Configurable dynamic load shedding method in distributed stream computing system

摘要

A computer implemented method of load shedding used in stream computing system that considers the relative importance of each of the applications processing the incoming input data or events. The method of load shedding method also accounts for system physical constraints, such as memories and CPU utilization. The load shedding method first observes the workload of each application and arriving rate of the incoming input data or events. If the system is under an overloading condition, calculate a input data or event drop ratio for each application such that the projected sum of all applications' workload will be at or below the system capacity when the unprocessed input data or events are dropped according to the drop ratio for each application.

主权项

1. computer implemented method for load shedding in a distributed stream computing system, comprising:
detecting a processing latency; calculating a target projection point for system load; if the processing latency is greater than a latency threshold:
calculating a drop ratio for each of one or more applications running in the system based on one or more drop ratio computation factors comprising:
the target projection point for system load,arriving rate of data or events,processing rate of data or events,amount of system resources for processing data or events, andsystem resource capacity;determining an incremental drop ratio delta for each of the one or more applications, wherein the incremental drop ratio delta is the drop ratio for the corresponding application divided by a number, the number for dividing the drop ratio being proportional to available buffer in the corresponding application for holding unprocessed input data or events;determining a load shedding percentage for each of the one or more applications;dropping a fraction of unprocessed data or events by the load shedding percentage for each of the one or more applications; repeating the determining of the load shedding percentage and the dropping of the fraction of unprocessed data or event for each of the one or more applications until the processing latency is not greater than the latency threshold, wherein the load shedding percentage is initially equal to the incremental drop ratio delta for the corresponding application and increments by the same delta for each cycle.