Benchmarking key-value stores

Posted on by cvrweb

This project analyze the shortcomings of using IRM-based benchmarkings to evaluate cloud key-value stores. Also propose a trace replay model suitable for these systems, and develop KV-replay, an open-source replayer that implements this model.
Team: Cristina L. Abad, Edwin F. Boza, César San-Lucas, José A. Viteri

Clustering of Stochastic Processes from Request Streams for Workload Modeling and Synthetic Generation

Posted on by cvrweb

The goal is to improve storage workload modeling via unsupervised clustering of stochastic processes, with the goal of synthetic workload generation to improve the state-of-the-art in benchmarking and simulation based evaluations. This project is funded through a Google Faculty Research Award
Team: Cristina Abad, Edwin Boza, José Viteri, Jorge Cedeño, Sixto Castro, César San Lucas

Towards autonomous distributed storage systems

Posted on by cvrweb

We are working on building improved distributed storage systems (object stores, file systems, key-value stores). Our work is mostly focused on dynamic and autonomic methods that can be used by storage systems to maximize performance as the workload changes, adapting to client demands. Team: Cristina Abad, Edwin Boza, Jorge Murillo, Luis Lucio, Gustavo Totoy, Jordy Vásquez, Andrés Abad