Load balancing for real-time, location-based event processing on cloud systems

For large-scale, and real-time processing, cloud systems are widely used due to their high scalability and availability. In this paper, we propose workload distribution methods for location-based event processing on the cloud systems. We define a measure of the workload, and focus on the balanced distribution of workload because in cloud systems the workload distribution is very important with respect to the system performance. For the balanced distribution of workload, we propose four methods: (1) round-robin data distribution, (2) round-robin query distribution, (3) data/query distribution via space partitioning and (4) skew-aware distribution. The round-robin data distribution method focuses on a balanced distribution of event data whereas queries are replicated in all cluster nodes. In the round-robin query distribution method, queries are evenly distributed whereas the event is replicated. The data/query distribution via space partitioning distributes event data and queries based on their spatial attribute values. Lastly, the skew-aware distribution method considers the non-uniformity of event data and queries. With extensive experiments, we evaluate the performances of our proposed methods.