Despite unifying frameworks like Akka, distributed systems often look like a patchwork of special-purpose tools, libraries, and frameworks. For example, the essential task of persisting objects by pickling (or serializing) them to a binary or text representation is typically outsourced to a third-party serialization framework. Many of these frameworks, like Google's Protocol Buffers, lack support by the Scala compiler or integration with the Scala standard library, resulting in suboptimal performance, unsatisfying Scala support, or both. Furthermore, to obtain good performance using one of these frameworks, it's often necessary to roll your own type-specialized custom serializers, requiring significant boilerplate. This talk presents a new pickling framework designed for Scala with a few attractive properties: (1) using the framework requires little to no boilerplate, (2) using Scala's implicit parameters, users can add their own easily-swappable pickle format enabling users to persist to a number of formats, binary, JSON, or your own custom format, (3) using the type class pattern, users can provide their own custom picklers to override the default behavior of the pickling framework, (4) static, macro-based generation of picklers enables significant performance improvements. In microbenchmarks our framework achieves a factor 6 speedup over Java Serialization, and performs on par or up to 3 times faster than popular fast Java serialization frameworks like Kryo. We'll show how this new framework can be used as a drop-in replacement for other serialization frameworks, and what's in store for Java compatibility. Finally, we'll present an experimental abstraction, called spores, well-behaved, composable pieces of functional behavior that can be cleanly pickled, transported over the wire, unpickled, planted, and more.

No outline is available