RoboComAI

Cincinnati, OH, United States

RoboComAI

Cincinnati, OH, United States
SEARCH FILTERS
Time filter
Source Type

Fecko M.,Applied Communication science | Chang K.,Applied Communication science | Cichocki A.,Applied Communication science | Kim H.,Applied Communication science | And 11 more authors.
Proceedings of the International Telemetering Conference | Year: 2014

In an iNET telemetry network, Link Manager (LM) dynamically allocates capacity to radio links to achieve desired QoS guarantees. Under the T&E S&T iMANPOL program, we developed an enhanced capacity allocation algorithm that can better cope with severe congestion and misbehaving users and traffic flows. We compare the E-LM with the LM baseline algorithm (B-LM), which employs priority-weighted allocation. The B-LM is expected to perform well for the majority of traffic patterns, but does not prevent an ill-behaved traffic class from causing excessive latency on other radio links. The E-LM ensures that each class has a "guaranteed" portion of the total available bandwidth that is proportional to the weight of the class. If the traffic loading of a class is lower than its quota, the difference can be flexibly shared by other classes across multiple links. If the traffic loading of a class is higher than its quota, its demand may still be satisfied, provided that the capacity is not taken away from well-behaved traffic classes that stay below their quotas. The qualitative analysis shows the E-LM provides lower latencies for the well-behaved links in overloading conditions and increases the overall system throughput when the traffic is unbalanced. We conducted extensive experiments to confirm that analysis, with the E-LM reducing latency of well-behaved flows up to 90%, and increasing overall throughput up to 65% over the B-LM.

Loading RoboComAI collaborators
Loading RoboComAI collaborators