Look Who it Was Random video

RAD Lab: Rain Workload Generation Toolkit

670 views · Published 11 March 2011 · 10:36 · Indexed 28 September 2026

Channel: CITRIS and the Banatao Institute · 2011 · Science & Technology

Watch on YouTube

Speaker: Rean Griffith

Abstract: 
Rain is an offshoot of our work on dynamic resource management in the Web Director and SCADS Director projects. In the context of those projects, the Directors need to be able to handle different workload regimes, while being able to adapt to transitions between regimes. In the context of cloud-computing we posit that handling workload variations is as important for dynamic resource management as is delivering good peak performance. We find that existing workload generation tools are not flexible enough to generate workloads that vary in three key dimensions relevant for making resource allocation decisions for Cloud Computing applications: variations in the amount of load, variations in the mix of operations performed by clients (e.g. changes in reads vs. writes or customer usage-patterns) and variations in the popularity of the data accessed, i.e., data hotspots. In this talk we describe the architecture and implementation of Rain, a statistics-based workload generation toolkit that uses parameterized and empirical distributions to model the different classes of workload variations. Our toolkit's implementation is flexible, allowing us to support the full spectrum of load generation strategies (closed-loop, open-loop and partly open-loop), time-varying loads in the three dimensions listed above (amount, mix and data), single-tenant as well as multi-tenant (aggregated) workloads and easy extensibility to target new systems/applications.

More from this channel