There are times when an administrator may want to move data between two volumes that are located on the same storage system. There are a slew of options, depending on how the move should be done but here, we will limit ourselves to SnapMirror.
Based on a few tests I have conducted, it appears that when doing local SnapMirror (meaning SnapMirror from and to the same controller), the loopback interface is indeed used by default. Run the command “netstat -I lo -i 5” while the system is doing a local snapmirror and you’ll see what I mean.
Next question then is
What is the maximum throughput that I can expect out of a local snapmirror over loopback ?
Let’s see. On my trusty FAS980, the maximum window size I can set is 1994752 bytes. I also ran a ping -s 127.0.0.1 and saw that I am seeing an average round trip time (RTT) of 0.060ms.
The formula I am going to use is
Bandwidth (bytes/s) * Delay (s) = Window Size (Bytes)
Bandwidth = 1994752 / ( 0.060 / 1000 ) = 33.24 GB/s
So a FAS980 loopback adapter gives me about 33.24 GB/s (266Gbps) throughput ?
[...] Data ONTAP 7.2.5, the SnapMirror TCP receive window size has been increased. Recall in my earlier posting about SnapMirror over loopback, we tried to ascertain the theoretical bandwidth of the SnapMirror [...]