Why fuse-sshfs performance is so poor?

Stephen John Smoogen smooge at gmail.com
Tue Apr 27 18:10:52 UTC 2010


On Tue, Apr 27, 2010 at 12:32 AM, Kirby Zhou <kirbyzhou at sohu-rd.com> wrote:
> fuse-sshfs performance is poor? It can only reach 10M/s over the
> gigabits-ethernet.
> On the both endpoint, ssh and sshd processes take the CPU usage less than
> 20% with a single core of Inteal-E5520.
> It seems the speed should be at least 40M/s.
>
> ]# dd if=olddata/diskdata count=100000 bs=64k skip=8000 | md5sum
> 316211200 bytes (316 MB) copied, 32.2354 seconds, 9.8 MB/s
>
> ]# dd if=olddata/diskdata count=100000 bs=4k skip=8000 | md5sum
> 28954624 bytes (29 MB) copied, 3.80448 seconds, 7.6 MB/s
>
> ]# dd if=olddata/diskdata count=100000 bs=512k skip=8000 | md5sum
> 30408704 bytes (30 MB) copied, 4.33369 seconds, 7.0 MB/s

It can depend on multiple things like encryption, network latency,
latency of the fuse system itself. You would need to test with say
arcfour as the encryption method or use a method that has no
encryption to see what the base speed would be.


>
> Regards,
>   Kirby Zhou
>   from   SOHU-RD   +86-10-6277-8261
>
>
> _______________________________________________
> epel-devel-list mailing list
> epel-devel-list at redhat.com
> https://www.redhat.com/mailman/listinfo/epel-devel-list
>



-- 
Stephen J Smoogen.
“The core skill of innovators is error recovery, not failure avoidance.”
Randy Nelson, President of Pixar University.
"We have a strategic plan. It's called doing things.""
— Herb Kelleher, founder Southwest Airlines




More information about the epel-devel mailing list