Raid Decessions...

[ripken204]

wow, have no clue why i said 300, nice catch, must have been looking at zxian's post before.

and do u have proof of larger hdds being poor in raid0? interesting point that ive never heard before. looking at benchmarks i dont seem to notice a difference.

i found this using google

http://os.vault9.net/forums/Hitachi_SATA_I...iew-t21127.html

[teqguy]

Large hard drives are poor performers, period.

Platters in general have a deficiency when it comes to reading and writing at their ends, because of the physical structure of the platters themselves(think in terms of a race track).

If you look at the HDTach benchmarks on the page you sent me, you'll see that halfway through, the performance drop is tremendously significant.

Single platter drives will also demonstrate this gradual decline at the end, but the decline will be less severe.

[Maleko]

Maybe larger capacity drives do loose performance, i dunno...but with the way apps are at the moment, and how they coudl be, especially in new PC games, you need about 3GB of space for each game!

Also people who do movie editing.....

People like that are going to go for larger capacity drives...so I can;t see the point to be honest.

Also, RAID0 should make up for the lose of performace in larger drives if that is the case, no?

[ripken204]

teqguy- thx for the info on that, and i have noticed how smaller hdds barely reduce in performance at all

[teqguy]

For me, putting two large hard drives that you know individually will not perform as well as a lower capacity, single platter hard drive in RAID0 is like beating a dead horse.

The most intelligent solution, which is also the solution that allows you to have your cake and eat it too, would be to get two 80GB single platter drives and RAID them, then get a large capacity hard drive for storage.

What I do when it comes to encoding is divide the file into 256MB chunks, then devote 384MB of ram to a ramdrive, which I then encode my file to. This improves encoding times tremendously, because the processor isn't waiting on the hard drive to write data.

If you have enough memory, the best solution is to put your source and destination files in memory, so the processor is reading from and writing to memory. Although, using the method I described above, you wouldn't have to have a lot of memory.... you would just need to divide the file into smaller portions.

[ripken204]

http://i4memory.com/showthread.php?t=1026

well now i found this site, still seems to be a drop towards the end

btw, what program do u use for your ramdisk, i have 2gigs of ram and havnt been able to find a good program

[teqguy]

btw, what program do u use for your ramdisk, i have 2gigs of ram and havnt been able to find a good program

Ramdisk from SuperSpeed Software is probably the easiest to work with, as mounting and configuring drives is all done within a GUI:

http://www.superspeed.com/desktop/ramdisk.php

[jcarle]

You know, there's nothing that frustrate me more then watching someone give improper advice to people. The unaccurate opinion of a single member can misguide hundreds of members in such a large forum as this one. It is not by talking with terminology that people are unfamiliar with that the opinion expressed contains correct information. teqguy, it is of my opinion that you are one of those people.

There are many points that need to be clarified, that I will do one at a time. And I will attack all points discussed in this thread, and if I've missed any, then please point them out to me and I'll clarify them.

So what affects a single drive's speed? There are a few factors to consider.

The most obvious factor, and the one SCSI drives have long been reknown for is the revolution speed. Older drives spin at 5,400rpm while most current drives spin at 7,200rpm. Only the ever so famous Western Digital Raptor series turn at 10,000rpm in the desktop (non-SCSI) world. Of course, the faster the revolutions, the faster the overall drive response, transfer speeds and seek times.

The amount of cache is also responsible in speeding up a hard drive when information is gathered repeatedly, such as the case in things like video games. The more cache, the better. Cheaper drives are only equiped with 2MB of cache, most drives today have 8MB of cache. More and more drives today are starting to come out with 16MB of cache.

The interface plays little role in the speed of the drive. Most drives today cannot reach even the transfer limits of PATA technology. The advantage that can come through the interface is with SATA's ability to use Command Queuing (TCQ/NCQ). Command queuing on a desktop setup does not give advantage to speed, but instead hinders it because of the linearity of command issued on a single drive from a single user.

The platter density has a direct affect on speed. Increasing the platter density has a two fold advantage. It reduces seek times because the head has less distance to travel to reach two given points and it increases through put because more data passes under the drive's head at the same revolutionary speed then a drive that has a lower drive density.

The number of platters also has a direct affect on speed. Increasing the amount of platters creates the same speed gain as increasing the platter density. Data is spread throughout all platters and for each platter there is a matching set of drive read/write heads. It's much faster to read/write from multiple heads on multiple smaller platters then to read/write on a single head on a larger platter. This is why a 320GB 7,200RPM 8MB hard drive is FASTER then an 80GB 7,200RPM 8MB drive of the same make, model and type.

Now what about RAID? First off, the differences in RAIDs.

RAID 0 is done by stripping data across drives. This provides no data reduncy so if one drive fails, everything is lost. However, the advantage is purely for speed. RAID 0 is one of the simplest RAID logistics, therefore requiring very simple RAID controllers and/or requiring very little overhead on the system to run. Drive reading and writing is improved substancially due to the distributed nature of the reading/writing. The overall bandwidth of the array can be seen as a total of all drives together, minus some minor overhead for the RAID controller. The total capacity of the array is equal to the total combined capacity of all drives in the array.

RAID 1 is done by mirroring the data from one drive to the other. This provides direct data reduncy so that if one drive is to fail, no data is lost. Mirroring is even simpler then stripping and creates a similar speed increase as RAID 0. The difference is that although the reading bandwidth is increases per total number of drives, writing is equal to the writing speed of a single drive, minus some minor overhead for the RAID controller. Also of significant importance is that the total capacity of the drives is equal to exactly half of the total combined capacity of all drives in the array.

So why RAID 0? If you want read speed and write speeds to both increase and do not intend on storing valuable data on that RAID array.

So why RAID 1? If you want to make sure you never lose your data from disk failure.

With the size of today's data, it's not conceivable for people to make backups on remote servers. You simply cannot backup GIGS worth of personal data on a remote server. People nowadays have music collections, movie collections, personal videos, personal pictures, documents, large e-mail folders, programs they've collected, source code, photoshop source files, and the list is endless. You cannot just Ghost, backup onto CD/DVD or store on a remote server that sheer amount of data. It's senseless. The only sensible solution is to RAID current a couple of large drives together and keep all valuables there. Hard drive also have a hugely higher safeguard life (when from a reliable company such as Western Digital, Hitachi or Seagate) then optical media.

Due to the chemical composition of optical media, optical media deteriorates within a matter of a few years. Optical media is also easy to damage with scratches, temperature changes and general simple abuse. Even the highest quality media will no resist the damage a child can do to it, and nor can it survive more then 5, 6 or even 7 years at the most. Hard drives, because of their metallic and magnetic nature, are much more reliable at keeping the data safe. Yes, it's true that a fire can destroy a fire, but a fire will also destroy the optical backups you may have in your house as well, remember that.

There's a reason why in my personal machine I have two 10,000rpm hard drives that are functioning in RAID 0 and in RAID 1 mode. I keep my operating system on a RAID 0 array for speed and I keep my personal data on the RAID 1 array for safe keeping. If there was no speed advantage to the RAID 0, I would have stuck with a single drive, and if there was a more cost alternative method for data safeguarding, I would have taken it instead of a RAID 1 array. The cost incurred for building a RAID system on two 10,000rpm hard drive is a VERY costly choice. But that cost brings gain and all my experience is not wasted when I made that choice.

[Maleko]

Thanks for that sweet post Jcarle!

so, what im going to look at now is getting 2 80GB SATA's, whiel keeping my exisitng 200GB for storage etc.

So, ill have to look at getting a PCI SATA controller to get more SATA ports

[jcarle]

I highly recommend the Promise SATA300 TX4 4-port SATA II PCI card. It does not support RAID, so it's much cheaper then it's RAID equivalent.

[nmX.Memnoch]

jcarle,

Thanks for posting almost exactly what I was getting ready to post.

Stating that larger hard drives are slower than smaller drives is entirely wrong. As was mentioned, today's drives have much higher data density in the platters. This results in less movement required for the heads to find data, making seeks faster. As was also mentioned, there is a direct correlation between the spindle speed as well as the amount of cache the drive has. SCSI manufacturers have known this for quite some time...16MB caches have been around for years on SCSI drives but are now making their way into PATA/SATA drives as more and more enthusiasts, as well are systems administrators, are looking for cheaper alternatives to high-end SCSI drives.

You can thank the new SAS (Serial Attached SCSI) for making this happen as most SAS external enclosures can be configured for either SCSI or SATA drives. SATA drives come in larger capacities at much cheaper prices. For example, the largest SCSI drive is 300GB and will cost you roughly $700-800US but the largest SATA drive at 500GB will cost roughly half that. That's a no brainer when you're looking to store terabytes worth of information, and considering that today's SATA drives are mostly derivitives of what's been learned in the SCSI world to make them so reliable.

[Maleko]

thnks guys!

got a final question regarding RAID, as im new to it so i gotta learn!

What storage do you get? ive heard different thing.

So, lets say i get x2 80GB and RAID0 them, do i get either:

80GB total Storage, or 160GB total Storage?

Ive heard that u get just the size of one drive, but ive also heard this:

Array Capacity: (Size of Smallest Drive * Number of Drives).

which to me says, 80 x2 = 160?

I know with RAID one i wud get just 80, as the drive is mirrored.

just cant figure out RAID0 total storage.

[jcarle]

You can in fact mismatch drives in a RAID array. However, in the cases of RAID 0 and RAID 1, the smallest drive within the array will be used as the baseline for the array size.

RAID 0: Stripping

[Total Capacity] = [smallest Drive Capacity] * [Number of Drives]

RAID 1: Mirroring

[Total Capacity] = [smallest Drive Capacity per Pair of Drives]

Example:

80GB + 200GB in RAID 0 = 160GB

80GB + 200GB in RAID 1 = 80GB

200GB + 200GB in RAID 0 = 400GB

200GB + 200GB in RAID 1 = 200GB

40GB + 60GB + 80GB + 100GB in RAID 0 = 160GB (First two drives paired, then next two drives paired)

40GB + 60GB + 80GB + 100GB in RAID 1 = 120GB (First two drives paired, then next two drives paired)

[nmX.Memnoch]

It should be noted that while it definitely does work, mismatching drives isn't recommended though. It's better to have drives that are paired up.

Also, don't forget about JBOD...which is literally Just a Bunch Of Disks. Most people don't recommend this either as there's no real advantage to it other than taking a "bunch of disks" and making them look like one disk. With this type of configuration mismatching drives isn't a big deal at all. What happens is that the controller will simply use one drive until it fills it up, then moves to the next and fills it up, then the next and so on. So if you had a bunch of different sized disks the total drive space would be whatever the total of the drives is:

40GB + 80GB + 120GB = 240GB JBOD "array"

Again though, JBOD offers no performance or redundancy increases at all. There's a slight possibilty of read performance gains, but only if multiple files are being read from different disks. However, due to the nature of the way they're written it's very unlikely this would ever happen.

[ripken204]

well jcarle told you a bunch of stuff on raid that should have helped

i doubt that you have enough money for two large hdds in raid and it seems like you want 2x80 in raid0, that will give you 160gigs total

if so then i would go with these:

http://www.newegg.com/Product/Product.asp?...N82E16822145082

they support sataII so that if you get a new mobo that supports sataII then you'll be all set