Raidsonic RAID1 box would probably be quite a tight bottleneck -- solutions like this tend to be very slow, even doing simple stuff like raid1. far from 40-60MByte/s, anyway.
I also would not recomment Via SN boards (ain't got any experience with the EN series, though) for this task. I've got an SN10000EG powering a small server (debian gnu/linux, unstable tree, custom-built 2.6.28 kernel, one 2.5" system drive and a couple of 1TB EcoGreens in an md raid1 array, using XFS, write-cache is disabled on the EcoGreens). There's a bunch of serious issues with this board:
- the AHCI implementation of the SATA-controller is seriously flawed. using NCQ in AHCI mode can lead to controller lockups. thus, no NCQ in AHCI mode.
- I wasn't able to use any of the cpu-offloading features of the gigabit controller. setting MTU to any value above 1500 makes the NIC stop transmitting packets after a short while (immediately with a kernel before 2.6.26, AFAIR). could be an issue with the via-velocity driver in the linux kernel -- I gave up trying after a while, so I'm not quite sure -- NFS or CIFS performance is not that important to me. NFS throughput while reading large files max out at ~25Mbyte/sec after some tweaking, CPU load at ~20-30% maximum.
- it definitely does cost too much to have such serious flaws
on the other hand, it has some nice features, like redirecting bios/bootloader console output to a serial port, small size, and a 16x PCIe slot, which should also accept some really nice Intel 1x PCIe Gbe card, solving the second issue. The AES/SHA-accelerator also works nicely, improving perfomance on encrypted volumes.
Devonavar wrote:
eSATA should be 100% transparent; it's basically an internal SATA connection with a different cable.
...and subtly different electrical specifications. should not be a problem with most hardware, though. sorry for nitpicking
