Research Interests#
Operating Systems, Computer Architectures, File and Storage Systems
Research Summary: The goal of my research focuses on fundamentally improving the performance and scalability of modern cloud systems. Today, cloud computing form the foundation of many critical computing infrastructures. However, existing technologies for memory, storage, and even CPU have high performance overhead and scalability issue, especially with increasingly fast hardware and complex stacks. The central theme of my research is to to close the performance gap between virtualized systems and native systems and enable scalable and flexible system software/hardware stacks, without losing isolation or compatibility. I am currently working on providing efficient and scalable support for emerging software and hardware.
Highlighted Works#
For a complete list of all my published works, please click here
.
This paper presents the Extensible Memory Translation (EMT) framework, a pragmatic framework atop Linux to empower different hardware schemes of memory translation such as radix tree and hash table. We port Linux’s memory management onto EMT and show that EMT enables extensibility without sacrificing performance. EMT enables us to understand the OS perspective of these architectures and further optimize their designs. Read More
This paper presents Mux, a new tiered file system that accesses different device types indirectly through device-specific file systems, rather than directly through device drivers. Despite introducing an additional indirection layer, we show that Mux significantly outperforms Strata, a research tiered file system, because it utilizes specialized production-ready file systems. Read More
This paper first presents a CXL-driven profiling solution to precisely and transparently count the number of accesses to every 4KB page and 64B word in CXL DRAM. Second, using the profiling solution, this work uncovers that (1) widely used CPU-driven page-migration solutions often identify warm pages as hot pages, and (2) certain applications have sparse hot pages, where only a small percentage of words in each of these pages are frequently accessed. Besides, this work demonstrates that the performance overhead of identifying hot pages is sometimes high enough to degrade application performance. Lastly, this work presents M5, a platform designed to facilitate the development of effective CXL-driven page-migration solutions. Read More
This paper presents HugeGPT, a software approach to reducing two-dimensional page table walk overhead in virtualized environments. HugeGPT ensures that page tables used in guest systems are physically held in the huge pages formed in the host. HugeGPT can efficiently reduce address translation overhead and improve application performance in virtualized clouds. Read More
This paper presents HugeGPT, a software approach to reducing two-dimensional page table walk overhead in virtualized environments. HugeGPT ensures that page tables used in guest systems are physically held in the huge pages formed in the host. HugeGPT can efficiently reduce address translation overhead and improve application performance in virtualized clouds. Read More
This paper identifies host-guest page size mismatch as a main cause of high TLB misses and low performance in virtualized systems. This paper presents Gemini, a VM-hypervisor-based technique to mitigate the issue. Gemini can reduce TLB misses by up to 83% and improve application performance by up to 126%. Read More
This paper proposes DASEC, a task scheduler for edge clouds. DASEC makes application performance less sensitive to the interference between workloads by detecting and protecting critical paths. DASEC can reduce the latencies of edge workloads by 32% ~ 52%. Read More