Anyone Know much about RAID.....................

[BritishBeefEater]

Currently I have the Abit Hot Rod Raid Controller Card (UDMA/ATA 100) with 2 20.4Gb Hard Drives also UDMA 100, runningf Raid 0. However my current motherbaord is only UDMA 33. Does this reduce the performance of the Controller Card? (My guess is that it does - I ran some bench Marks)

Therefore I am considering upgrading my motherbaord and processor to the Abit KT7A and an AMD 1Ghz. My main concern is....

If i bought the KT7A-RAID (onboard raid card) would I need to reformatt the Hard Drives in order for them to work with the Onboard Raid Controller? Also is the performance of the Onboard raid better due to it not going through a PCI slot?

Its whether i should pay the extra and have (a hopefully faster) onboard raid controller. Or whether the difference in performance between the onboard and PCI Controllers is so minimal that I would only need to buy the standard KT7A (without Raid)

Any help/advice would be appreciated

 

[ElFenix]

it kills bugs really quick... they see it, then yell, "ra-aid," and poof!, they're dead!

oh... wait...

okay, your ata 33 controller (i'm guessing a bx chipset, but not sure) is completely independent of the hot rod controller. completely. they are two different devices on the pci bus.

that said...
if you put those hard drives in a new mother board and your operating system is on those hard drives, you will need to reinstall. operating systems don't like having that big of a component swapped out. the on-board raid still uses the pci bus, just like a card in a pci slot would, so it would not be any faster due to that. so the difference between the two is minimal.

lastly, the hot rod isn't all that great a controller from what i've heard (i'm running the promise myself). and RAID on an IDE controller like that is simply software RAID, even windows 2000 can do this with the regular IDE controllers in the south bridge on the motherboard. so its not terribly fast.

and if you manage to get something with a hub-like architecture (not gonna happen since you're not getting an intel board and via won't have something like it for a while) then the southbridge's (I/O Hub in intel parlance) ata ports are actually faster than a pci based solution. the reason is that the new hub architecture gives the southbridge a dedicated bus to the northbridge thats twice as fast and also less congested than the pci bus.

 

[BritishBeefEater]

So your telling me that if I got an ATA 100 motherboard i would be running faster without the Hot Rod Controller and just using my 2 Hard Disks through the ATA ports on the actual motherboard???

I was also under the impression that this was not software RAID!! but then again i don't know much....

I also didn't make my system spec quite clear, my motherboard itself us only UDMA 33!

thanks anyways

 

[PCResources]

<< RAID on an IDE controller like that is simply software RAID >>



And so is the Promise card you are using, both are software.

The only HW based IDE RAID card i know of is the Adaptec.

Patrick Palm

Am speaking for PC Resources

 

[BritishBeefEater]

ok ok I believe you,

but do recommned using that software IDE Raid with a 1Ghz Athlon and a ATA 100 Mboard, or would it be faster to just use the ATA ports on the mboard?

 

[MasterMind]

The definition of RAID disk types are:

Software based

The array is managed by software running in a host computer using the host CPU bandwidth. The disadvantages with this method are the load on the host CPU and the need for different software for each operating system.

SCSI to SCSI

The array controller resides outside of the host computer and communicates with the host through a SCSI adapter in the host. The array management software runs in the controller. It is transparent to the host and indepentdant of the host operating system. The disadvantage is the limited data transfer rate of the SCSI channel between the SCSI adapter and the array controller.

Bus-Based

The array controller resides on the bus (for example, a PCI or EISA bus) in the host computer and has its own CPU to generate the parity and handle other RAID functions. A bus-based controller can transfer data at the speed of the host bus (PCI, ISA, EISA, VL-Bus) but is limited to the bus it is designed for.


These definitions are from before IDE RAID and does not define IDE RAID at all. IDE RAID is a hybrid controller incorporating features of both software and bus-based RAID solutions. The very definition of bus-based RAID is that it &quot;can transfer data at the speed of the host bus&quot;. The PCI bus can tranfer data at up to 132 MB/s. But it also uses the CPU for calculations controlled by the hardware on the controller card. Don't sell it short, IDE RAID works!

 

[BritishBeefEater]

So with a 1Ghz Athlon, 256mb PC133 RAM, and the IDE Raid Card with two ATA 100 HDDs (Raid 0) I should get faster HDD access than I would wihtout the controller, and the effect upon the system should be minimal due to the large amount of RAM and fast CPU - correct?

 

[PCResources]

<< Don't sell it short, IDE RAID works! >>



I agree, it does work, it's just that RAID 0 is risky (which is what most IDE RAID users will be using), IDE in itself uses CPU power, then add the CPU power required for the RAID calculations (which in fact uses as much CPU power as the SW solution in NT4/W2K) and you got a system that is crippled.

1. Less CPU power
2. Higher seek times
3. More than twice the risk of loosing data, and not just a little piece of info, ALL of it. (and NO, it doesn't require a HW crash, just a minor error in the right place will do it)

Patrick Palm

Am speaking for PC Resources

 

[smn198]

Use a hardware raid controller like the one you mentioned on the mother board. That way you don't use system resources like you would with software raid (so hardware is faster).

 

[PCResources]

<< Use a hardware raid controller like the one you mentioned on the mother board >>



That controller is not a HW based solution, it is as much SW based as the controller he is using now.

Patrick Palm

Am speaking for PC Resources

 

[smn198]

Use a hardware raid controller like the one you mentioned on the mother board. That way you don't use system resources like you would with software raid (so hardware is faster).

I also think I remember hearing something about RAID 1 (mirroring). With some RAID cards and identicall hard disks it is possible to run each disk half a revolution out of sink whith the other. This would therefore reduce (1/2?) the seek time for reads with a performance hit for writes. I would be interested in doing this but is it only available on SCSI or with special HDDs or was my source speaking BS?

 

[PCResources]

<< I also think I remember hearing something about RAID 1 (mirroring). With some RAID cards and identicall hard disks it is possible to run each disk half a revolution out of sink whith the other. This would therefore reduce (1/2?) the seek time for reads with a performance hit for writes. I would be interested in doing this but is it only available on SCSI or with special HDDs or was my source speaking BS? >>



No, the seek time stays pretty much the same, with a slight decrease. The transfer rate on reads goes up about 45% though, and the risk for data loss is slightly less than 100%. It is as possible to use as the RAID 0 config on all IDE RAID solutions, no need for special drives.

Your source needs to know what he is talking about, which he obviously doesn't, ask him which &quot;special drives&quot; that can do mirroring he is referring to.

Patrick Palm

Am speaking for PC Resources

 

[chuckieland]

so who make the total Hardware base Raid card?

 

[PCResources]

<< so who make the total Hardware base Raid card? >>



For SCSI, there are many HW RAID solutions, for IDE, only Adaptec (as far as i know, i haven't had a chance to test it out yet, but the specs look right).

Patrick Palm

Am speaking for PC Resources

 

[Pariah]

3Ware and Promise also produce hardware IDE RAID solutions.

CPU utilization is an overused term when referring to RAID0. RAID0 is very simple and barely taxing on the CPU at all, especially in a typical 2 or 3 drive array. ARSTechnica did a comparison between a Promise FastTrak and WinNT software RAID and in 2 drive arrays, NT (which is obviously straight software) beat the Promise card in CPU utilization. Both were under 10% on what I believe was a Celeron 450 which is slow by today's standards. As drives were added, the Promise card performed better but also had quite a bit higher CPU utilization. I would say that the higher througput had more to do with the higher CPU utilization than anything else.

 

[MasterMind]

<< and the IDE Raid Card with two ATA 100 HDDs (Raid 0) I should get faster HDD access than I would wihtout the controller? >>



IDE &amp; SCSI RAID will do nothing to improve access or seek times. It will only increase the data transfer rates. For OS performance the access time is more important than transfer rates. There is a (OS) performance increase when running IDE RAID since a single disk can't saturate the PCI bus but the performance is minimal. There is a unbelieveable difference (2X faster) when tranfering files or writing to the drive. With the MTBF rates over 500,000 hours I think the chances of losing data due to drive failure is about the same as winning the lottery. Setting up and maintaining a array is a different story. I wouldn't run a RAID 0 array as the primary OS drive without having a backup on a different drive.

 

[PCResources]

<< I wouldn't run a RAID 0 array as the primary OS drive without having a backup on a different drive. >>



You almost got me thinking that you were one of these &quot;RAID 0 is totally safe&quot; kind of guys, i have had too many arguments with those guys. ;-)

I totally agree with you, with a current backup (and don't fool yourself, you will need it some day) RAID 0 is great for most people.

However, don't expect a 100% increase in read or write transfers, 50-75% is more realistic (if you are using your computer for other things than benchmarking your HDD subsystem.


Patrick Palm

Am speaking for PC Resources

 

[BritishBeefEater]

But still (orignial question/s) Is my RAID system slowed by the slow/oldness of my motherboard - being a bx chipset at UDMA 33 (i think)???

And I would be better off running through the Controller card if I have one (which i do) than abandoning it and reverting back to 2 seperate HDD through the conventional ports????????

 

[MasterMind]

<< don't expect a 100% increase in read or write transfers, 50-75% is more realistic >>




Actually, I've noticed > 80% increase in write transfers with a IDE subsystem and from SCSI bus to IDE RAID it is nearly 100%. I'm talking performance I've noticed with Windows and DOS transfers and a/v capturing, not benchmark results.

BBE


<< okay, your ata 33 controller (i'm guessing a bx chipset, but not sure) is completely independent of the hot rod controller. completely. they are two different devices on the pci bus. >>



Your controller card is running UDMA 100 and is not limited to the UDMA 33 of the m/b.

Whether you run RAID 0 or not is a personal decision depending on what &quot;performance&quot; you're looking for.

 

[cvlegion]

I remember that Windows 2000 SW RAID is strictly data drives. It cannot be used up as a boot drive, which sucks since this is at least the purpose of RAID 0, so that you programs will run superfast and your computer will startup pretty fast. Who cares about data being fast anyways as long as it doesn't take forever.

I also know that these IDE RAID cards are software based. However, they don't have the deficiency of not being able to boot off of them. I have a hacked Promise 66.

 

[BritishBeefEater]

Well Can i ask this then........

I ran Norton HDD benchmark on my Raid ATA 100 system with the old m/b, and it came out lower than a m8 of mine using ATA 66m/b and single HDD, granted it was higher than the ATA 33 HDD i have as well, but is this not where raid excels???

What applications are gonna be most helped by having a Raid based HDD system??

 

[Pariah]

Cvelegion, I wasn't comparing NT to Promise, just showing that software RAID, of which both are, does not kill the CPU in RAID 0 arrays. As fot NT software RAID, it is bootable off of RAID 1 which is a true RAID, unlike RAID 0. I fail to see what the fascination is with faster boot times. How often a day do you reboot? And most of the time you will gain from a RAID array is lost during the added time to initialize the array during boot.

&quot;Who cares about data being fast anyways as long as it doesn't take forever.&quot;

You've missed the whole intention of RAID with this statement. RAID is intended for servers, A/V editing, databases, which are all strictly mass data reads and writes. RAID was never designed with the intention of making home user's system boot faster.

If you want programs to run &quot;superfast&quot; you should first buy more RAM. You will get a small additional boost by moving everything but your OS to the RAID array.

Norton is a horrible HD benchmark, right up there, or down there, with Sandra.

 

[Dameon]

Bottom line - the BX chipset is not the problem in the equation.

#1 - If you are using Earlier rev Maxtor drives, they are notorious for problems in RAID configurations.

#2 - This is very much a mostly software controller, much like a winmodem. Technically it is a piece of hardware, but almost all of the information flow is being handled by drivers / software.

 

[cvlegion]

Okay, Pariah, I completely understand and know what you are talking about.

However, you did not understand my intention. It doesn't matter what RAID was initially intended for; the fact is that most of the people that will run across this string of messages who are using RAID will be using striping. This was the reason IDE RAID was created: to make a cheap solution and bring it down from being an exclusive solution for servers and workstations to high-end and even mid-range performance home users. You talk about servers, in fact, if these IDE RAID/SW RAID solutions were meant for servers, why is it that among these RAID solutions, there is only one RAID 5 solution, Windows 2000 Server. My statements were directed to the majority of forum users who do not do audio/video editing and do not run servers.:disgust:

You also denigrate the fact that I might like my computer to start up faster and you assume that I have POS RAM because of my desire for my programs to boot up faster. You assume too much. I have 256Mb of Mushkin PC150; my FSB is 150MHz on a BX by the way. I also have a stick of PC2100 waiting for my upcoming dual Palomino rig. You ignore the fact that it takes a lot of time in computer terms for program files to load up from the HDD to memory. This is the same with startup, where the computer must load all files from the HDD to memory.:frown:

 

[BritishBeefEater]

So...........

When i get 256Mb of PC133SDRAM I should start seeing more out of Raid???

Also could it be slow due to: 'SPD Not Found on Dimms 1 and 2' Error - which now always comes up on boot up..... I have noticed that I sometimes don't have enough memory to run winamp staright from boot up (ie the first application)!!!!