Special Seminar - Anurag Khandelwal
— 11:00am
Location:
In Person
-
Newell-Simon 4305
Speaker:
ANURAG KHANDELWAL
,
Assistant Professor, Department of Computer Science, Yale University
https://www.anuragkhandelwal.com/
Recent improvements in intra-rack interconnect technologies have driven advances in data center resource disaggregation, promising better resource utilization, support for hardware heterogeneity, and application resource elasticity. However, actualizing these benefits while ensuring application performance requires support from the operating system (OS). Unfortunately, existing approaches expose a hard tradeoff between application performance, resource elasticity, and application transparency.
Our approach to navigating this tradeoff space argues for a ground-up redesign of the OS stack, central to which is a performant shared memory abstraction. In this talk, I will first talk about MIND, our memory management subsystem for disaggregated architectures that places its logic in the interconnect fabric to enable efficient, disaggregated shared memory, achieving performance and resource elasticity for real-world workloads without application modifications. I will also describe CORD, an extension of MIND’s in-network approach to enable efficient release consistency for disaggregated memory that is being incorporated into real-world hardware coherence interconnects. I will then present Spirit, a multi-user framework for fair resource allocation that addresses a challenge unique to disaggregated memory systems — the interdependence between cache and disaggregated memory bandwidth resources, where larger cache allocations can reduce an application’s need for memory bandwidth, and vice versa. To this end, Spirit employs a novel algorithm that takes application-specific dependency between cache and network bandwidth into account and ‘trades’ cache and bandwidth resources across users at runtime to guarantee fairness. Finally, I will present some of our ongoing work on developing other OS abstractions for disaggregated architectures.
—
Anurag Khandelwal is an Assistant Professor of Computer Science at Yale University, where his group focuses broadly on problems in computer systems and networks. He received his Ph.D. from UC Berkeley, where he was advised by Ion Stoica. His work has been recognized by an NSF CAREER award, two NetApp faculty fellowships, a Roberts Innovation Award, best paper awards at USENIX Security’20, ISCA’23, EuroSys’24, ISCA’25, and an IEEE Micro Top Picks selection in 2024.
Faculty Host: Rashmi Korlakai Vinayak
For More Information:
karenl@andrew.cmu.edu