Performance Raid Setup - 187MB/sec sustained

[Jeff Kilgroe]

Could it be a power/heat issue?

That's really the only reason I can think of, but it still doesn't seem to add up.

3.5" drives do consume more power and put out more of that consumed power in the form of heat. But in general, they offer superior performance and capacity, meaning you can use fewer drives to accomplish the same goal and usually at lower cost.

[number6]

That's really the only reason I can think of, but it still doesn't seem to add up.

3.5" drives do consume more power and put out more of that consumed power in the form of heat. But in general, they offer superior performance and capacity, meaning you can use fewer drives to accomplish the same goal and usually at lower cost.

Could it be platter size? and therefore, arm travel? Seems like I read something to that effect on Tom's Hardware Guide maybe six months ago where they compared 5400 RPM 2.5 inch notebook drives in a RAID ? against a 3.5 inch HDD RAID ? I think their conclusion was that the 2.5 was just as good or better, but that's not my "final answer". On second thought, was probably more than six months previous since they were using 5400 RPM notebook drives instead of 7200's, which would have been available and seemingly, the more logical choice.

[Chris Kenny]

If money is no object, 2.5" drives are going to deliver better performance per unit of volume, and probably per watt of power consumed. Sure, they're 1/2 of the speed of 3.5" drives, but they're a less than a third of the size. And a smaller platter gives you faster seek times. (Though seek time isn't that important to video, which is most sequential reads and writes.)

Seagate actually makes a 2.5" 10K RPM drive targeted at the server market.

[Simon Blackledge]

yep sorry it's 40 drives.. 4 sleds of 10 drives, with quad channel fibre.

http://www.ciprico.com/Products/pdf/MV4440Data_Use.pdf

800MB/s min though.


How would monitor refresh effect raid speeds? :-/ ekk..

[Simon Blackledge]

This looks intresting... a switch for the caldigit raids, but looks like each box only feeds 1x not 2 so I take it 2 is the minimum. But if you start with 1 raid feeding a box direct with 2 feeds.. and then get the switch and add another box all the media on box 1 will be half data rate :-/

[Gavin Greenwalt]

How would monitor refresh effect raid speeds? :-/ ekk..

Do video card refresh speeds affect the timing at the southbridge? You might have to get your video card tweaked in such a way that you don't congest your PCI-X buses.

[Rocket]

We will be building a real-time 4K post pipeline from the ground up here in Johannesburg, and we have chosen a Bright Systems 4K BrightDrive setup as the storage sub-system for Scratch. This will give us sustainable real time uncompressed 4K bandwidth at around 1.2GB/sec for what I believe to be a reasonable cost.

My gut feel is to leave storage solutions at this kind of performance level to the experts. Even if all the individual components theoretically deliver the required performance, for commercial use reliability is crucial and any money you might save will be lost the first time you have to try and explain to a client what happened to their data.

Redundancy is another story altogether, in the event that one drive in the array packs up, you are going to be in trouble depending on the RAID configuration. We are even considering a second mirrored drive array that we can bring online at any point if necessary.

Rich