Tonegawa: Highly Scalable Distributed Web Server with Data Stream Processing

We developed Tonegawa, which is a distributed Web server on top of a stream processing system called System S under development by IBM Research. Tonegawa takes staged execution approach and a web request is processed over multiple stages, each of which can be processed by one multi-core node or even distributed to a cluster of nodes. We conducted the performance evaluation of Tonegawa with the use of both static contents and dynamic contents, and showed that Tonegawa on multiple nodes can be utilized in the case where large computational effort is required. In addition, we demonstrated the qualitative software productivity with regards to the implementation of a web server using such a stream-style programming, which decreases potential bugs and brings highly extensibility to the system.