IRIS-HEP Fellow: Eshan Bhargava



Fellowship dates: Jun – Sep, 2021

Home Institution: University of Southern California


Project: SkyhookDM - Ability to push back query execution to Client in case of overloaded OSDs

Currently, SkyhookDM allows pushing down compute operations such as selection and projection into the Ceph file system, which is the Storage Layer. However, when a large number of clients try to push down computation into OSDs at a time, the CPU and memory pressure of the OSDs may quickly increase, causing run-time side effects such as blocked and slow OSD operations. The goal of this project is to create functionality, such that if there is high CPU and memory pressure in the OSDs, for which they cannot process filters or expressions efficiently, the query execution is pushed back to the client for processing.

More information: My project proposal

Mentors:

  • Jeff LeFevre (UC Santa Cruz)

  • Jianshen Liu (UC Santa Cruz)

  • Jayjeet Chakraborty (NIT, Durgapur)

Presentations and Publications
Current Status
April 2023 - Incoming Machine Learning Intern at NASA Jet Propulsion Laboratory

Contact me: