Zhou and Colleagues Work to Enhance Serverless Computing Efficiency

With high scalability and flexibility, serverless computing is becoming a promising computing model. However, as pointed out in a new study by TSYS School computer scientist Yi Zhou, existing serverless computing platforms initiate a container for each function invocation, which leads to a huge waste of computing resources. Examinations by Zhou and colleagues from Jinan University and Auburn University reveal that executing invocations concurrently within a single container can provide comparable performance to that provided by multiple containers (i.e., traditional approaches) while reducing memory resource waste and longer execution times. The researchers propose such a framework, FaaSBatch, that batches invocations and minimizes resource utilization by mapping groups of batched invocations into a single container. According to Zhou, "We evaluated the effectiveness and performance of FaaSBatch by comparing it to three state-of-the-art schedulers. Our experimental results show that FaaSBatch effectively and remarkably slashes invocation latency and resource overhead." In fact, as Zhou and his team discuss in the study, which is set to appear in a future issue of IEEE Transactions on Computers, FaaSBatch reduces the resource overhead of traditional approaches by anywhere from 43% to 98%, depending upon the particular comparison.