SnapMirror Replication - TCP Window problem

[Clifford McGlamry]

Customer has two WAAS units running 4.4.3 code.  Initial goal was to optimize snap mirror replication traffic over a 1 gig link. 

When the units are rebooted, the initial acceleration is great....75% reduction and good throughput.  But after a few hours, the maximum through put we get is around 120mbps....about exactly what you'd get if you calculate the BDP.

Went in and manually set for high BDP values, and restarted things.  Same result. 

I really scratching my head as to why this thing is not handling the TCP Windowing correctly.  There is no current documentation on SnapMirror optimization with WAAS (only version I could find is for code version 4.0.19, and that's so old it shouldn't even be posted).  The documentation sets available are pretty non helpful on this issue as well.

Looking for any suggestions on how to troubleshoot and correct this. 

TIA

Cliff

[adrianjames]

I can't help with the problem but when I noticed poorer than expected throughput with snapmirror, I ended up setting the optimization to TFO only and turning on snapmirror compression. I would think that DRE would not be too useful as deduplication on the netapp should take care of duplicate blocks before they are sent.