The Advanced Technologies Section (ATS) of the National Center for Computational Sciences at ORNL is a world leader in developing and deploying scientific and technical solutions for leadership-class computing environments. The R&D activities of ATS are organized around designing and deploying leadership class systems, developing artificial intelligence solutions for science and smart facilities of the future, and stewardship of data and workflows at scale to enable science. The ATS Seminar Series is a forum for learning from experts and engaging with collaborators to advance their scientific mission.

Seminar Title:

How to leverage multi-tiered storage to accelerate I/O

Modern system architectures include multiple tiers of storage organized in a hierarchy. The goal is to mask the I/O gap between compute nodes and remote storage. However, this adds complexity to the end user resulting in under-utilization of these specialized I/O resources. In this talk, we will demonstrate how the existence of multiple tiers of storage, often organized in a hierarchy, can help mitigate the I/O bottleneck. We will present recent developments in two NSF-funded projects designed for and around multi-tiered storage.

Specifically, we will present Hermes, a heterogeneous aware, multi-tiered, dynamic, and distributed I/O buffering system. Hermes aims to remove the complexities associated with multi-tiered storage environments. It offers a simple, yet powerful, buffering API that abstracts the existence of tiers of storage. Further, the Hermes adapter layer is designed to support existing legacy I/O APIs (POSIX, STDIO, MPIIO) transparently to the user via interception. Hermes approaches its first public beta release and we are proud to demonstrate the developments in this area. Further, we will discuss ChronoLog, a new distributed and tiered shared log store. This project is motivated by the activity (or log) data explosion we experience today. This data production stems from the proliferation of modern sensors, IoT devices, web activity, mobile and edge computing, telescopes, as well as by computer-generated traffic (e.g., process synchronization, system utilization monitoring, error debugging, etc). ChronoLog uses physical time to provide total ordering on a log as well as auto-tiering across multiple storage tiers to seamlessly scale the capacity of a log. We will highlight how a distributed shared log can be used to build several other services that require a source of strong consistency, fast appends and “commit” semantics, and transactional isolation.

Speaker:

Anthony Kougkas

Research Assistant Professor, Department of Computer Science, Illinois Institute of Technology

Dr. Anthony Kougkas is a Research Assistant Professor and the Director of I/O Research in the Scalable Computing Software Laboratory at Illinois Institute of Technology. He earned his PhD from the Computer Science Department of Illinois Institute of Technology, advised by Dr. Xian-He Sun, and his dissertation is titled Accelerating I/O using Data Labels: A Contention-Aware, Multi-Tiered, Scalable, and Distributed I/O Platform. He holds a B.Sc. in Military Science and a M.Sc in Computer Science, both received in Athens, Greece. His research is focused in Parallel and Distributed systems, Parallel I/O optimizations, HPC systems, and Key-Value Store solutions. He currently manages several NSF-funded projects related to distributed and parallel I/O that leverages multi-tiered storage. He works closely with Argonne, Lawrence Livermore, and Sandia National Laboratories. Previous to his PhD, Anthony was a military officer serving in several positions in the Army and has extensive experience as an intelligence officer, operations and security, public relations, and personnel development and training.

Recorded Presentation:

Vimeo link: https://vimeo.com/574086376