Which Areca would suit me

J

Jeroen1000

Limp Gawd
Joined
Sep 17, 2010
Messages
266
So, the new 1880 range is really really tempty as I like the lastest and greatest. However, I've been wondering whether I truly need it.
I'm building around a Norco 4220 atm. As it stands now I'll be putting the beloved 2TB Desktar Hitachi's in (because of the lack of a proven alternative). I'll opt for RAID 6 and _probably_ only create 1 array. I know the 1680 series have problems when using multiple array's...

I'm looking to achieve read + write speeds around 250 MiB/second. I don't think I'll need more anytime soon.

For expansion, I'm going with a HP expander and, in the future, perhaps a second Norco with an addtional HP expander. So my card needs at least 1 internal and 1 external port.

There is a 1680LP with might be what I need. What would you recommend?
 
Also consider it's trivially easy to take internal ports and use them externally. Simple cabling and brackets. Yes, planning for this ahead of time is better, of course.
 
Partially a bump but I noticed something more.1680 uses a PCIe x8 interface. And a 1880 uses a PCIe 2.0 x8.

So technically, couldn't I get away with putting a 1680 in a PCIe 2.0 x4 slot?
 
As long as the slot has an open end and can accommodate longer cards, it will work just fine.
 
So bandwidth en technological wise, no restrictions? Only a physical obstruction may be there.
 
Technology, isn't it great. I never ever had to use a PCI express slot. Everything was always on the motherboard.

Thanks Blue Fox
 
I switched from Areca to percs because I seem to be getting better benches in my large array setups.

Unless someone had direct experience with this, I would caution about using an x4 slot for an x8 card, regardless of the tech revision. All my Areca cards (older ARC models) really need to be put in x16 slots (I don't have mobos with x8 slots) to function.
 
No damage should occur. It simply would not post. The chances I get a board with open ended slots are small anyway. And I refuse to dremel:D

Anyhow, I don't know what to buy 4 sure, yet...
 
Areca support warned me that the 1680 series (not sure about the 1880 series which is PCIe 2.0) can have a PCIe link speed problem on 5520 chipset motherboards (Dual 5500 / 5600 Xeons). Down to a bug in the 5520 which Intel aren't going to fix apparently. On system start up, the speed might be negotiated at x1 or x4 instead of x8. The card will still work though. You have to reboot until you get a x8 link negotiated.

This would suggest that you would be fine with a x1 or x4 slot, just you wouldn't get full speed. IIRC some of the shorter length slots are keyed so you can't physically fit a longer connector into them, also it's common to have components on the mobo that will obstruct a x8 card in a shorter slot.

On the other hand, I'm currently having a problem with a 1680ix-24 and a 5520 based motherboard, where the motherboard is faulty and not supplying enough juice to start the card properly. Not sure if you'd need a x8 or x16 connector for watts rather than throughput.
 
Jeroen1000 said:
I'm looking to achieve read + write speeds around 250 MiB/second. I don't think I'll need more anytime soon.
Click to expand...

Good thing that part should be easy to achieve. I mean its easy to achive with linux software raid so any current generation true hardware raid should not sweat on that assuming you use 5 or more disks with 500GB platters or better on raid6 or 4 or more on raid 5.
 
How do you calculate the throughput numbers?

Roughly speaking, on the older cards, you have 3 gigabit/seconds. This times 4 for the link from the controller to the expander. So 12 gigabit times 8 divided by 10 (encoding overhead) is 9,6 gigabit / seconds = 1200 megabyte / second.

Is that enough for 20 drives? Sequentially I'd say "no". Modern drives can get to 100 MiB/second so I'd need 2000 instead of 1200 MiB/second

But all of that is moot because I only have gigabit ethernet. Even when teaming 2 NIC's I'll never ever go over 250 MiB/second. Morever, the STP-protocol seems to cap out on 800 MiB/second. Don't know how high the bar is for 6 gigabit cards.

So am I making sense it is useless for me to buy a 6 gigabit card? Or will I regret it?
 
I would not worry much about trying to achieve 100% throughput efficiency unless you have 10G NICs.

Although with the expander not allowing full bandwith to the drives hung off of it you will lengthen the rebuild/resync times.

Modern drives can get to 100 MiB/second
Click to expand...

~100 to 140 MB/s on the outer tracks for modern SATA disks.
 
You must log in or register to reply here.
Back
Top