Design and Performance of a Split Protocol Architecture on Distributed Network Protocol 3 (DNP3)

DNP3 (Distributed Network Protocol) is a recent protocol having been standardized by the IEEE 1815 standard in 2012. It is employed predominantly in the electric utility industry. The advent of DNP3 over TCP/IP made many devices and networks available on the internet which prompted attacks and exploits by several researchers and organizations. In the period 2007 to 2009 there were 27 attacks [1]. In this paper we present an out of the box approach to increasing speed of response for DNP3 devices by using a split protocol design that can be selectively deployed in high traffic areas of a utility network. We implemented a 'mini' DNP3 protocol stack with a reduced feature set that facilitated Class 0 and Class 1 data polls. Master and Outstation devices with this 'mini' stack were built and tested with Third Party tools to verify compliance with the IEEE 1815 standard. 20 Class 0 counters were configured in the Outstations. The time taken to complete the Class 0 poll was measured with and without the split design. Measurements with the split design used 1, 2, 3 and 4 Data Servers. We saw a maximum improvement of 98% reduction in the time taken for the Class 0 poll using 2 Data Servers. The Class 0 poll duration decreased as we added more Data Servers going from 1 to 2 and 3 Data Servers showing scalability. We are encouraged to conclude that using a split design in DNP3 can significantly improve response times from Outstations.