The case has an eSata port and cable.
I've run a speed test (Xbench) on a 2TB internal drive and the external RAID.
Here is the result. It seems the speed is good.. is it correct?
System Info Xbench Version 1.3 System Version 10.7.3 (11D50d) Physical RAM 16384 MB Model MacPro4,1 Drive Type WDC WD2001FASS-00W2B0 Disk Test 98.66 Sequential 175.94 Uncached Write 212.82 130.67 MB/sec [4K blocks] Uncached Write 181.19 102.52 MB/sec [256K blocks] Uncached Read 123.16 36.04 MB/sec [4K blocks] Uncached Read 227.40 114.29 MB/sec [256K blocks] Random 68.55 Uncached Write 21.96 2.33 MB/sec [4K blocks] Uncached Write 441.42 141.31 MB/sec [256K blocks] Uncached Read 155.38 1.10 MB/sec [4K blocks] Uncached Read 242.86 45.06 MB/sec [256K blocks] System Info Xbench Version 1.3 System Version 10.7.3 (11D50d) Physical RAM 16384 MB Model MacPro4,1 Drive Type Netstor H/W RAID10 Disk Test 97.50 Sequential 150.65 Uncached Write 182.92 112.31 MB/sec [4K blocks] Uncached Write 144.62 81.83 MB/sec [256K blocks] Uncached Read 104.63 30.62 MB/sec [4K blocks] Uncached Read 216.80 108.96 MB/sec [256K blocks] Random 72.07 Uncached Write 22.29 2.36 MB/sec [4K blocks] Uncached Write 205.88 65.91 MB/sec [256K blocks] Uncached Read 360.62 2.56 MB/sec [4K blocks] Uncached Read 332.37 61.67 MB/sec [256K blocks]
I believe that Nestor enclosure has a JMicron RAID chipset, in testing that chipset as implemented into a RAID chassis by another manufacturer we were getting near 200 MB/sec (sequential large file transfer tests) in a RAID5 over eSATA using a SATA card with a Silicon Image 3124 chipset. Drive model, as long as they are decent, like Seagate Barracuda or those WD Blacks, should not be a factor as the combined speed of that many drives is a lot faster than the RAID chipset is capable of. Ran same large file tests with same performance maximums - (200 MB/sec) running a simple RAID0 and a couple SSDs as well - to find the maximum throughput of the RAID chipset. Was faster over USB3 than eSATA.
The E2P SATA card uses a Silicon Image 3132 chipset. That chipset will top out at 150 MB/sec, or around that. So yeah, it will slow that RAID a little bit, 25%-30% anyway.
In a RAID10 you are effectively getting the striped speed of 2 drives. With those WD blacks, when the drives are empty, mechanical speed capability of 2 blacks would be up over 250 MB/sec. So the place where the speed limitation is when the drives are empty is first the 150 MB/sec E2P, then the 200 MB/sec JMicron RAID board.
With RAID10 you will show a significant slow down of the RAID as you fill up the space with data. The RAID10, because it uses fewer drives in a pair of stripes, than a RAID5 which uses basically all but one of the drives in a larger stripe, the RAID10 will show the effects of filling up the space much sooner than the RAID5. If your RAID is anywhere near full it will slow a lot!
I found the JMicon RAID quite functional. I would probably choose a RAID5 over a RAID10, but that is me. A faster SATA card will help a little, but part of the problem is the limitation of a RAID enclosure that is not really designed for speed. You are not going to make it a lot faster with a different card or a different RAID type.
Make sure you always maintain a backup of your data - RAID10 does not mean backed up. All it takes is a failure of the RAID software or hardware to lose everything. A separate copy is essential.
I always gave the E2P more like 130MB/s in theory with less in practice. But that was on PCIe 1.1 and older drives at the time also.
Sounds like a good setup all around, but would not deliver what 4-port card and enclosure with - but maybe not worth the cost to do so depending on the needs. Or a 4x/8x card which I assume means getting into SAS too.
One new WD Black empty was 135MB/sec (internal, SATA II) or the speed of one drive is all you really get in that setup.
You may want to look at SATA 3 boards like the Highpoint RR2721 which can provide SATA 3 RAID performance (both internal and external) at sane prices.
I have a RR2720 in my Mac Pro 3,1 and get single spindle (JBOD) speeds of 145MB/sec using the Black Magic Design speed test. I'm setting up for a 6-spindle RAID 5 and a pair of eSATA ports. The RR2720 requires a few tricks, but the RR2721 and RR2722 should be plug-and-play.
but it is HPT; and there are other makes and cards. There is even bootable PCIe populated with SSD chips - bootable.
Once I saw RAID6 and the limitations of RAID5, and given the size of today's drives people often use.
HPT also has 4-port SATA3/6G $249 on Apple Store and elsewhere. Said to be bootable.
Yes it is HPT, and I've have good experiences with them. (However, their tech support isn't for newbies...)
As far as I have been able to determine, a bootable 8-port SATA 3 board is close to $1000 and not internal. The (non-bootable) HPT I am using was about $170. The new HPT 2700 series have been getting good performance reviews. It's not the only solution out there but it is the cleanest and least expensive I've found.
The hatter wrote:
One word: bootable
Also, Areca and ATTO.
In my application, two words: Tiny and expensive.
PCIe Slot 2 carries the RR2720 (which does not need to be bootable) to support the internal RAID. The Mac Pro 3,1 only has two PCIe 2.0 slots, so the Accelsior would be pointless. I'm planning on using an SSD on the on-board SATA II controller as a boot drive.
BTW, do you know of a PCIe 2.0 expansion chassis that is sanely priced?
Total bandwidth is a function of slot speed, spindle count, and drive transfer rate. My RAID is not yet implemented because I only had Hitachi 3TB drives before the Thailand flooding (and ensuing shortages) drove prices completely out of sight. It will complete when prices normalize or I save enough money to buy 5 (maybe 6) Seagate 3TB drives (either way, cost is a significant issue.) I believe a RAID should use identical drives.
The Mac Pro 3,1 has two 16x PCIe slots and two 4x. My RR2720 is in slot 2 (16x) so it should perform as advertised. If you actually require speeds of 800MB/sec and beyond, don't take the free advice of anyone on this (or any other) forum; work with a professional and contract for a specific level of performance.
In the end you'll probably decide that reliability is your first criterion and all the speed you can get reliably is good enough. RAID 0 is like a racecar - sexy and exciting until it crashes. I'm looking for a RAID 5 that's fast enough (500-600MB/sec) to boost my overall throughput in disc-intensive work.