Query Condition Cache
A predicate cache in DuckDB’s filter-pruning path that skips decompression of non-qualifying blocks for repeated analytical queries.
YUN-TANG(Andrew), Chang
張昀棠
About
I am currently a Computer Science and Information Engineering student at National Cheng Kung University. My work focuses on enhancing system performance while balancing resource-usage trade-offs, including improving network I/O in Confidential Virtual Machines (CVMs), database query optimization, and stateful autoscaling algorithms for streaming data systems.
I conduct research with the Secure System Lab at National Taiwan University (NTU), advised by Prof. Shih-Wei Li, and with the Computer Engineering research group in the Department of ECE at the University of Toronto, advised by Prof. Ashvin Goel. I also develop open-source DuckDB extensions — including Condition Cache and Table Inspector — with engineer Hao Jiang.
Research interests
My research interests span operating systems, database systems, and data streaming systems. More recently, I have been exploring systems support for AI workloads, particularly scheduling, caching, and resource management for efficient and predictable execution. I am also interested in how these ideas might extend to confidential VMs, where performance, resource efficiency, and isolation must be considered together. I am always open to research, open-source collaboration, and systems engineering opportunities.
Selected work
View all projectsVsock-net: Making Paravirtualized Network I/Os for Linux-based Confidential VMs Safe and Fast
Removed the large, attack-surface-heavy Linux network stack from confidential VMs’ Trusted Computing Base by delegating network traffic to the host stack, then optimized virtio-vsock and eBPF sockmaps to recover performance.