Re: RAIDframe write performance below expectations on a RAID-1 of two magnetic disks on NetBSD/amd64 9.1

[mlelstv%serpens.de@localhost (Michael van Elst)]

mp%petermann-it.de@localhost (Matthias Petermann) writes:

>       # gpt create raid1
>       # gpt add -l data -a 4k -t ffs raid1
>       # newfs -O 2 -b 16k -f 2k NAME=data

>This was formatted with an FFS filesystem (with the recommended
>parameters from [1]) and mounted with the mount option "log".

The best filesystem parameters depend on the size of your I/O
operations. [1] uses "unpacking source tarballs". But that
doesn't make such a large difference.


>However, the write throughput remains well below my expectations and I
>am despairing. When writing a 1 GB file, I achieve write rates of about
>2 MB/s.

How did you measure this?  



>[     3.400024] wd2(ahcisata0:2:0): using PIO mode 4, DMA mode 2,
>Ultra-DMA mode 6 (Ultra/133) (using DMA), WRITE DMA FUA EXT
>[     3.560024] wd3(ahcisata0:3:0): using PIO mode 4, DMA mode 2,
>Ultra-DMA mode 6 (Ultra/133) (using DMA), WRITE DMA FUA EXT

"using DMA" is the message here. So that's ok.  


N.B. in a similar setup (larger disks and layered RAID-10 on 4 drives)
I get:

# dd if=/dev/zero of=testfile bs=1024k count=10240 oflag=creat,direct
10737418240 bytes transferred in 29.037 secs (369784007 bytes/sec)

Each disk is good for about 230MByte/s, but with filesystem and
relatively small I/Os (MAXPHYS = 64kByte) you only get 180-190MByte/s
each. Striping + Mirroring over 4 disks results in 360-380MByte/s
net write speed which matches the measurement.


-- 
-- 
                                Michael van Elst
Internet: mlelstv%serpens.de@localhost
                                "A potential Snark may lurk in every tree."