Optimizing Serverless Computing and Operating Systems

The field of serverless computing and operating systems is witnessing significant advancements, driven by the need for efficient, scalable, and cost-effective solutions. Researchers are exploring innovative approaches to optimize serverless computing, including the design of new operating systems, execution environments, and resource management techniques. A key direction is the development of lightweight and high-density serverless platforms, which can minimize cold start latency and reduce memory usage. Another area of focus is the improvement of snapshot and restore mechanisms, enabling faster and more reliable state restoration. Additionally, there is a growing interest in leveraging specialized hardware and architectures, such as CXL and GPUs, to accelerate serverless workloads. Noteworthy papers in this area include: TrEnv, which presents a co-designed serverless platform that reduces startup latency and memory usage through repurposable sandboxes and memory templates. Taming Serverless Cold Starts Through OS Co-Design, which introduces Spice, an execution engine that delivers near-warm performance on cold restores from disk, reducing latency by up to 14.9x over state-of-the-art process-based systems.

Sources

{\mu}Fork: Supporting POSIX fork Within a Single-Address-Space OS

TrEnv: Transparently Share Serverless Execution Environments Across Different Functions and Nodes

WebAssembly and Unikernels: A Comparative Study for Serverless at the Edge

XBOF: A Cost-Efficient CXL JBOF with Inter-SSD Compute Resource Sharing

Taming Serverless Cold Starts Through OS Co-Design

Cost-Performance Analysis: A Comparative Study of CPU-Based Serverless and GPU-Based Training Architectures

Built with on top of