RAID-0 for capturing??

[DiViNeLeFT]

i do a massive amount of different tasks on my comp... from ripping dvds/cds, encoding video, capturing video... would a RAID-0 array really benifit me for capturing without hindering performance on other tasks. i realize the trade off of higher HDD performance for more cpu usage.

[bigmicka]

if you had a raid 0 array (stripe set without parity) and it was on a seperate controller than you O/S drive, you would get better performance from your disks for capturing.

with a stripe set without parity, it will split the data use over both drives. with one block written to one disk and then one block written to the next disk. This alternates between the drives in your array. it is supposed to give you an increase of roughly 10% on your acces times to this drive. as it has extra time to move the head to get the next block of data, while the other drive is sending through its data. we are talking milliseconds here but it is a difference.

mic

[sommersby]

most of the latest reviews including Maximum PC witch most here highly regard say the speed increase does not justify the expence or trouble of raid, if price is no object then a 4 drive raid will give you about as fast a drive system as you can hope for but still only a 30 percent increase over a single udma drive, and if you lose one drive you lose all your data, if you want safty and security then a two drive raid 1 system is great because you duplicate your data but at a cost of speed and disk size, and with serial ata due out anytime, which will be equal to raid or scai on debute and promises to to double or triple within a couple of months of it's debute the recomended set up is either the IBM deskstar or the new Western digital with the 8mb buffer, either one of these will serve you well untill the new serial ata arrives.

[mope]

There is a major difference in speed. I run a four IBM 7200 Deskstar (RAID 0+1) configuration and it rocks. You don't lose data if one drive fails (that's the point of mirror and stripe). If you only have two drives then you can still do a mirror array and obtain a slight increase in speed. The major benefit of that, however, is that if one drive fails you can switch to the other one. Keep in mind that in order to do a mirror the drives have to either be exactly the same or the second one has to be bigger. The capacity won't do...it can't be off by a single bit. Most RAID drives are purchased at the same time since you lose whatever capacity you have to purchase in excess of your first drive. If you have any data you value don't even think about running a pure stripe array.

[sommersby]

Well, I don't have a raid system but I do trust what I read and according to Maximum PC you NEVER gain speed with a mirror system, how could when you are writing the same data twice! This is what Maximum PC says , and I quote, " you will see about a 10 to 15% in creae in speed in a striped array and a 10 to 15% decrease in speed in a mirror array. with a 4 drive striped you will see about a 25 to 30% increase in speed, however, this is an increase in write speed, not read speed, in most cases you will see a decrease in read speed in any raid arrangment over a single drive system" I have always trusted maximum PC as they tell it like it is and at the start of the article they even stated that their are some BIG claims out there for raid systems but they have never been able to duplicate which they more or less blame on hype and wishfull thinking. Get the mag and read it for yourself, I would before I would take anyones advice seriously, it is a pretty big investment, one I thought about untill I read more about it and got past the hype.

[phelix]

This is what Maximum PC says , and I quote, " you will see about a 10 to 15% in creae in speed in a striped array and a 10 to 15% decrease in speed in a mirror array. with a 4 drive striped you will see about a 25 to 30% increase in speed, however, this is an increase in write speed, not read speed, in most cases you will see a decrease in read speed in any raid arrangment over a single drive system"

Well, that is the biggest pile of crap I've ever read (what you quoted, I have not read the entire article). Here's some real truth:

1) ALL Raid on the Motherboard solutions are crap

2) Many cheap raid cards are also crap. Here's a hint: if you raid controller allows you to hang two drives off a single port, it's crap.

3) RAID-0 imroves one thing, Sustained transfer rate for both read and writes. Depending on your setup, you may pay a 1 or 2 ms penalty in latency. Improved STR does not have a large effect on everyday things like word processing, email, etc. (the 10% - 15% Maximum PC talks about), but when working with big files, it can yield significant imrpovement.

4) RAID-1 (mirroring) on a good raid card can improve reads without hurting writes compared to a single drive. It does this by balancing load among the different drives. It's not a spectacular improvement, but at least you're not worse than a single drive.

5) Good Raid cards can deliver good performance. My 3 drive RAID-0 array (using 3ware escalade 6410 controller) delivers 82MB/s reads and 50MB/s writes. The writes should be faster, but there's some inefficiencies in my setup. As a comparison, the drives I'm using rate 38MB/s reads in a single drive config.

6) You don't need RAID for video capturing. Using huffyUV, I've never seen a capture exceed 10MB/s, which any modern 7200rpm drive can easily sustain.

7) Some benchmark programs don't work with raid arrays. Avoid Sandra and HdTach. Winbench is OK. Iometer and ATTO are good.


So, unless you're willing to spend the money for a good raid card, don't waste your time. If you're really interested in this, I recommend looking at http://www.storagereview.com

[DiViNeLeFT]

7) Some benchmark programs don't work with raid arrays. Avoid Sandra and HdTach. Winbench is OK. Iometer and ATTO are good.

are iometer and ATTO free and if so where can i download them???

i did go ahead and set up the raid-0... it is using onboard highpoint raid. im not looking for massive proformance gains... i just had to identical drives settin around and figured ive try it. but now i wanna benchmark it a little.

[Spicuzza]

I have a RAID-0, it makes a BIG difference when rendering and encoding, by BIG difference I mean from minutes to seconds. Let me clarify I don't do long projects, just short fun stuff like personlized music videos for friends and relatives from their photos and my son's High School wrestling matches. I have 4 HDD's in my system. I use the RAID-0 for all video capturing/editing/encoding, the RAID-0 array is 2 WD 20 gig 7200 drives ATA 100, 40 gig total capacity, the primary master is a WD 40 gig 7200 ATA 100 that I use to store finished video projects, attached to it is a WD 20 gig 7200 ATA 100 that I use to store music and backup files. One word of caution a RAID-0 is for performance not redundancy, if one drive fails you will lose ALL infomation. There is no recovery from this.
Gary Spicuzza

[DiViNeLeFT]

I have a RAID-0, it makes a BIG difference when rendering and encoding, by BIG difference I mean from minutes to seconds. Let me clarify I don't do long projects, just short fun stuff like personlized music videos for friends and relatives from their photos and my son's High School wrestling matches. I have 4 HDD's in my system. I use the RAID-0 for all video capturing/editing/encoding, the RAID-0 array is 2 WD 20 gig 7200 drives ATA 100, 40 gig total capacity, the primary master is a WD 40 gig 7200 ATA 100 that I use to store finished video projects, attached to it is a WD 20 gig 7200 ATA 100 that I use to store music and backup files. One word of caution a RAID-0 is for performance not redundancy, if one drive fails you will lose ALL infomation. There is no recovery from this.
Gary Spicuzza

your setup is exactly what im trying to acheive...
i just finished my 2 40gb western digital RAID array ata 100, 7200rpm tests... it benchmarked pretty good...
i have an attached 30 gb ata100 7200rpm music and video HDD...
and i have a ata66 rpm 5400 removable HDD...

[phelix]

7) Some benchmark programs don't work with raid arrays. Avoid Sandra and HdTach. Winbench is OK. Iometer and ATTO are good.

are iometer and ATTO free and if so where can i download them???

Iometer:
http://developer.intel.com/design/servers/devtools/iometer/
http://sourceforge.net/projects/iometer/

ATTO:
http://www.attotech.com/software/index.html
select ATTO ExpressPro-Tools and ATTO BridgeTools, fill out the form, then ATTO ExpressPro-Tools and download PC SCSI Version.

[DiViNeLeFT]

well, with that atto program ive gone from regular 30 gig ata100 7200rpm drive at 35953 write 31300 read to a highpoint duel ata100 7200rpm 40gb to come to 80gb RAID write 60787 and read 56871...

so raid... even onboard shines...

[SiCN]

I have a ABIT KG7-RAID motherboard with on-board HighPoint 370A RAID controller. To that I have four IBM Deskstar 7200RPM disks connected (RAID0+1). For the first: Yes, RAID0 does make harddisk intense operations faster (that includes capturing video to the harddisk). But therefore the overhead that RAID1 generates makes the RAID0 configuration a little slower again. But still faster than a single drive. Still, I figured that I am better off mirroring my stuff on a second pack of disks because it is still a pain in the ass to lose all data in case one disk fails.

Do not forget: Also the Block Size with which your RAID controller works is of importance. If you say that anything should be split in 4KB blocks then it is unlikely that you have a big speed gain since the controller will have lots to do just to get all the 4kb pieces together. Some website once said that 64KB is ideal. I am not quite sure if I agree because I have had 32KB before I read that article and switched to 64 afterwards, but it felt just as fast.

Hehe, so now go for it with 4x Western Digital 120GB w/ 8MB Cache Who wants to donor some of those disks?

[phelix]

well, with that atto program ive gone from regular 30 gig ata100 7200rpm drive at 35953 write 31300 read to a highpoint duel ata100 7200rpm 40gb to come to 80gb RAID write 60787 and read 56871...

so raid... even onboard shines...

Have you played with the ATTO settings. In particular, make sure the "Total Size" is 32MB and Direct I/O is checked. If you leave the total size at the default 4MB, you could end up just hitting the drives' buffers (2 drives at 2MB buffer per drive = 4MB). Having said that, your numbers sound right for a two drive raid0 array.


I think In my next system, I'll use the latest 3ware escalade controller, the 7450. It boasts RAID5 with 45MB/s sequential writes in a 4 drive config. 4 WD 120GB drives will give me 360GB of RAID 5 video editing goodness. Now, that's somthing those Motherboard RAID controllers could never achieve.

[Spicuzza]

[quote="phelix"]

..........6) You don't need RAID for video capturing...........

I totaly agree with that statement. My old 550 mhz Athlon with one WD 7200, 20 gig, ATA 66 captured video just fine. The big improvment comes when you edit/render/encode your final project. If fact until I set up the RAID drives on my current system there wasn't that big of a difference between the 550 mhz Athlon and the new 1.33 ghz Athlon. Like I said before for the type of projects that I do the render/encode time was cut from minutes to seconds.
Gary Spicuzza
cic7@juno.com

[sommersby]

How can a raid array possibly increase encoding speed, that is just a ridiculas statement, I would like to see the processor that could posibly encode fast enough to push even the slowest hard drive running in pio mode. lololololol

The article I was refering to was the jan issue of Maximum PC page 30.

they used a Promise FastTrak IDE RAID controller using Western Digital 1200bb driives in various RAID scenerios. Here are the results in a nutshell, Oh, and they did use HD Tach, I guess you could do the ATI thing and search around till you find a benchmark that fits your expectations.

HD TACH READS HD TACH WRITES
1 drive 40.3 MB/sec 23.9 MB/sec
2 drives striped 39.0 MB/sec 28.7 MB/sec
4 drives striped 27.2 MB/sec 32.0 MB/sec
2 drives mirrored 40.1 MB/sec 19.0 MB/sec
4 drives striped/mirrored 38.8 MB/sec 19.1 MB/sec


I would much rather trust what I read than exagerated or wishful thinking.

[Spicuzza]

For anyone considering setting up a RAID system I would recommend the following site.
http://www.raid.com/04_01_00.html
This specifically addresses the pros and cons of a RAID-0

Recommended Applications for RAID-0

Video Production and Editing
Image Editing
Pre-Press Applications
Any application requiring high bandwidth

Gary Spicuzza

[phelix]

How can a raid array possibly increase encoding speed, that is just a ridiculas statement, I would like to see the processor that could posibly encode fast enough to push even the slowest hard drive running in pio mode. lololololol

I agree. Encoding is and will probably always be CPU bound. However, I have noticed some improvment in muxing operations (which is probably more I/O bound).

The article I was refering to was the jan issue of Maximum PC page 30.

they used a Promise FastTrak IDE RAID controller using Western Digital 1200bb driives in various RAID scenerios. Here are the results in a nutshell, Oh, and they did use HD Tach, I guess you could do the ATI thing and search around till you find a benchmark that fits your expectations.

HD TACH READS HD TACH WRITES
1 drive 40.3 MB/sec 23.9 MB/sec
2 drives striped 39.0 MB/sec 28.7 MB/sec
4 drives striped 27.2 MB/sec 32.0 MB/sec
2 drives mirrored 40.1 MB/sec 19.0 MB/sec
4 drives striped/mirrored 38.8 MB/sec 19.1 MB/sec


I would much rather trust what I read than exagerated or wishful thinking.

And I'd rather trust benchmarks from people who know what they're doing. There are tons of benchmarks over at http://www.storagereview.com done by people who really know their stuff. If you think about it, it should be obvious that there is something seriously wrong with the above numbers. A 2 drive RAID1 (mirrored) has better read performance than 4 drive RAID0 (striped)? I don't think so.

Did Maximum PC at least run the Winbech disk tests? Or did they base their entire article on only one bechmark.

Theres a big difference between shopping for benchmarks and recognizing that one particular benchmark is flawed.

BTW, I did a quick google usenet search and found several reports of people seeing 60+MB/s with HD Tach.

[mope]

HD TACH READS HD TACH WRITES
1 drive 40.3 MB/sec 23.9 MB/sec
2 drives striped 39.0 MB/sec 28.7 MB/sec
4 drives striped 27.2 MB/sec 32.0 MB/sec
2 drives mirrored 40.1 MB/sec 19.0 MB/sec
4 drives striped/mirrored 38.8 MB/sec 19.1 MB/sec


I would much rather trust what I read than exagerated or wishful thinking.

OK. I don't know why I'm about to waste my time...but here goes. I checked your source since you seem to want to rely on it so heavily. I had a hard time finding your article since EVERY time RAID came up in the search hits Maximum PC raved about RAID.

Here are some URLs (maybe you could post yours so I could review their test for myself).

"We do know that dual 75GB IBM hard drives offer awesome real-world performance when striped together in a RAID 0 config." - http://www.maximumpc.com/reviews/systems/review_2001-04-01.html

Article titled "Just Shy of Kick Ass"
"dual ATA/66 hard drives in a RAID 0 config were benchmark breakers...The double-barrel data shotgun is comprised of dual 13.6GB IBM Deskstar ATA/66 hard drives in a RAID 0 configuration that improves performance by disk striping, which involves breaking files into chunks and writing across both disks simultaneously...We loved the Athlon, GeForce, and RAID array" - http://www.maximumpc.com/reviews/systems/review_2000-04-02.html

"While the Polywell reaps the benefits of its ATA/100 RAID array (see this months feature on IDE RAID)," - http://www.maximumpc.com/reviews/systems/review_2000-10-01.html

Article titled "The Definition of Kick Ass"
"But chew on these for starters: Its got an ATA/100 RAID array with two 75GB IBM 75GXP hard drives (for math dummies, thats 150GB of storage)," - http://www.maximumpc.com/reviews/systems/review_2001-01-01.html

You might want to recheck what you read and how you interpreted the data.