Machine Learning to Detect Complex Workloads in Real-time and its Applications

For cloud-based storage and compute systems optimal performance is possible when one understands characteristics of workload and dynamically take action based on the needs. To detect and classify mix of workloads for large number IO-streams in real-time is thus essentially a pattern recognition problem.

This paper presents application of supervised neural network that can recognize a variety of customer workloads. The network was trained to recognize workloads such as, video streaming, virtual desktops, analytics, file-share and databases. The method can also identify a mix of these workloads. Such a detection of workloads in real-time can be used to auto-tune the systems, optimal scheduling of jobs, remove performance bottlenecks and monitor service level agreements.

Learning Objectives:
1. How ML can be used to improve storage management and performance
2. Understanding supervised neural networks used for pattern classification
3. Applying this ML tool to help make practical decisions and to improve performance in in real-time