BOOM: experiences in language and tool design for distributed systems (keynote)

Abstract

With the rapid expansion of cloud infrastructure and mobile devices, distributed systems have quickly emerged as a dominant computing platform. Distributed systems bring significant complexity to programming, due to platform issues including asynchrony, concurrency, and partial failure. Meanwhile, scalable distributed infrastructure--notably "NoSQL" systems--have put additional burdens on programmers by sacrificing traditional infrastructure contracts like linearizable or transactional I/O in favor of high availability. A growing segment of the developer community needs to deal with these issues today, and for the most part developers are still using languages and tools designed for sequential computation on tightly coupled architectures. This has led to software that is increasingly hard to test and hard to trust. Over the past 5 years, the BOOM project at Berkeley has focused on making it easier to write correct and maintainable code for distributed systems. Our work has taken a number of forms, including the development of the Bloom programming language for distributed systems, tools for testing and checking distributed programs, and the CALM Theorem, which connects programmer level concerns of determinism to system-level concerns about the need for distributed coordination. This talk will reflect on this work, and highlight opportunities for improved collaboration between the software engineering and distributed systems research communities.