Machine Learning for Performance Prediction of Spark Cloud Applications

Abstract

Big data applications and analytics are employed in many sectors for a variety of goals: improving customers satisfaction, predicting market behavior or improving processes in public health. These applications consist of complex software stacks that are often run on cloud systems. Predicting execution times is important for estimating the cost of cloud services and for effectively managing the underlying resources at runtime. Machine Learning (ML), providing black box solutions to model the relationship between application performance and system configuration without requiring in-detail knowledge of the system, has become a popular way of predicting the performance of big data applications. We investigate the cost-benefits of using supervised ML models for predicting the performance of applications on Spark, one of today's most widely used frameworks for big data analysis. We compare our approach with Ernest (an ML-based technique proposed in the literature by the Spark inventors) on a range of scenarios, application workloads, and cloud system configurations. Our experiments show that Ernest can accurately estimate the performance of very regular applications, but it fails when applications exhibit more irregular patterns and/or when extrapolating on bigger data set sizes. Results show that our models match or exceed Ernest's performance, sometimes enabling us to reduce the prediction error from 126-187% to only 5-19%.

A parallelization model for performance characterization of Spark Big Data jobs on Hadoop clusters

“…Maros [22] conducted a cost-benefit analysis of a supervised machine learning model for Spark performance prediction and compared their results with Ernest [23]. In this investigation, they considered the black box and gray box techniques.…”

Advancing Design and Runtime Management of AI Applications with AI-SPRINT (Position Paper)

“…The tools will automate the AI application performance profiling and identify the ML model providing the highest performance prediction accuracy supporting model selection and hyper-parameters tuning. Preliminary results in [4,5,6] have shown that ML models allow to achieve good accuracy (with average percentage error between 5 and 15%) in cloud environments. AI-SPRINT will extend the use of such models to consider AI-based sensors and deep networks partitioned and deployed across computing continua.…”

“…Both private and public clouds will be valid targets. Applications will be described as OASIS TOSCA templates 4 describing the topology of their components and their software dependencies. Application templates will support restrictions for the deployment on multiple heterogeneous resources (i.e., including hardware accelerators, e.g., GPGPUs) also at the edge layer, by specific attributes (e.g., the target image, performance constraints or privacy requirements), which will be instantiated and managed by a single interaction with the AI-SPRINT framework.…”

An Enhanced Parallelisation Model for Performance Prediction of Apache Spark on a Multinode Hadoop Cluster

“…A cost-benefit Spark performance prediction model based on a machine learning algorithm was proposed by Maros [30]. They have proposed both black-box and grey-box models based on four machine learning algorithms.…”