bachviet

I just bought a cheap PCI RAID card (Silicon Image) and use it with my 40GB IBM & 60GB Maxtor (RAID 0). I lost 20GB but I gain in hard drive speed. I used to get around 31K with the 120GB WD SE but now I get around 38K with the RAID drive (using Sandra). Both drives are partitioned @ around 10GB for Windows XP and I use NTFS. Here is the pix:

__________________
Dell Dimension 9200 | Intel Core 2 Quad Q6600 (2.4GHz) | 4x1GB DDR2 | 256MB nVidia GeForce 8800GT

Dell Studio 17 | Intel Core i7-720QM (1.6GHz) | 2x2GB DDR3 1066MHz | 1GHz ATI Mobility Radeon HD 4650

Intel P4-C 3.0GHz | ECS 865PE-A | 3x512MB PC3200 | 128MB PNY GeForce 6600GT

Hiro

__________________
-Corsec-

PrObLy

I'll hvae to pick me up a couple of them, it's exactly what I've been looking for, are they usually available through their website for online ordering?

__________________
Facebook me if you so desire.
Guns don't kill people.
Chad Greenway kills people.

Jeffbx

Simple rule of thumb: If the card controlls the RAID, it's hardware RAID. If the OS controlls the RAID, it's software RAID.

LegendKiller

That card DOES do software RAID. Its much like WinModems, there ARE drivers for it, BUT there is no on-board processor to process the transactions the RAID controller needs to process.

More or less, the RAID interface on the board tells windows "Hey, I need this much CPU power to do these transactions". Windows then interrupts the processor and says "here, we need to access this disc, do the calculations". The processor then does the calculations, your RAID is accessed and you get the data.


The HUGE difference is this. When a HARDWARE raid controller needs to access that disc the interface tells the ONBOARD HARDWARE processor "Hey man, I need this data", the onboard processor accesses the data, calcuates the stipe and sends it to the OS through the interface.

What didn't you include in the second process? The CPU (Athlon or P4...whatever. A HARDWARE raid card has an onboard RISC or other processor (intel makes several, as does LSI and Adaptec) that does all of the RAID calculation. However, on your "cheap" raid board, the good ole CPU needs to process the RAID.


Difference? In terms of disc access they are both the same, more or less. A dedicated RAID controller might give you even more throughput and accurate calcuations instead of the jack of all trades, good at none CPU.

As far as CPU power being sucked, the difference in MASSIVE! Your typical IDE controller will use 7-12% of CPU power (SCSI comes in at 3-5). However, IDE SOFTWARE raid (what you have) can take up to 50% of your CPU power.

Furthermore, your disc access times go down, so when you are NOT getting huge files off of your HDD but many small files from different places on your RAID your actually slowing yourself down with RAID 0 when comparing it to a single normal HDD.

So, in conclusion, IDE Raid in general slows down your computer by increasing access times. Furthermore, IDE SOFTWARE raid further slows down your computer by killing the availible time for your CPU to work on other things.

You are taking 2 steps forward and 3 steps back....


LK

Jeffbx

Huh, you learn something new every day!

I never knew there was such a thing as a RAID controller with no processor...

Thanks for the info!

LegendKiller

Any motherboard raid controller does not have a processor. The ones that do are $150+ dollars.


LK

jase71

Of course, CPU cycles are the cheapest, most freely available, and most inexpensive to upgrade resource in your pc.

So gaining hard drive speed at the expense of CPU cycles is often a very favorable tradeoff.

Software RAID also offers greater flexibility than hardware RAID.

SCSI hardware RAID is the best pure performance solution, but as we've discussed before, IDE software RAID has some applications where it's a logical and viable alternative.

__________________
And I found that hope and a lucky card
were all I had to walk with me...

LegendKiller

How does Software RAID offer ANY flexibility? I do not see anything at all besides being cheap.


Ok, lets do a quick analysis of this...


I go out and buy a 1.8ghz Athlon (whatever the PR is). Now, in order to gain about 30% in HDD speed by doing RAID 0. However, I lose 40% of my CPU speed.

So, effectively, I have a 1.08ghz CPU. However, I have gained the ability to load Windows faster. BUT, anything that is CPU intensive AND HDD intensive (games, Photoshop, pretty much anything BUT transfers) I actually lose. Why? Because you need your CPU MUCH more than you need HDD speed...

Software RAID 0 is a waste of time unless you are doing strait transfers.


LK

jase71

No?

I've got two IDE drives. With those two IDE drives, I can have a RAID 0 partition for my OS to boost transfer speed. I can have a RAID 1 partition for redundancy for my /home directory in case of drive failure, to preserve my user accounts and user data. And I can have standard partions (non-raid) for things I need neither speed nor redundancy for, like print spool swap space, or non critical data storage.

Software RAID works on partitions, not soley on individual drives. That gives you the flexibility to use whatever form of RAID you choose, depending on the situation. Transfer intensive data can be placed on RAID 0 partitions. Important data can be placed on RAID 1 partitions for safety. And non-transfer intensive and non-critical data can be placed on standard partitions. ALL on the same two drives.

Try all that with 2 hard drives in hardware RAID.

Quote:

I go out and buy a 1.8ghz Athlon (whatever the PR is). Now, in order to gain about 30% in HDD speed by doing RAID 0. However, I lose 40% of my CPU speed. So, effectively, I have a 1.08ghz CPU. However, I have gained the ability to load Windows faster.

First off, I don't lose 40% cpu speed. My numbers fluctuate between 15-25%, as noted before in the earlier thread. That's not painfully higher than standard IDE.

Second of all, you don't lose those cpu cycles all the time. You lose them for the time spent in the high volume transfers.

The rest of the time, your cpu is operating at its full potential. So it's very misleading to categorize your CPU as falling from a 1.8 gig CPU to a 1.08 gig CPU. It's still a 1.8 gig cpu. And it's still giving you the full 1.8 gig the majority of the time.

Not to mention, booting Windows faster can be a big perk, since the average Windows user ends up doing a lot of it.

Quote:

BUT, anything that is CPU intensive AND HDD intensive (games, Photoshop, pretty much anything BUT transfers) I actually lose. Why? Because you need your CPU MUCH more than you need HDD speed...

Do you?

Let's take the 40% number you claim. Let's assume I have to make up that 40% cpu hit.

What does it cost me to set up an 80 gig SCSI Hardware RAID array?

What does it cost me to set up an 80 gig IDE RAID array with a 40% faster processor? Care to guess which one is cheaper? Not to mention, for the majority of time, you'll have a 40% faster CPU.

It's much cheaper to offset that 40% of cpu time( even though 20-30% cpu utilization is a more fair estimate given today's increasingly speedy cpus) than it is to make up that speed on the RAID side.


Software RAID (which does not just include RAID 0) is a viable alternative in some situations. It's more cost effective, and more flexible. What it lacks in reliability, it more than makes up in cheap speed, flexibility, and accessibility.

It's not an enterprise solution, but for home use, or small office use it can be a legitimate alternative as long as you understand the caveats.

Jeffbx

So maybe I'm a little confused here -

If these cards are just running software RAID, what's the purpose of buying them?? Why not just let the OS handle the RAID, since the PC's processor is handling the transactions anyway?

I guess I could understand for an OS that doesn't support RAID...

Merlin

If I am running Win2k as my operating system and am looking for a raid0 soultion to help boost performance of my os (faster boot times, faster access to the swap file, etc) which would I need to go with. I presume that software raid needs to have the os running and therefore won't help me. Is that correct? And if I am looking for a raid solution for my os drive(s) what should I look for?

Note - I've got 2 40.0gig 7200 Maxtors just waiting for the right solution.

__________________

jase71

If you want to boot the OS off the RAID partition, you'll need to pick up one of the many IDE RAID cards out there. By loading the appropriate driver during the Win2K install (S to specify driver, and all that...) you can install the OS to the RAID Array you've created.

The card essentially is just the interface, and doesn't do any of the processing. But it DOES allow you to install the OS to the software RAID array.

You can also run software RAID without a seperate card, using just the standard IDE controller. However, you cannot install the OS to the RAID array in this case. The OS must be on a non-RAID partition. In linux this is no big deal, since you can boot the kernel off of a small boot partition and load the bulk of the OS off the RAID array, essentially like booting directly off the array. However, in Windows, this is a bigger limitation, so a seperate card is a good idea.

If you're looking for a RAID solution for your two 40 gig drives, pick up an IDE RAID card, set 'em up, and go. But do a little digging into which card is best for your purposes. I use linux, so my experiences won't necessarily apply to your situation.

LegendKiller

Actually, you CAN do that with hardware RAID. 3ware and LSI make cards that you can raid between partitions. However, that is like buying a car with an 8-cyl engine and disabling 2 of the cyl's. Why? Because partitions are taking different tracks and separating them. However, when you stripe between those tracks you are also increasing the amount of seeks the head has to do for one string of data.

Lets make an example. You have your 6MB MP3 file, on a 32KB stripe you have that MP3 broken into 192 stripes. So, in order to read that file the head will need to keep switching back and forth ever 32KB's. This makes no sense at all.

The WHOLE point of RAID is to take the burden of reading data OFF of one head and switch it to two. Thus you can have one head reading and another writing, two writing, to reading, or other combinations. Thus, you get a transfer rate that is much higher.

My example is like having 2 people trying to pick the alphabet out and put it into sequential order. 2 people with 2 bowls, coordinating can do it MUCH faster than 1 person with 1 bowl. However, what your saying is that you should only have 1 person with 1 bowl only able to use 1 hand...doesn't make sense.

Furthermore, the other reason for RAID is redundancy. Why the heck would you want to Mirror to another partition? The whole point of mirroring is so that if one drive goes out, you still have the mirror on the other. If you put a mirror on one partition, the only gain would be is if the one partition became corrupted (something that doesn't really happen) then you have the mirror on the other that you can switch to. However, I do not think this would work anyway, especially since the FAT table wouldn't recognize that second partition as bootable.

Your arguments for software raid do, I am sorry to say, make any sense at all. Furthermore, in even more CPU intensive RAID's, which most software RAID cannot do, you NEED a hardware raid, such as RAID 5, RAID 0+1, or RAID 7. Calculating XOR's and other parity scheme's is impossible without a dedicated raid processor. Again, your argument for software raid does not make any sense.

Quote:

First off, I don't lose 40% cpu speed. My numbers fluctuate between 15-25%, as noted before in the earlier thread. That's not painfully higher than standard IDE. Second of all, you don't lose those cpu cycles all the time. You lose them for the time spent in the high volume transfers. The rest of the time, your cpu is operating at its full potential. So it's very misleading to categorize your CPU as falling from a 1.8 gig CPU to a 1.08 gig CPU. It's still a 1.8 gig cpu. And it's still giving you the full 1.8 gig the majority of the time. Not to mention, booting Windows faster can be a big perk, since the average Windows user ends up doing a lot of it.

I used to run a software RAID0, and I set up a log file which recorded my CPU utilization throughout different events. A friend of mine, who is a master programmer was also able to write a program which tracked which processes took the most CPU usage (driver level processes). We tracked CPU utilization from the Highpoint DLL and found that during certain tasks, like Quake3 texture loading, Photoshop calc's, Windows loading, and other things that the CPU was being utilized by 40%+ during many of these tasks.

We also found, through PC benchmarks as a whole that many times a software RAID LOWERS your PC mark. Why? Because they benchmark the system as a whole, not just transfer rates. Thus, if you want to use your system as a whole, you are losing.

[QUOTE}


Do you?

Let's take the 40% number you claim. Let's assume I have to make up that 40% cpu hit.

What does it cost me to set up an 80 gig SCSI Hardware RAID array?

What does it cost me to set up an 80 gig IDE RAID array with a 40% faster processor? Care to guess which one is cheaper? Not to mention, for the majority of time, you'll have a 40% faster CPU.

It's much cheaper to offset that 40% of cpu time( even though 20-30% cpu utilization is a more fair estimate given today's increasingly speedy cpus) than it is to make up that speed on the RAID side.
[/quote]
What I really get a kick out of is these people trying to overclock to make their games faster, but they are killing themselves by running software RAID. Furthermore, why would you want to buy a processor for 220+(current price for Athlon XP +shipping from a reputable place) bucks, spend another 40 on a good HSF (also with a VERY loud fan...), overclock it by 20% and then end up losing that AND 20% MORE from software RAID? It makes no sense at all. The way your saying is that it does, so lets do this.

Lets calculate what you actually get. 100+100+30 for the drives/card, 260 for CPU/HSF. Total investment for a software RAID with the best CPU. Now, since your killing your overclock by 20% AND your extra CPU power by 20% your reduced to an Athlon XP 200-, which is a full 80 bucks cheaper. Then, how much do you value reliability? To me, its worth whatever amount of work I put into installing, and then keeping papers, GotApex reviews, and other such things, not to mention RMA'ing a dead HDD. I get paid more than 25 bucks an hour currently. Thus, an install of WinXP with full programs, takes me more than 4 hours. Then, you have to include RMA'ing the drive, crosshipment takes care of time, but then I have to pay to get it back, another 20 bucks+time on phone=45 more dollars. Then, since I lose my papers + gotapex stuff, I would say thats another 15 hours of work per semester (I am a graduate student), thats another $375. Then we get pain and frustration for losing all of that, which I would put at about 100 bucks (plus damage from the sledgehammer when I hit the stupid computer for dying..).

So, whats the butchers bill? For my time lost, data lost, RMA time, and then money I have to spend on my GF so she isn't so pissed at me for spending 2 days fixing my computer...and not to mention the CPU time you lose from running IDE raid...It all comes to 620 bucks, PER incident. I have lost data 3 times, so your talking almost 1900 bucks IDE RAID0 cost me. Why not run RAID10? Why not just buy a SCSI raid card with 2 SCSI drives and run RAID0, you get the same reliability and 2x the performance of RAID0.

Your point for a 80gig SCSI RAID array, again, makes no sense at all. What do you gain from Software IDE RAID? Transfer rate, and some dedundancy...but not both. 90% of the people I see running Software RAID do not use 4 drives (striped mirrors), they just use RAID0, like this gentlemans case.

Thus, you are just gaining speed. The cost of 2-IDE 80gig drives is ~100 bucks. You get a nice, large, 160gig IDE raid0 partition, cool. HOWEVER, its very unreliable, can die any time and you lose EVERYTHING! What good is that? Furthermore, you cannot compare SCSI raid to IDE raid. Why?

1. The cost of SCSI raid is higher because they have onboard processors, large cache's, and demand the higher premium for SCSI.

2. The drives are faster. If you get 41MB's from TWO IDE drives, you can get that +10MB's, from ONE 36LP. With SCSI raid you would get probably about 80MB's from two 36Lp's (I might have to try this out...maybe Seagate and LSI would work with me...). Yes, 80MB's...why? Because you also get 32+MB of cache on the SCSI raid card, thus increasing your burst speed and overall transfer rate.

You are comparing a Toyota Camry to a Mercedes SL500 and saying you are going to buy the Mercedes because its cheaper AND you get more speed. However, you dont, and why would you even compare the two cars? They are in totally different brackets....

A more fair comparison would be with ONE 36LP AND ONE 80gig IDE. Thus, you pay 120 for your SCSI controller, 300 for your 36LP and 100 for your IDE. Total investment of 520. Two IDE's and a controller is 220. Sure, 300 buck difference. HOWEVER, is reliability AND speed important? ONE IDE has a higher failure rate than ONE 36LP, TWO IDE's have a failure rate thats about 3x a SCSI... Take your chance if you want, I lost too much data to do that again.

Quote:

Software RAID (which does not just include RAID 0) is a viable alternative in some situations. It's more cost effective, and more flexible. What it lacks in reliability, it more than makes up in cheap speed, flexibility, and accessibility. It's not an enterprise solution, but for home use, or small office use it can be a legitimate alternative as long as you understand the caveats.

I know Software RAID doesn't just include RAID0, however, thats what most of these people are using it for. Hardly any at all use it as RAID10, or even RAID1. Is it cost effective? Sure. Is it flexible? NO! It doesn't give you that many options, surely not RAID 5, which is probably the best raid you can run.

Sure, you can run a RAID0, but you are taking your chances. All your really doing is betting that you will be able to run faster while keeping your data. Personally, I would rather spend a little more and buy some quality stuff than let my data go out the door.

I HIGHLY recommend that anybody who is running RAID0 to backup their data DAILY and never trust anything on a RAID (unless its 0, 0+1 or 10). Either that or dont use it at all...

LK

jase71

So... is this a problem with the RAID array? Or a problem with your choice in stripe size not being tailored to the type of data? Seems to me this is much less of an issue if you properly configure your array, isn't it?

Quote:

Furthermore, the other reason for RAID is redundancy. Why the heck would you want to Mirror to another partition? The whole point of mirroring is so that if one drive goes out, you still have the mirror on the other. If you put a mirror on one partition, the only gain would be is if the one partition became corrupted (something that doesn't really happen) then you have the mirror on the other that you can switch to. However, I do not think this would work anyway, especially since the FAT table wouldn't recognize that second partition as bootable.

Who says it has to be bootable? In my case, if I had a RAID 0 partition for my OS, and a RAID 1 partition for my /home directory, and a drive fails, I can simply replace the dead drive, recreate my RAID 0 partition, restore my OS backup, and then recover my RAID 1 data from the original working drive. This is the way I upgrade my OS, and it works quite nicely, thank you.

I can survive a hard drive crash, with my critical data intact, and be fairly quickly back to original operating condition.

As easy as hotswapping a SCSI drive in a RAID 5 array? No... but then, I don't have nearly as much money invested, do I?

Quote:

Your arguments for software raid do, I am sorry to say, make any sense at all. Furthermore, in even more CPU intensive RAID's, which most software RAID cannot do, you NEED a hardware raid, such as RAID 5, RAID 0+1, or RAID 7. Calculating XOR's and other parity scheme's is impossible without a dedicated raid processor. Again, your argument for software raid does not make any sense.

I can run software RAID 5 on my system. Doesn't much pay obviously, because the cpu utilization does rise dramatically.

But then, who the heck needs a hardware RAID 5 array at home? And if they do, let them spend the money. For those who don't, there's a less expensive alternative.

Gosh, claiming that software RAID is worthless because Hardware RAID 5 is better is sort of like claiming a Subaru WRX just plain 'ol sucks because it's no Ferrari F40. Sure, the Ferrari is the better car... but it costs a LOT more, and it's just not pratical, feasible, or realistic for most people.

Whether or not it's valuable to be able to run RAID 5 or RAID 7 really only applies to those people who have a NEED to run RAID5 or RAID7, doesn't it? How many home users/hobbyists have a legitimate need for the expense of RAID types that are more suited to enterprise applications?

Quote:

I used to run a software RAID0, and I set up a log file which recorded my CPU utilization throughout different events. A friend of mine, who is a master programmer was also able to write a program which tracked which processes took the most CPU usage (driver level processes). We tracked CPU utilization from the Highpoint DLL and found that during certain tasks, like Quake3 texture loading, Photoshop calc's, Windows loading, and other things that the CPU was being utilized by 40%+ during many of these tasks.

That's nice. What CPU? Matters a lot, doesn't it? How much memory? Matters a lot. What OS? Matters a lot. Even the RAID chipset matters a fair amount. How does that compare to cpu utilization of a single IDE drive under similar circumstances? Matters a lot.

40% of cpu utilization on a P3 1 gig chip does not mean you'll see 40% cpu utilization on your P4 2 gig.

What percentage of the time was the cpu actually being utilized at 40%? If the CPU hits 40% utilization for only 5% of the time, that's a small price to pay, isn't it? If it's 80% of the time, you either need more memory to quit swapping, of you're in the market for a much higher end system, and SCSI might be wise.

Rather than making a blanket statement that one standard is always best, isn't it wiser to calculate the needs of the individual, and tailor your recommendations to those needs?

Quote:

What I really get a kick out of is these people trying to overclock to make their games faster, but they are killing themselves by running software RAID.

Who's talking about overclocking?

Although, I suppose you could make the argument that overclocking just makes the CPU hit "free". If I take an additional 20% cpu hit with software RAID, but I overclock my pc by 20%, I'm right back where I started, but with faster transfer. Sounds like a deal to me.

But I don't overclock. Well not much. I'm running 48mhz over stock right now. I'd rather have a quiet system with stable temps, without the expense of radical cooling solutions.

Quote:

Lets calculate what you actually get. 100+100+30 for the drives/card, 260 for CPU/HSF. Total investment for a software RAID with the best CPU. Now, since your killing your overclock by 20% AND your extra CPU power by 20% your reduced to an Athlon XP 200-, which is a full 80 bucks cheaper. Then, how much do you value reliability?

Let's try it with real numbers from a specific site. And lets toss out the $40 heatsink and fan bit, since we're not talking about overclocking, unless you want to apply the same to the SCSI system and overclock it as well. All prices from Newegg.

2 Seagate Barracude IV 40 gig drives. - $78 x 2 ($156)
Tekram Highpoint IDE RAID controller (RAID 0,1,0+1) - ($33)
Athlon 2100+ Retail box (1.73Ghz, no HSF needed) ($165)
Total ($354)


2 Atlas 10K 36GB drives - $235 x 2 ($470)
Intel ServerRaid controller - ($230)
AMD Athlon 1.1 gig cpu (~60% of 1.7 gig) ($60)
Total ($760)

How much do I value reliability? Considering I'll swap these drives out long before they run out of their 3 year warranty, I'm pretty safe, financially. At any rate, I could buy quite a few replacement hard drives for the $400 price difference.

Now.. those 10K SCSI drives are gonna make for a faster RAID array than my 7200RPM IDE drives. But for any time NOT spent transferring data, my 1.7gig cpu is gonna pummel that 1.1 gig chip by the same kind of margin.

As for my data, proper backups isolate the risk there. There's no excuse for data loss if proper procedures are followed.

Quote:

To me, its worth whatever amount of work I put into installing, and then keeping papers, GotApex reviews, and other such things, not to mention RMA'ing a dead HDD. I get paid more than 25 bucks an hour currently. Thus, an install of WinXP with full programs, takes me more than 4 hours. Then, you have to include RMA'ing the drive, crosshipment takes care of time, but then I have to pay to get it back, another 20 bucks+time on phone=45 more dollars. Then, since I lose my papers + gotapex stuff, I would say thats another 15 hours of work per semester (I am a graduate student), thats another $375. Then we get pain and frustration for losing all of that, which I would put at about 100 bucks (plus damage from the sledgehammer when I hit the stupid computer for dying..).

That's nice. I prefer to use backups. That way I don't have to do full reinstalls, nor risk losing papers, or even spend a ton of hours rebuilding my system if something goes down. I have a cron job scheduled to back up my critical data to a network drive. Highly critical things are burned to CD monthly in addition. Thus, at any one time, I have 3 potential sources for my critical data.

If you want to add all that extra work into the cost, well, that's your prerogative. But it's unnecessary, and inefficient.

Quote:

So, whats the butchers bill? For my time lost, data lost, RMA time, and then money I have to spend on my GF so she isn't so pissed at me for spending 2 days fixing my computer...and not to mention the CPU time you lose from running IDE raid...It all comes to 620 bucks, PER incident. I have lost data 3 times, so your talking almost 1900 bucks IDE RAID0 cost me.

Is it fair to chalk your failure to adequately back up your data against the price of the hardware?

Quote:

Your point for a 80gig SCSI RAID array, again, makes no sense at all. What do you gain from Software IDE RAID? Transfer rate, and some dedundancy...but not both. 90% of the people I see running Software RAID do not use 4 drives (striped mirrors), they just use RAID0, like this gentlemans case.

Actually, with a RAID 0 partition for the OS, and a RAID 1 partition for my /home directory, I certainly CAN gain both.

I can gain speed on the RAID 0 partition, and security with my /home partition on RAID 1. If one drive fails, I restore/reinstall my OS on RAID 0, and recover the data on the RAID 1 partition.

Quote:

Thus, you are just gaining speed. The cost of 2-IDE 80gig drives is ~100 bucks. You get a nice, large, 160gig IDE raid0 partition, cool. HOWEVER, its very unreliable, can die any time and you lose EVERYTHING! What good is that?

Speed is good for me. I like speed. That's a good reason to do it.

And with my backup scheme, I really don't risk losing that much. Oh, an hour or so of recovery, maybe two. But that won't kill me.

Quote:

You are comparing a Toyota Camry to a Mercedes SL500 and saying you are going to buy the Mercedes because its cheaper AND you get more speed. However, you dont, and why would you even compare the two cars? They are in totally different brackets....

EXACTLY! And yet you're trying to tell us that SCSI is always the best alternative, and always better than IDE RAID.

Is a Mercedes SL500 a better choice for an economical family car than a Camry? Of course not. They both have their places.

Is the Mercedes a better car? Sure. But sometimes, the less expensive Camry is a perfect fit for a family on a budget. Just like IDE raid can be a perfect fit for someone on a budget.

Blowing a ton of money on the best equipment isn't always an option, especially for those of us who are married and/or have kids. House payments, bills, vacations, those sorts of things take precedence. We'd all love to have SCSI RAID 5 arrays in our systems. But it's not practical, and it's not cost effective, just like it's not cost effective for most of us to buy a Mercedes SL500, now matter how much we might want one.

So we go with the less expensive option, and plan accordingly.

Quote:

Sure, you can run a RAID0, but you are taking your chances. All your really doing is betting that you will be able to run faster while keeping your data. Personally, I would rather spend a little more and buy some quality stuff than let my data go out the door.

Personally, I'd rather back my data up properly, and save myself some money, while still getting a nice speed boost. SCSI drives fail too, even if it's less often. If you're not backing your data up, you're still vulnerable to data loss. It just means the equipment that fails will be more expensive.

LegendKiller

Well, what it comes down to is you saying IDE raid is good to save money and get some sort of redundancy.

I say why not just run 1-drive scsi, that way you dont have to worry about that much redundancy, you get faster transfer speeds than IDE raid, you get no CPU overhead (and even lower than IDE regular), and its not THAT much more (120+309 compared to 200+30).

Well, whatever. Some people would rather take the chance and deal with backups. I would rather deal with small backups, take no chance, and still have more speed.

Financially, it makes much more sense to deal with SCSI, especially on the enterprise level. Even if IDE were closer in speed, and you used a hardware raid card, it still would make less sense to use 2 drives to do the work of 1.

Will I ever run software raid again? Nope.
Will I recommend to anybody that it be run? Nope.


LK

jase71

1 drive SCSI still puts all your eggs in one basket. It lessens the odds of failure, but it doesn't eliminate them. Your data is still at risk for data loss from the failure of just one drive. I would never run a single drive system without frequent backups. Mirroring my data partition and running frequent backups leaves me less vulnerable to data loss than running a single SCSI drive.

Quote:

Financially, it makes much more sense to deal with SCSI, especially on the enterprise level.

Yep, and when I'm running an enterprise server, or making hardware recommendations for one, I'll recommend SCSI. I've done it before, and I'll do it again. I actually like SCSI a lot. But for some applications, it's just plain 'ol overkill.

When I'm looking for some simple performance boosts and/or redundancy for a home pc on a budget, I'll go with something more cost effective, like IDE RAID.

Quote:

Will I ever run software raid again? Nope. Will I recommend to anybody that it be run? Nope.

That's your prerogative.

Personally, I would prefer not to rule out an entire class of hardware and standards just because of my personal bias against them.
Sometimes there may be an application where they fit perfectly into the needs of the user, and if I ignore a solution merely because of my personal bias, I do the user/customer a disservice, not to mention cost them more money.

To each his own.

sbp

Looks like GAM will have to increase the message length again.

jase71

Bah, you're just jealous of the 31337 H4x0r typing skills LK and I 0wnz...

I've overclocked my fingers by 23%...

LegendKiller

HAH! I have you beat, I am running them in RAID0 by using my toes at the same time! MUAHAHAHAHAH!


LK

jase71

Oh yeah? Well I'm typing this with my d....

...never mind.