EECS Special Seminar: Geoff Ramseyer: Scaling the Unscalable: Solving Worst-Case Contention with Better Economic Mechanisms

Details

• Date: Tuesday, April 9
• Time: 11:00 am - 12:00 pm
• Location: Grier Room A

Additional Location Details:

Abstract:
Brute-force engineering can squeeze only so much throughput from
applications with limited parallelism.  Instead, at some point, we need to
attack scalability from the other side: squeezing functionality from
scalable systems, even though this requires applying domain expertise at
much lower levels of software design.  In this talk, I show how
better economic mechanisms can achieve near-linear scalability on two
problems that previously flummoxed the fintech industry:
decentralized asset exchanges and general smart contracts.  In
both cases, trading a small amount of latency and semantics can
improve performance by orders of magnitude over state-of-the-art
systems.  Perhaps more surprisingly, the new semantics offer
additional security and economic benefits--including increased liquidity,
more efficient auditing, and the ability to avoid at least
one class of smart-contract vulnerability.  Ultimately, the key to
scalability lies in studying software architecture in tandem with the way a
system fits into the outside world.  Doing so opens exciting new
parallel systems architectures and challenges while simultaneously
raising intriguing theoretical and economic questions.

Bio:
Geoff Ramseyer is a postdoctoral researcher at Stanford University.
His research interests span  
building scalable, high-performance systems and analyzing
the economic tradeoffs implicit in computer system architectures.
He received his PhD from Stanford in 2023 and
his BS in 2017 from the University of Chicago.

Host

• Frans Kaashoek