OPTIMIZATION OF POLICY MANAGEMENT ON ISTIO SERVICE MESH

• Main Author: Philip Gunadi, Michael
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/81782
• Institution: Institut Teknologi Bandung

As development of software goes fast, new architecture is needed to fulfill application’s requirements. Microservice architecture come to answer the limitation of monolithic architecture in building increasingly complex application. Microservice architecture enables the development of specific purpose and independently developed application. Various technologies have emerged to support this architecture, such as kubernetes which is an orchestration tool to automates deployment, scaling and application management process. The presence of kubernetes make microservice application even more sophisticated and open up huge opportunities for even larger application. Microservice nature that divide application into smaller one brings a new challenge on how to establish communication, later known as east-west communication, that enables those services to be integrated. Service mesh technology emerges to help kubernetes in providing such communication service and management. One of the widely known and used service meshes is Istio, which implement sidecar architecture in its implementation. This architecture introduce overhead in which every service or application will have sidecar in their pod, that has a role as a communication broker that will receive and transmit messages. This research tries to optimize Istio by reducing the overhead introduced by reducing the number of sidecar involved on microservice communication without disabling or removing Istio’s communication policy enforcement capability. It was found that by minimizing the number of sidecar involved in the context of study, there was a 20% reduction in end-to-end latency, 14.4% to 66% reduction in CPU usage, 23.1% to 29.75% reduction in RAM usage, and up to 7.2% reduction in network I/O pressure. Therefore it’s concluded that Istio optimization can be done by reducing the number of sidecar involved during communication.