Picocenter: supporting long-lived, mostly-idle applications in cloud environments

Abstract

Cloud computing has evolved to meet user demands, from arbitrary VMs offered by IaaS to the narrow application interfaces of PaaS. Unfortunately, there exists an intermediate point that is not well met by today's offerings: users who wish to run arbitrary, already available binaries (as opposed to rewriting their own application for a PaaS) yet expect their applications to be long-lived but mostly idle (as opposed to the always-on VM of IaaS). For example, end users who wish to run their own email or DNS server. In this paper, we explore an alternative approach for cloud computation based on a process-like abstraction rather than a virtual machine abstraction, thereby gaining the scalability and efficiency of PaaS along with the generality of IaaS. We present the design of Picocenter, a hosting infrastructure for such applications that enables use of legacy applications. The key technical challenge in Picocenter is enabling fast swapping of applications to and from cloud storage (since, by definition, applications are largely idle, we expect them to spend the majority of their time swapped out). We develop an ActiveSet technique that prefetches the application's predicted memory working set when reviving an application. An evaluation on EC2 demonstrates that using ActiveSet, Picocenter is able to swap in applications in under 250 ms even when they are stored in S3 while swapped out.