Random hanging and low speed of syncing

Hello! We have a workbench with two servers:

Server 1: device1

CentOS release 6.6 (Final)
2x Intel® Xeon® Processor E5320 (8M Cache, 1.86 GHz, 1066 MHz FSB)
12Gb memory:
    2x 4GB DDR2-667MHz
    2x 2Gb DDR2-667MHz

Server 2: device2

FreeBSD 10.1-RELEASE
Intel(R) Core(TM)2 Quad CPU    Q9400  @ 2.66GHz
12Gb memory:
    2x 4GB DDR2-800MHz
    2x 2Gb DDR2-800MHz

We have syncthing v0.11.5 installed on both servers. Servers are directly connected via 1000BASE-T twisted pair. Firewall is turned off on both servers. We are trying to syncronize 6Tb of data (about 50K files of size from 100kb to 10Gb) stored on device1 with device2 When Syncthing is running we see full utilization of physical memory and 3 of 8 physical cores loaded at 100% on "device1" Maximum transmission speed we ever noticed is 170Mbps. Maximum size of successfully transfered data is 1Tb. Both “device1” and “device2” randomly lose connections Sometimes sync just freezes while connections are still active. It can be resolved either with restart Syncthing on both servers or, sometimes, with just entering Syncthing web interface. We’re having same issue with all versions of Syncthing. With rsync we have a stable transfer at 500-600Mbps

Links to syncthing config files from both servers are below

"Device1" http://pastebin.com/N9Mgpmvh

"Device2" http:// pastebin.com/6ezhjGqs

Can you see something useful in the logs? You can try enabling STTRACE=model environment variable which should give more verbose logging which might help, though pipe the output to file, as there will be a lot of it.

Regarding the speeds, it would be interesting to see where the bottleneck is, if you check IO stats, memory consumption and CPU consumption, and see which one is maxed out, and perhaps provide a CPU profile to see where most of the time is being spent.

Upgrade immediately! v0.11.7 contained a significant performance improvement for syncing, especially for large files.