Queryable Kafka: An agile data analytic pipeline for mobile wireless networks

Owing to their promise of delivering real-time decisions, today's streaming analytics platforms are increasingly being used in the communications networks, where the impact of the decisions go beyond sentiment and trend analysis to include real-time detection of security attacks and prediction of network state (i.e., is the network transitioning towards an outage). Current streaming analytics platforms operate under the assumption that arriving traffic is small (order of kilobytes) produced at very high frequencies. However, communications networks, especially the telecommunication networks, challenge this assumption because the arriving traffic in these networks is big (order of gigabytes) but produced at medium to low velocities, and furthermore, these large datasets may need to be ingested in their entirety to render prediction decisions in (near) real-time. Our inter- est is in subjecting today's streaming analytic platforms - constructed from state-of-the art open software components (Kafka, Spark, HDFS, ElasticSearch) - to traffic densities observed in such communications networks. We find that handling filtering on such large datasets is best done in a common upstream point instead of being pushed to, and repeated, in downstream components. To this extent, we modify Apache Kafka to perform limited native data transformation and filtering, relieving the downstream Spark application from doing this. We show that our approach out-performs the "out-of-box" analytic pipelines with negligible overhead compared to standard Kafka.