El 11/08/2015 a las 06:37 a.m., Götz Reinicke - IT Koordinator escribió:
Hi,
i have two servers, connected to to the lan by 10Gb with 10Gb and DAS
hardware raid.
Each system con read and write locally or to the 10G iscsi by more than
200 MB/s.
Now I have to transfer backups form A to B and doing this with rsync
always stuck at +- 48-50MB/s no matter which options, compressions,
encryption etc I use. Even the plain default rsync is at that 50 Mb limit.
coyp by scp goes up to 200 MB/s.
Copy from and to my workstation with scp from or to both servers is at
1Gb limit (so +- 100 MB/s)
Why is rsync stuck at +- 50 MB/s ? Any suggestions hints ...
rsync is not a like a copy, it checks from both sides that the file to
copy is not the same on the other side. Copying something is straight
forward, it dont wait to a remote check of the files metadata or
attributes (last modification, permissions, etc). Access disk time could
explain this...
I dont know how you are using rsync... if you are using the network
protocol or a nfs mount.
check this also, if the rsync is using a single tcp connection for each
file, and you have a lot of small files, the problem could be the slow
start congestion algorithm of tcp (scp uses a single conection to copy,
im sure of that. i dont know how rsync protocol works)
Also, if you have a lot of small files, it will take a lot of access
time (disk seek time) for each to fetch this information and then start
the copy.
See if you can do a iostat on each server doing a scp and a rsync to see
the disk usage diference... if you find that you have a large queue on
disk when using rsync, thats the problem...
I hope this helps to point the way to solve your problem.
Saludos, Juan.
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
