M4 read performance

Paaland

I was curious on the M4 read performance so I did a quick benchmark to try to find out if this read-striping works. I made a small tool that reads a 4 GB file using unbuffered read (bypassing Windows file cache).

Reading duplicated file from pool using read-striping:

Read 4771020800 bytes, elapsed time 00:00:40.4042590, transfer speed 115314 Kbyte/s

Reading duplicated file from pool without read-striping:

Read 4771020800 bytes, elapsed time 00:00:40.2170578, transfer speed 115851 Kbyte/s

Reading the same file directly from the disk:

Read 4771020800 bytes, elapsed time 00:00:43.6802800, transfer speed 106665 Kbyte/s

The difference is neglect able and not something you'd notice in practice. The good thing is that reading from the pool is just as fast as from the native drive. Alex: How optimized is M4 when it comes to performance at this time?

As for write speed I've not tested since I use duplication and write speed will equal the time to write on the slowest drive used of the two.

Alex

Read striping may not double your read speed. It works well in limited BUS scenarios like USB 2.0. Still testing / tweaking  this feature to see if it's worth it.

M4 read / write performance should be optimal.

Opening many small files is not optimized at this time.