If SATA performance stays this way ...
___________________________
In fact, it is not sata performance, it is the ext4 driver's performance
Begin with some bogus benchmark:
ext4
[root@nas ~]# dd if=/dev/zero of=/test.bin bs=64K count=100
100+0 records in
100+0 records out
6553600 bytes (6.6 MB) copied, 0.125591 s, 52.2 MB/s
ext2 using ext4 driver
[root@nas ~]# dd if=/dev/zero of=/media/Opt/test.bin bs=64K count=100
100+0 records in
100+0 records out
6553600 bytes (6.6 MB) copied, 0.137121 s, 47.8 MB/s
ext2
[root@nas ~]# dd if=/dev/zero of=/media/Opt/test.bin bs=64K count=100
100+0 records in
100+0 records out
6553600 bytes (6.6 MB) copied, 0.070848 s, 92.5 MB/s
Notice the total write size I choose here, there are no real IOs, they are totally bound by cpu power.
Previously, I use ext4 driver to mount all ext2/3/4 partitions.
But seems there is no much can be done to improve ext4 write performance.
So I add ext2 driver back, we can choose ext4 for stability, ext2 for performance.
Anyone willing to test ext2 write performance is highly appreciated.
Prebuilt kernel image
https://skydrive.live.com/?cid=06815211 ... E6E5%21105
Known issue: you will get a random mac address.
By the way, ethernet driver's hardware offload is not ported yet, so network performance (around 50MiB/s) is bound by cpu too.