An objective comparison of two prominent virtual actor frameworks: Proto.Actor and Orleans

Recently there has been a significant increase in developing distributed systems easily and rapidly. Driven by the demand of software communities, developers seek tools and frameworks that abstract away low-level details of the underlying distributed system and the need to understand complex details on how the system works. Researchers have explored serverless frameworks, distributed key value stores, distributed stream processing frameworks and distributed actor frameworks. Currently, stateful serverless applications and distributed actor models may be the answer to what developers need. In this paper, we present a review of stateful distributed computing frameworks, and the results of experiments that compare Orleans and Proto.Actor - two popular actor model frameworks - running on Kubernetes. We discovered that the Proto.Actor performs at least two times faster than Orleans, but is more complex to learn. We present the results of these tests, and provide a discussion of future research opportunities highlighting virtual actor model frameworks.