who can really use this stack to communicate with kuka iiwa robot smoothly???

[BoWangFromMars]

I have read the java code running on the kuka iiwa controller, it is based on rosjava. When i wrote a rosnode on PC (Ubuntu OS) to publish joint position to the robot (50 Hz), and I also printed the time interval data in the callback function of Subscriber node, the time interval shown on the pendent ranges from 5 ms to 50 ms. It is so unsteady that the robot runs in jerky motion.
And I suppose that the communication goes wrong. So I write a simple test code only based on TCP/IP, not in the frame of ROS, which shows the network works well.
Further, I suppose that the author created too many threads in rosjava code, which makes OS on the robot controller call the callback function not in real time. So I simplified the java code to only subscribe the joint position commands, with killing all other ros nodes and threads, the time interval for calling the callback function of Subscriber node still varies in a large range. Unfortunately, calling the callback function is compelted by OS, I cannot do nothing about it.
Now, I suspect that the communication scheme based on ROS is really feasible??? Or my robot controller works in wrong way? Who can stand up to verify the effectiveness of this scheme?