Short bio. Salvatore Dipietro is a thrid year PhD student at the Imperial College of London. His interests include performance modeling and prediction of distributed applications and in particular NoSQL databases. He previously earned a Master’s degree in advance computing research and a master in Computer security and Forensic. He also worked in industries for some years as Cloud consultant and Linux system administrator.
Thesis abstract: NoSQL databases have emerged as one of the most widely adopted database technologies used in Big Data. Solutions such as Apache Cassandra allows to store huge amount of data using a scalable decentralized architecture, suitable for cloud deployment. However, deploying and managing NoSQL databases in enterprises requires specialized techniques to ensure that operational costs and quality-of-service meet expectations. He has developed stochastic models to predict the performance and allocate computing capacity to popular NoSQL databases, including Apache Cassandra and ScyllaDB. To control operational costs, he has invented PAX, a new elastic auto-scaling mechanism that dynamically configures cloud resources used by the database by profiling resource usage and data partitions. Results indicate that PAX is able to reduce the operation cost up to 60% compared to a traditional deployment with static resources. He is actually working on a different demand estimation algorithm which looks on the system state of a dynamic system to infer the demands and creates models that reflect the real system behaviour.