Your thoughts on RAID...

[utopia]

RAID topic.

What?s your favorite RAID set up (0, 1, 2, 5, 7, 10)? How many people are running RAID 0? What HD's are you using, and how many in an array? What controller? How do you like it?

I know this sounds like a survey but I am just curios because I am building a new RAID 0 set up with Dual 75GXP?s.

 

[Windogg]

RAID0 suck IMHO. Costs more than a single large drive, is NOT true RAID, and doesn't have true gains performance for average users. It a nice novelty to have and tell yout friends about the neat setup you have. My biggest problems it that one failed drive will kill the entire array with no possibility for data recovery. Many with RAID0 do not take proper precautions such as having a routine back up.

At work I administer 3 servers (1 with 5 x 36GB IBM UltraStars, 2 w/ 5 x 18GB Quantum Atlas) with RAID5 and 3 with RAID1. The controllers are all Intel i960 based. I really love true RAID since it makes everything fault tolerant. In a few moths I will be installing hot spares as a preemtive strike against hardware failure.

Windogg

 

[AMDfreak]

I love my RAID 0 setup. HD Tach results give me average speed over 60MB/s and burst off the chart over 80/s. While it's not "true" RAID, it is hardware assisted so that you can boot from it. The CPU does most of the work, but with the latest drivers/BIOS CPU utilization is around 25%. This is a fair tradeoff IMHO for the big speed boost. Us home users/gamers don't really care much (not as much as a server admin anyway) about losing data. We can back up what is important easily on a CDRW or other removable media. Windogg, good point about back ups, but most people with a single drive don't do proper back ups either. As overclockers we look for anything and everything that can boost more speed from our system. I've found in my case RAID 0 has helped alleviate the biggest bottleneck in any system.

 

[pdo]

I had a RAID-0 setup using 2 GXP75 and never notice a difference in speed. Just sold it and got a bigger HD. All you're getting is bragging rights.

 

[Brig]

BX133-Raid, HPT370, (2)75gxp's in array

I love it. I also disagree with the statement about "doesn't have true gains performance for average users". I posted just one of my 'real' gains in a reply in another thread.

numbers

Definitely agree that a backup is a necessity 'just in case', but that's always true.

 

[Argo]

I do notice speed increases, however, they aren't that big. The whole RAID 0 setup cost only extra $70, so it's well worth it.

 

[Windogg]

Brig:

Here's some real world numbers for you. I timed these on a watch, not using benchmark software.
Time from pressing the power on switch, until computer was ready at desktop screen with W98:

C300A@374, BH6 (1) WD 5400 - 120 secs
P3700@933, BX133Raid (2) 75gxp's ATA100, Raid1(mirror) - 32 secs
P3700@933, BX133Raid (2) 75gxp's ATA100, Raid0(stripe) - 15 secs

You are comparing a Celeron @ 374 w. 5400RPM HDD to a P3 @ 933mhz w/ 7200RPM HDD. Of course the P3 will boot faster. The massive FSB difference is also a factor. Not only that but other system factors such as drives, network protocol, "cleaness" of install will impact performance. I've played around with RAID0 and have seen no noticalble performance gains.

 

[Brig]

I know what you're saying, but the key is not the C300A vs a P3. I just threw that in for anyone trying to make a decision between 5400 rpm and 7200 rpm. The real comparison for this thread is between using a single drive (basically raid1) and a two drive raid0 setup, at the same speed.

 

[Windogg]

Again that is deceptive since RAID0 and RAID1 both read the same way. Data is pulled from both drives. The numbers should be very close to each other on read performance. Bootup is not a good indicator of disk performance since other factors are at play. Transfer of files of various sizes would give a more accurate picture of what is at work. IMO, your methods are flawed.

Windogg

 

[Noriaki]

Keep in mind that all IDE RAID is software based, at least right now. Apperntly prommise is working on it though, but there is no IDE Hardware RAID yet.

What this means is that yes, your hard drive perfromance will increase greatly in RAID0, but it uses CPU time to make the RAID.

You have to decide what's more important to what you do.
If you do stuff that's really disk limited and you have an overpowered CPU the RAID0 may really help you out.

If you do stuff that's CPU limited RAID0 is going to hurt your performance.

You have to determine where your apps balance.

 

[xtreme2k]

i would love to see some systematic benchmark with raid vs single drive

i really hate it when all these tech website just basically 'copy' each others 'topic' and never actaully think of something new to benchmark

 

[Brig]

Windogg,

You know, if I actually HAD any methods, they might be flawed. But since, as I mentioned before, that my lab consisted of an on/off switch and a wristwatch, I think the term 'flawed' might be a bit too strong.

I also forgot to say earlier that I know these numbers won't remain as good on bootup when I have to install an Ethernet connecter for DSL; I'm sure they'll go back to approaching the 5400 rpm drive times. They are, however, indicative of the overall improvement I see by using raid0.

 

[AdamK47]

Performance on reads from RAID-1 and RAID-0 are not identical. I don't know where you got that from. With RAID-1, whatever drive has the data ready is read from that drive first, therefore decreasing drive access time. With RAID-0, the data is read from the blocks on all drives in an array to increase bandwidth. Also, with the disk access time being the same as the slowest drive in the array, or equal to the access time of an identical drive. Data is not being read from both drives in the same way with RAID-1 as it is with RAID-0.

I can't respect someones opinion if they do not have the correct facts. I don't justify myself by stating what job title I have, or what useless qualifications I may have in the networking field pertaining to this topic, but what I do have are the facts.

 

[xtreme2k]

Raid 1 (mirroring) performs actually SLOWER than a Single Driver Configuration

since Each command has to be repeated twice (one to each hard disk)

 

[kranky]

Echoing stuff that's been said already, I guess.

RAID has a place on some systems, not all systems. I have no use for it at home (additional risk of losing data, no need for faster I/O), but I wouldn't be without RAID 1 and 5 on the servers at work. RAID 1 for the operating system disks and RAID 5 for the data disks is right for our situation.

In my situation, I won't give up CPU cycles for RAID at home. Others feel differently. It's not one of those situations where there's an obvious right answer.

 

[Brig]

I think that RAID for what the average Anandtechie uses his computer for is overkill

This is just a hobby for me, so I couldn't agree more with that statement. Unless it really matters to a gamer that their 'save game', or 'load game', takes 6 seconds instead of 10 (for instance), why risk raid0 ? Although there is a huge performance difference between raid0 and raid1 (found my notebook, so here are the Sandra score comparisons on my setup), it's strictly an individual preference as to whether or not it's worth it.

Sandra: Raid1-21676, Raid0-41972

Even though those benchmarks are obvious, I got raid for the auto backup capability for my 15gig drives. It'll stay that way until/unless I find a 'can't say no' deal on a 30gig drive. Then raid0 will be 'gravy'.

 

[Windogg]

AdamK47 - 3DS:
I can do a more deatailed search but I just wanted something quick:

RAID Explaination

An excerpt:

RAID 1 creates a duplicate of all the contents of one drive on another. This protects the data, since the failure of one drive means that the data is still available live on the other. Read access is also faster because data can be retrieved from either drive; writing data, however, is slower because you need to write the data twice.

Yes writes are slower because you have to swite to both disks and make sure the data is correct but you can read from bothh drives like RAID0. The mirrored drive does not just sit there uselessly.

I may not have any fancy papers saying I'm certified at stuff but I am pretty good at what I do.

Windogg

 

[Windogg]

AdamK47 - 3DS

I can keep them coming if you wish?

Another Piece on RAID1:
RAID Level 1 provides redundancy by writing all data to two or more drives. The performance of a level 1 array tends to be faster on reads and slower on writes compared to a single drive, but if either drive fails, no data is lost. This is a good entry-level redundant system, since only two drives are required; however, since one drive is used to store a duplicate of the data, the cost per megabyte is high. This level is commonly referred to as mirroring.

Another Site on RAID

 

[Windogg]

AdamK47 - 3DS

Please read before you comment again. I don't wish to misrepresent my qualifications.

Under "Advantages":
Twice the Read transaction rate of single disks, same Write transaction rate as single disks

Please Read A Complete Explaination On RAID1

Also reread my post. I never "Identical" I said "Close" Big difference.

 

[Windogg]

AdamK47 - 3DS

Another One For Good Measure

RAID level 1 refers to maintaining duplicate sets of all data on separate disk drives. Of the RAID levels, level 1 provides the highest data availability since two complete copies of all information are maintained. In addition. read performance may be enhanced if the array controller allows simultaneous reads from both members of a mirrored pair. During writes, there will be a minor performance penalty when compared to writing to a single disk. Higher availability will be achieved if both disks in a mirror pair are on separate I/O busses, known as duplexing.

My qualifications may be "useless" to you but what benefit do we get from having your specs in yout sig?

 

[AdamK47]

This is what you wrote:


<< Again that is deceptive since RAID0 and RAID1 both read the same way. >>



They do not read the same way! The transfer rate is slower in a RAID-1 array than a RAID-0 array. What don't you get? Even one of those links you gave me said this:

Transfer rate per block is equal to single disks

Now I'm going to start a new reply to exaggerate my response. (expect more bolding too!)

 

[Hardware]

Raid 0 Fast
Raid 1 safe
raid 5 blows (how do you spell 3-5 mb/s)?
Best raid 0+1

 

[AdamK47]

Heres another quote from the same link you gave me:

In addition. read performance may be enhanced if the array controller allows simultaneous reads from both members of a mirrored pair.

We're talking about IDE drives. If I remember correctly the IBM 75GXP is an IDE drive. The IDE RAID controllers do not offer simultaneous reads with RAID-1. Heres a link to one such controller:

FastTrack6 Review

Heres a link to the Promise site explaining how they handle reads on the FastTrack in a RAID-1 array:

IDE RAID on Promise.com

And heres a bit from that page:
Under mirroring, FastTrak uses a concept called Load Balancing to improve data handling -- particularly in multi-tasking operating systems. Load balancing sends multiple data read requests to the mirrored drives. As one drive is searching for the first data read request, the idle mirrored drive seeks a second data request (see Fig. 6).

As you can see by that, it's very obvious that the reads are not done in the same way as they are in an IDE RAID-0 array.

 

[AdamK47]

And to make an even more exaggerated reply heres another one...

This shows that the FastTrack is not the only one that handles RAID-1 reads the same way. The HighPoint 370 does the same thing:

FAQ for HPT370

Q: What does the RAID 1 &quot;Mirroring&quot; feature do?
A: RAID 1 (Mirroring) configures an array of 2 drives to write duplicate data and read in parallel. The HPT370 performs reads using advanced data handling techniques that distribute the workload in a more efficient manner than using a single drive. When a read request is made, the HPT 370 selects the drive positioned closest to the requested data, then looks to the idle drive to perform the next read access.

 

[Madcowz]

I love how a simple raid question turns into a HUGE blown out arguement between two or more people... That seems to happen all the time and that's why I love these forums so much! Thanks for the chuckles guys!