeable in practical terms. The implementation could be so bad that CPU utilization is so high that performance is woeful. The system could be so powerful that CPU utilization, no matter how it peaks, is not noticeabel to the end user.

Short answer, until someone sits down and generates the data, do not accept anyone's broad sweeping opinions one way or another. It pays to be a skeptic from time to time.

-Card

still doesn't change the fact that:

1) RAID was designed to be run in SCSI arrays, not IDE channels.
2) So many poeple are posting that "I'm gonna get RAID 0 'cause its faster" all over the place in the 8790 and 8890 threads. Its disgusting, I bet most of these people don't even know what RAID is. I have physically seen true RAID arrays with banks on banks of hard drives with hot-swapable capabilities running data servers. That's what RAID is for, not taking two slow IDE drives together on the same channel and hoping for the best. No one else seems to realize that you must have at least 3 hard drives to do stripping, otherwise its spanning.

... sigh ..

Please, stop with the non-information. You certainly can have a 2 drive array with striping, ie. a 2 drive RAID 0 system. You are confusing this beyond any rational reason.
Raid 0 explanation

- Card

Ever think that striping and spanning are the same thing, with the difference that spanning is limited to 2 volumes and striping is for more than two volumes.

I have been doing a lot of research for LARGE open system storage and ATA using various flavors of RAID are offered by the Major vendors (EMC, HP, etc...)

Times change and references become obsolete.

I will run SANDRA on the RAID-0 and provide results. The someone else that does not have RAID enabled could do the same thing and we can then all compare the results.

I suspect that some will never change their opinion even when confronted with facts.

SEMC, I will put my credentials up against yours anyday.
Given you have the words "IT Trainee" under your handle I would be careful slagging off against some of the people on this forum. Whilst there are people here who's knowledge is a bit warped (including your own my friend) there are some *very* experienced professionals who frequent this board.

Before I refute some of your incorrect assertions let me lay down my resume:
Degree in Chemical Engineering, honours in Computer Science, Electronics Engineering sub-major.
Degree in Law
Computer experience: Started 1972.
VLSI/ASIC design for Canon, computer design from chip, circuit board level, bios and operating system. Robotics, process control and automation design for 20 large scale factories. Traffic control systems. Unix system admin since 1976 (version 6) - systems supported Sun, DEC, HP, IBM. Coding - everything from microcode, C, C++, Fortran, Cobol, Pascal, assembler (12 different chip sets), Occam, Perl, OO modelling. Developer of several automation languages used by Philips (ETX). Webmaster, database modelling, Oracle, Informix, DB2. MVS, VMS, CP/M, Unix, Linux. Largest database server: 3 TB. I could go on but I think you get the picture. You have probably read some of my posts that should also give you an idea of my experience.

First point: I don't know who wrote your Unix textbook you quoted from but it is a *very* poor description (I would say misleading) description of how RAID works. (Using three disks to explain striping - that has to be confusing and their description of RAID 3 is just plain wrong)

Second point: whilst RAID is most often used with SCSI disks it is a fallacy to say it does not work with IDE drives. The interface is not relevant except as you correctly point out that IDE cannot release the bus as SCSI does during command execution. It does not mean that RAID 0/3/5 implemented with IDE drives will not speed up data transfers. They will not be as fast as SCSI but please get your facts right.

Third point: rotational speed has nothing to do with the drive's interface. You have the cart in front of the horse. Fast rotational disks are normally released with SCSI interfaces first because that is where the best price margins are. Companies who run servers are not as price sensitive as consumers. You can have IDE drives that rotate as fast as SCSI drives. Again your facts are just plain wrong.

Fourth point: RAID (0) will give you nothing - wrong again. The industry did not stop using it decades ago. It is still very much in use thank you very much. I am sure some of my clients would like to take up that point with you. The risk of failure is there but it is no different to losing a disk from a JBOD (just a bunch of disks) array. Still have to restore from tape or whatever. Normal commerical tradeoff for certain applications.

Point five: Whilst the 8890 supports 3 disks, only two of them can be configured as a hardware RAIDset. The third disk is on an independent bus.

Point six: whilst I applore your swearing whilst trying to explain a technical concept (shows poor language skills imho) the 8890 RAID setup with two disks *does* increase data throughput. Fact. If you actually owned an 8890 (which I gather from you sig you don't) you would know this. The reason Clevo have the Promise RAID controller in the 8890 is to overcome the limitations of IDE interfaces when it comes to RAID. Go to my review and follow the link to the Promise specifications. I doubt you have actually read the specs for the chip. So stop mouthing off about things you obviously no nothing about.

Finally without putting too fine a point on it, if you are an IT trainee, I suggest you go back and do a hell of a lot more study because what you currently know is wrong.

Aussie,

From time and experience comes wisdom. For some, wisdom is overcome by time and technology. In your case, the latter does not apply. Points very well stated.

Thanks Aussie,

For writing a very coherant reply and saving me the time, from my workstation running raid 0 on IDE, to write it myself.

I dont own a Clevo yet, but I do have some benchmarks somewhere on access times / data transfer using the desktop versions of the promise ATA raid controller cards. The drives in desktop systems I would imagine are slightly different (other than the obvious size difference), perhaps having more cache. I honestly dont know much about laptop IDE drives, but to cut to the chase, raid 0 on desktop IDE drives works very well for me.

Let me dig up the benchmarks.

SEMC, there are some very smart, and very helpfull people in here. I am not a regular contributor, but I do read the forum a lot. I can back Aussie up on his claims. Did you get microwaved from all that data center time?

Regards,

Tristan

- Looking forward to his laptop ordering party when someone invents 17" wuxga...

Not quite sure how to react

I'm not quite sure what to do about all of this wonderful debunking of just about everything I know about computer hardware I have learned either from personal experience (10 of my 23 years) or from my AS in Computer Science degree from Trident Technical College. These are my limited credentials. I am not quite ready to accept defeat on this one yet. I still want benchmarks as proof before I go to my professors and tell them to their face that they are all wrong. Wouldn't be the first time they were completely wrong (from misconceptions of Active Directory, wrong syntax and poor programming designs, and bad hardware configurations), but I can't believe that I can't trust any of my information I have learned from my courses.

Few questions from the experts.
1) If IDE RAID is so wonderful, why is it that I have only started seeing RAID as an option from major PC manufacturers in the past year alone? Was the limiting factor price alone or price to performance inefficency?

2) 10,000 rpm SCSI drives have been around for years. I find it hard to believe that IDE drives are limited to price alone instead of other physical limitations since IDE drives have only reached 7200 rpm within the last 8 years (?) and have firmly stayed there. Meanwhile SCSI has continued to increase in speed and performance and capacity.

3) What is wrong with the RAID 3 description. I have a test on that chapter in a few days and I'd like to not fail it just because the book is wrong.

4) Again if RAID 0 has any significacant benefits, why haven't I seens it used before in consumer level applications? You'd think companies would jump on any chance they could get to make money. The majority of my experience comes from the gaming side and the extreme modding/overclockers. The gaming extremists try to squeeze every last bit of performance out of their machines no matter how minute. I think it would be very odd that they would ignore such an obvious choice for increasing performance. Granted most gamers focus primarily on the CPU, Ram and the video card. But attention is also given to sound and controllers. Hard drives and storage are almost ignored. You either have SCSI or not. No mention of RAID anywhere. It should be noted that while I know others who overclock and mod, and read online about it; I myself do not overclock since I have only recently been able to afford my own machine with which to toy with.

Finally, my brain is okay and it has not been microwaved. I am just upset that I see post upon post from people, regardless of qualifications, flocking to RAID 0 and extolling the benefits as if it were some sort of manificent new toy or savior to solve all of their problems or needs. I got RAID 1 and I never have to back up again, or RAID 0 makes my machine lightning fast. I feel sometimes that I'm in Best Buy listening to the sales pitches.

SEMC, I posted my answers to your questions on the other thread here.

Just to reiterate, most of the reason you don't see high performance features in consumer disks is they cost more to produce. Consumers normally trade off performance for a lower price.

Hmm, well this will come as quite a surprise to my pair of SATA WD 10k rpm 74GB Raptors in RAID 0 in my desktop system :
Raptor Review
I get near equivalent to SCSI performance on my particular setup for a fraction of the cost. Very high performance SATA drives are here now. The price / performance ratio offered by this generation of IDE drive is going to put heavy pressure on the future of SCSI. Look for SCSI to become an even more marginalized option than it is now given the price/perfomance equation.

My own experience with on-board IDE Raid 0 recently has been very favorable, as contrasted to my nightmares of a few years ago using 3rd party IDE Raid cards. I now get blistering performance and rock solid reliability on my desktop workstation day-in day-out. Of course I do still back-up my workproduct to a non-raid 0 drive nightly, a moderate dose of paranoia never hurts .

I am very intrigued to give notebook RAID a good thorough evaluation, certainly won't believe someone tossing blanket statements around condeming the technology without the facts to back it up.

Bring on the numbers...

- Card

Some numbers for the doubting: http://www.hardwarecentral.com/hardw...eviews/1611/8/

Admittedly it isnt in a laptop nor does it use the same raid controller, but it does address the 'raid 0 does not help in an ide system' point.