I'm interested in operating systems and have been doing some systems stuff for a while. Datacenter I/O intensive applications have particular idiosyncrasies, big memory footprint, micro-second scale service time, high fanout pattern are just a few to name. Such characteristics introduce new challenges, and I worked on a few research projects targeting the challenges. I worked on memory systems, mostly designing better memory management mechanisms (address translation and virtual memory) for pointer chasing applications at NUS School of Computing. Prior to the memory systems research, I spent some time to address long-tail latency problems of data center applications in DSLab at IUST.
I recently started reading about datacenter networking. Particularly, I'm investigating datacenter congestion and flow control mechanisms. Meanwhile, I'm looking for a night watch!
- SPARTA: A Divide and Conquer Approach to Address Translation for Accelerators - TerrierTail: Mitigating Tail Latency of Cloud Virtual Machines - CTS: An Operating System CPU Scheduler to Mitigate Tail Latency for Latency-Sensitive Multi-Threaded Applications - Yawn: A CPU Idle-state Governor for Datacenter Applications - Experimental Validation of the Suitability of Virtualization-based Replication for Fault Tolerance in Real-time Control of Electric Grids
- Checkout my new blog post, Remote Page Table ~= Local Page Table - 27/12/2019 - Happy to announce that our paper Yawn: A CPU Idle-state Governor for Datacenter Applications was accepted to APSys'19 - 06/15/2019