GridStat Middleware Communication Framework: Systematic Adaptation
GridStat is a middleware communication framework with ultra-low latencies and high availability that is aimed at providing wide-area data delivery capabilities for the power grid. GridStat’s data plane is a tightly managed mesh overlay network that provides stringent, rate-based delivery guarantees. However, the data plane components are susceptible to arbitrary (Byzantine) failures and cyber-attacks that, if not addressed, have the potential to make those guarantees unachievable. Furthermore, even non-malicious changes within the operating environment—for example, a sudden burst of large subscription requests triggered by a power contingency or benign component failures—may also force reconfiguration in order to meet the guarantees, particularly for the most important applications, given the present power and cyber conditions. The objective of this activity was to develop adaptation services and supporting instrumentation services for GridStat in order to systematically adapt to changing conditions and available resources. In pursuit of that goal, we modeled and assessed the performance characteristics of GridStat under various constraints that affect normal functionality; determined the level of instrumentation that maximizes adaptation-related evidence-gathering with minimum effects on data delivery performance; studied existing Security Information Event Management (SIEM) and Complex Event Processing (CEP) techniques to discover analogous compound adaptation triggers based on multiple kinds of instrumentation inputs; and implemented an adaptation service for GridStat that is highly tailorable both in the steady state and under changing conditions. We also explored the use of utility functions to optimize the benefit of the data delivery service over an entire grid, and the use of pre-computed information on failures and their effects.
GridStat on DETER: Presented by Dave Bakken, Washington State University, during the TCIPG Industry Workshop in October 2012.