How Low Can You Go? Practical cold-start performance limits in FaaS
share
Function-as-a-Service (FaaS) has recently emerged as a new cloud computing paradigm. It promises high utilization of data center resources through allocating resources on demand at per-function request granularity. High cold-start overheads, however, have been limiting FaaS systems' such potential. Prior work has recognized that time redundancy exists across different cold function invocations and has proposed varied snapshots that capture the instantaneous execution state so allow for jump-starts through restoration. However, it remains unclear what the cold-start performance limits are as the previous snapshots employ different techniques and aredesigned for different environments. In this paper, we summarize these snapshots from a taxonomic perspective andpresent a model that depicts the cold-start performance fromfirst principles. To approximate the performance limits, we propose a snapshot design SnapFaaS. We evaluate SnapFaaS using real-world FaaS functions. Our empirical results prove SnapFaaS' efficiency. It is 2-10x as fast as prior work for most functions and achieves near-optimal cold-start performance.
READ FULL TEXT