Architecture-Specific Performance Optimization of Compute-Intensive FaaS Functions

Abstract

FaaS allows an application to be decomposed into functions that are executed on a FaaS platform. The FaaS platform is responsible for the resource provisioning of the functions. Recently, there is a growing trend towards the execution of compute-intensive FaaS functions that run for several seconds. However, due to the billing policies followed by commercial FaaS offerings, the execution of these functions can incur significantly higher costs. Moreover, due to the abstraction of underlying processor architectures on which the functions are executed, the optimization of these functions is challenging. As a result, most FaaS functions use pre-compiled libraries generic to x86-64 leading to performance degradation. In this paper, we examine the underlying processor architectures for Google Cloud Functions (GCF) and determine their prevalence across the 19 available GCF regions. We modify, adapt, and optimize a representative set of six compute-intensive FaaS workloads written in Python using Numba, a JIT compiler based on LLVM, and present results wrt performance, memory consumption, and costs on GCF. Results from our experiments show that the optimization of FaaS functions can improve performance by 18.2x (geometric mean) and save costs by 76.8% on average for the six functions. Our results show that optimization of the FaaS functions for the specific architecture is very important. We achieved a maximum speedup of 1.79x by tuning the function especially for the instruction set of the underlying processor architecture.

Towards Demystifying Intra-Function Parallelism in Serverless Computing

“…Evaluating the general performance of FaaS platforms and improving the performance of FaaS functions has also been actively researched [8,10]. In [10], the authors present the Serverless Application Analytics Framework (SAAF), to improve observability on the performance of FaaS functions on commercial FaaS platforms.…”

“…SAAF currently supports multiple FaaS platforms and several different programming languages. In our previous work [8], we examined the underlying processor architectures on GCF and demonstrated the usage of Numba [21], a Just-in-Time (JIT) compiler based on LLVM for optimizing and improving the performance of compute-intensive Python based FaaS functions. We showed that the optimization of FaaS functions can improve performance by 18.2x and save costs by 76.8%.…”

“…A possible explanation for observing two CPU cores at lower memory configurations for the different services, i.e., AWS Lambda and GCF can be hyperthreading or Simultaneous Multithreading (SMT) [11] present in modern Intel Server Family of Processors, i.e, Haswell-EP, Broadwell-EP, Skylake-SP, and Cascade Lake-SP. As shown in previous works [8,18,33], these are the family of processors found in the Virtual Machines of the commercial FaaS providers on which the function/container instances are launched.…”

“…Furthermore, after a certain memory configuration (>2048MB) allocating more memory does not considerably affect the function execution time. Serverless FaaS/CaaS platforms launch the function instances on the platform's traditional Infrastructure-as-a-Service (IaaS) virtual machines (VM) offerings [8,18]. However, the provisioning of such VM's is abstracted away from the user.…”

FedLess: Secure and Scalable Federated Learning Using Serverless Computing

“…These containers are commonly referred to as function instances [22], [23] 1 https://github.com/andreas-grafberger/fedless and contain the function's language-specific runtime environment. For commercial FaaS platforms, the function instances are launched on the FaaS platforms' traditional Infrastructureas-a-Service Virtual Machines as described in [24], [25]. The runtime is responsible for relaying the invocation events, context information, and responses between the FaaS platform and the function.…”