Resource-efficient SANDboxing for instantaneously scaling serverless computing platforms

In this paper, we present a novel sandboxing approach for serverless computing that uses resources more efficiently and improves startup latencies by an order of magnitude compared to the state-of-the-art. This allows a platform implementing our approach to take full advantage of the serverless, event-based computing model by allocating and de-allocating resources at very high speeds as demand arrives/departs while providing sufficient isolation of function instances against each other. Such a platform enables the class of low-latency applications such as Virtual Reality or Augmented Reality Systems to effectively use serverless computing.