Prediction-driven resource provisioning for serverless container runtimes

In recent years Serverless Computing has emerged as a compelling cloud based model for the development of a wide range of data-intensive applications. However, rapid container provisioning introduces non-trivial challenges for FaaS cloud providers, as (i) real-world FaaS workloads may exhibit highly dynamic request patterns, (ii) applications have service-level objectives (SLOs) that must be met, and (iii) container provisioning can be a costly process. In this paper, we present SLOPE, a prediction framework for serverless FaaS platforms to address the aforementioned challenges. Specifically, it trains a neural network model that utilizes knowledge from past runs in order to estimate the number of instances required to satisfy the invocation rate requirements of the serverless applications. In cases that a priori knowledge is not available, SLOPE makes predictions using a graph edit distance approach to capture the similarities among serverless applications. Our experimental results illustrate the efficiency and benefits of our approach, which can reduce the operating costs by 66.25% on average.

I. Introduction

Serverless computing, and in particular Function as a Service (FaaS), is becoming an increasingly popular cloud programming model [1], [2], fueled by the recent demand to host services on provisioned cluster infrastructures and the paradigm shift towards interconnected IoT applications, devices and platforms [3]–[5]. It offers an intuitive, event-based interface for developing cloud-based applications, that makes the writing and deployment of scalable microservices easier and cost effective. This computing model has additional advantages including lower operational and deployment costs due to its unique pricing policy (based on a pay-as-you-use model) where users do not explicitly provision or configure virtual machines (VMs) or containers but they only get charged based on the number of resources consumed by the application functions during execution [6], [7]. The serverless computing model has been successfully adopted in a wide range of application domains, including, processing event streams, next-generation web services and applications [8], [9], etc. All major commercial cloud service providers are now offering serverless computing platforms, including AWS Lambda (https://aws.amazon.com/lambda/), Google Cloud Functions (https://cloud.google.com/functions) and Azure Functions (https://azure.microsoft.com/en-gb/products/functions/).