Debugging Big Data Systems for Big Data Analytics

This chapter unveils the intricate art of debugging big data systems for optimal analytics performance, providing a comprehensive guide to navigating real-world performance challenges. The exploration commences by delineating the critical debugging steps essential for identifying and resolving issues within big data systems. Focussing on the specific problems that can afflict these systems, such as data locality, resource heterogeneity, network issues, resource over-allocation, unnecessary speculation, and poor scheduling policies, the chapter dives into the intricacies of root cause analysis. Emphasising the importance of this analysis in the context of big data analytics, the narrative elucidates the systematic steps involved, accompanied by insightful details on tools and techniques, challenges, and considerations. The chapter explores available diagnosis tools tailored for big data systems, including Mantri, Texas Advanced Computing Centre (TACC) Stats, Data Centre Data Base (DCDB) Wintermute, and AutoDiagn, empowering practitioners to effectively diagnose and address complex issues in their analytics infrastructure.