A Workload-Specific Memory Capacity Configuration Approach for In-Memory Data Analytic Platforms (1712.05554v1)

Abstract: We propose WSMC, a workload-specific memory capacity configuration approach for the Spark workloads, which guides users on the memory capacity configuration with the accurate prediction of the workload's memory requirement under various input data size and parameter settings.First, WSMC classifies the in-memory computing workloads into four categories according to the workloads' Data Expansion Ratio. Second, WSMC establishes a memory requirement prediction model with the consideration of the input data size, the shuffle data size, the parallelism of the workloads and the data block size. Finally, for each workload category, WSMC calculates the shuffle data size in the prediction model in a workload-specific way. For the ad-hoc workload, WSMC can profile its Data Expansion Ratio with small-sized input data and decide the category that the workload falls into. Users can then determine the accurate configuration in accordance with the corresponding memory requirement prediction.Through the comprehensive evaluations with SparkBench workloads, we found that, contrasting with the default configuration, configuration with the guide of WSMC can save over 40% memory capacity with the workload performance slight degradation (only 5%), and compared to the proper configuration found out manually, the configuration with the guide of WSMC leads to only 7% increase in the memory waste with the workload's performance slight improvement (about 1%)