8 drives in 1 system - what's the cheapest way?

[Undefined]

Hey all.

I buy & sell hard drives on eBay and I'm trying to consolidate 3/4 testing systems into one.
The problem I'm having is that I need to be able to benchmark ideally around 8 SATA II drives at once without a drop / much of a drop in speed compared to if I was testing one drive connected to the controller native to the board. The root I think would be the best is if I was to buy an 8 port raid controller and just use it in JBOD mode, but everything that I'm finding that’s like this is extremely expensive - £250/300 which I don't really want to spend if at all possible.
I'm trying to reuse equipment that I currently own which means that the system will be an old 775 based setup; the fastest slot available to me is a PCIE 2.0 8X, which I *think* should be enough bandwidth to be able to test 8 drives at once without slowing down so much that the benefits of having one system are negated.

I was wondering if I could get some thoughts about the best way forward with this build? Does anyone know of a raid card that will do what I want and that won't cost that much used? Technically it doesn't have to have 8 ports - I have 2 spare using ICH7, so 6 would be fine.

Any pointers at all would be great.

Cheers.

 

[ddrueding]

I'm fairly certain that the performance numbers will be significantly affected by having more than one benchmark going at once. In the spirit of most online forums, allow me to not answer your question but offer a different solution.

Get an eSATA hotswap toaster. At least you won't have to open up the chassis or even reboot between tests.

 

[Undefined]

Hi and thanks for the reply.

I don't mind a bit of a speed reduction - I think any decreases in speed would be because of the connection I'm using? When I test I actually look at how long individual sectors take to respond as opposed to an MBPS rating, although obviously you can calculate an MBPS rating from this information. In other words, if each sector takes a few MS longer to respond I don't mind - this sort of thing vairies per drive anyway, just so long as its not dreadfully slow to the point where I start to wrongly identify a healthy drive as bad. I'm quite bad at explaining this sort of stuff encase you couldn't tell.

I've tried those hotswap things before; 7 can actually hotswap internal SATA drives just fine in most cases, you just have to be a bit carefull when you plug power in and remember to use something like hdparm to power the drives down before you remove them. They are pretty cool devices, but I'd have to buy at least 6 of them which then brings me back to the problem of connecting them all.

Thanks anyway.

 

[Undefined]

Not sure what happened with ^; I don't seem able to remove it?

Hi and thanks for the reply.

I don't mind a bit of a speed reduction - I think any decreases in speed would be because of the connection I'm using? When I test I actually look at how long individual sectors take to respond as opposed to an MBPS rating, although obviously you can calculate an MBPS rating from this information. In other words, if each sector takes a few MS longer to respond I don't mind - this sort of thing vairies per drive anyway, just so long as its not dreadfully slow to the point where I start to wrongly identify a healthy drive as bad. I'm quite bad at explaining this sort of stuff encase you couldn't tell.

I've tried those hotswap things before; 7 can actually hotswap internal SATA drives just fine in most cases, you just have to be a bit carefull when you plug power in and remember to use something like hdparm to power the drives down before you remove them. They are pretty cool devices, but I'd have to buy at least 6 of them which then brings me back to the problem of connecting them all.

Thanks anyway.

 

[Adcadet]

Why do you need to benchmark all of these drives? And why do you need to bench multiple ones concurrently?

 

[LunarMist]

Why do you need to benchmark all of these drives? And why do you need to bench multiple ones concurrently?

I'm not sure that benchmark is the right term. Presumably the OP wants to ensure used drives are in good shape before flogging them on auction and there are many to be tested.

 

[Bozo]

Buy an 8 port RAID card and set it up as JBOD. Run multiple instances of the test program.

 

[Undefined]

Hi and thanks for the replies.

LunarMist got it in one and Bozo seems to be thinking of the same solution that I was thinking of.
Does anyone know something that I can pick up used for not that much? As soon as you start looking at 6/8 ports the prices skyrocket.
I know the costs of stuff like this are justifiable for someone who wants to build an array for their server, but £250 or so is a bit more than I was thinking of spending although I will be able to make that back from sales.

Cheers.

 

[LunarMist]

Can you use 2x4 port or even 1x2 and 1x4 port controllers?

 

[Handruin]

Hi and thanks for the replies.

LunarMist got it in one and Bozo seems to be thinking of the same solution that I was thinking of.
Does anyone know something that I can pick up used for not that much? As soon as you start looking at 6/8 ports the prices skyrocket.
I know the costs of stuff like this are justifiable for someone who wants to build an array for their server, but £250 or so is a bit more than I was thinking of spending although I will be able to make that back from sales.

Cheers.

How about a Dell Perc 5i/6i from eBay to fit your budget? I and a few others on this forum use them with good results in our personal systems. They fluctuate in price, but you might be able to find one and fit it in your budget and get two cables to attach 8 SATA drives to it. I'm running 8 SATA II drives in my NAS off of one Perc 6i. I paid $150/each on eBay for two of them with the batteries and then purchased inexpensive cables from a company in China.

 

[Mercutio]

The IBM M1015 supports modern (larger than 2TB) drives, unlike a Perc5/6. It also supports Expanders so you can plug in to one of those cheap 4-bay SiL things or an HP SAS expander. There's a firmware hack that essentially disables all the RAID features of the card. That is useful since most RAID controllers aren't going to be addressable by low level tools.

 

[Santilli]

http://www.youtube.com/watch?v=kztiseGvvis
These SATA bays have been the best I've tried.
I combine that with one of the toasters Mercutio suggests. I don't see why you couldn't buy two toasters or more toasters and run them through the USB ports.

As for the used raid card in SATA:

I haven't tried the Perc cards. I've used a couple Megaraid LSI cards, used, and 3 of 3 have failed.

Also 2 of the 3 had problems working with modern tech. Drive limits, the channels were limited to 60 MB/sec, etc.

If you want a bunch of drives it's probably cheaper to buy a modern motherboard with a bunch of SATA ports then a used raid card.

 

[Mercutio]

The Perc and IBM controllers are both some form of rebranded LSI controller. They're generally not used so much as pulled from systems and replaced with other controllers for one reason or other. "New system pull" is a very common condition to find them, anyway.

 

[Handruin]

I haven't tried the Perc cards. I've used a couple Megaraid LSI cards, used, and 3 of 3 have failed.

Also 2 of the 3 had problems working with modern tech. Drive limits, the channels were limited to 60 MB/sec, etc.

If you want a bunch of drives it's probably cheaper to buy a modern motherboard with a bunch of SATA ports then a used raid card.

The Perc and IBM controllers are both some form of rebranded LSI controller. They're generally not used so much as pulled from systems and replaced with other controllers for one reason or other. "New system pull" is a very common condition to find them, anyway.

This has been my observation also (system pulls), but on the flip side, tons of the Dell servers we have in the labs have a variety of Perc RAID controllers (which are of the LSI-rebrand like you mentioned between Perc 5i/6i/H700) and we do not see the failure rates that Santilli mentioned. Batteries have gone bad, but that's not uncommon after years of use. The controllers themselves seem to be rather reliable. They are finicky about the types of drives that connect to them, but otherwise perform very well with the appropriate drives attached.