EE Distinguished Speakers Seminar Series of IEL-EPFL: How the cloud is driving software and hardware specialization - a use case from the airline industry
Event details
| Date | 06.03.2020 |
| Hour | 13:15 › 14:30 |
| Speaker | Prof. Gustavo Alonso, ETH Zürich, Switzerland |
| Location | |
| Category | Conferences - Seminars |
Abstract:
Increasingly demanding workloads and the prevalence of data center and cloud computing as the primary computing platforms open up many opportunities for specialization at all levels of the stack, from hardware to software. In this talk, I will explain this trend as well as discuss the technology and economic reasons for it. We will then explore some of the challenges associated to modern search engines in terms of scalability, efficiency, and latency requirements. Using two use cases from the airline industry developed together with Amadeus (decision tree ensembles and rule engines), I will show how we have been able to develop proof-of-concept solutions based on FPGAs that represent orders of magnitude improvements along several dimensions (performance, efficiency, cost) over the state of the art. The system has been tested on premises using different FPGA products and also on the cloud (Amazon's F1 instances), showing where the strengths of the approach lie and where more could be invested to improve performance. Spectacular as the results are, they also shed light on important architectural questions regarding specialization and hardware acceleration: unified memory spaces, cache coherency, connectivity, and overall management. I will go over these questions and discuss several of our lines of research exploring this promising space.
Short Bio:
Gustavo Alonso is a Professor of Computer Science at ETH Zürich where he is a member of the Systems Group (www.systems.ethz.ch) and the Head of the Institute for Computing Platforms. He has a degree in electrical engineering from the Madrid Technical University as well as a M.S. and Ph.D. degrees in Computer Science from UC Santa Barbara. Gustavo's research interests encompass almost all aspects of systems, from design to run time. He works on distributed systems, data processing, and system aspects of programming languages. Most of his research these days is related to data processing on data centers and the cloud as well as hardware acceleration using FPGAs. Gustavo has received numerous awards for his work, including four Test-of-Time awards for contributions to databases, programming languages, mobile computing, and systems. He is a Fellow of the ACM and of the IEEE as well as a Distinguished Alumnus of the Department of Computer Science of UC Santa Barbara.
Increasingly demanding workloads and the prevalence of data center and cloud computing as the primary computing platforms open up many opportunities for specialization at all levels of the stack, from hardware to software. In this talk, I will explain this trend as well as discuss the technology and economic reasons for it. We will then explore some of the challenges associated to modern search engines in terms of scalability, efficiency, and latency requirements. Using two use cases from the airline industry developed together with Amadeus (decision tree ensembles and rule engines), I will show how we have been able to develop proof-of-concept solutions based on FPGAs that represent orders of magnitude improvements along several dimensions (performance, efficiency, cost) over the state of the art. The system has been tested on premises using different FPGA products and also on the cloud (Amazon's F1 instances), showing where the strengths of the approach lie and where more could be invested to improve performance. Spectacular as the results are, they also shed light on important architectural questions regarding specialization and hardware acceleration: unified memory spaces, cache coherency, connectivity, and overall management. I will go over these questions and discuss several of our lines of research exploring this promising space.
Short Bio:
Gustavo Alonso is a Professor of Computer Science at ETH Zürich where he is a member of the Systems Group (www.systems.ethz.ch) and the Head of the Institute for Computing Platforms. He has a degree in electrical engineering from the Madrid Technical University as well as a M.S. and Ph.D. degrees in Computer Science from UC Santa Barbara. Gustavo's research interests encompass almost all aspects of systems, from design to run time. He works on distributed systems, data processing, and system aspects of programming languages. Most of his research these days is related to data processing on data centers and the cloud as well as hardware acceleration using FPGAs. Gustavo has received numerous awards for his work, including four Test-of-Time awards for contributions to databases, programming languages, mobile computing, and systems. He is a Fellow of the ACM and of the IEEE as well as a Distinguished Alumnus of the Department of Computer Science of UC Santa Barbara.
Practical information
- General public
- Free
Contact
- Elison Matioli: [email protected] Philippe Gay-Balmaz: [email protected]