David Shim

Hui Sub (David) Shim

About Me

I am a PhD student in Electrical Engineering at Stanford University, advised by Professor Philip Levis. My research focuses on rethinking memory abstractions at the hardware–software boundary to improve performance and efficiency in datacenter systems.

Before Stanford, I spent nearly a decade at Samsung Electronics in the Memory Division. As a firmware engineer and associate architect, I developed enterprise solid-state drives used worldwide and contributed to products such as the world’s first PCIe Gen5 enterprise SSD and the world’s first 32TB SAS enterprise SSD.

Research

My work centers on rethinking how memory systems interface with CPUs and accelerators.

Flow-Based Addressing for SmartNICs

SmartNICs speed up packet processing but are limited by the mismatch between variable-sized packets and page-based DRAM. I developed flow-based addressing, which provides CPUs with a contiguous, byte-addressable view of each flow without copying packets into new buffers. Using a flow translation table, a custom DMA engine, and a per-core flow cache, this design reduces redundant DRAM accesses. An FPGA prototype showed a 67% reduction in DRAM traffic and a 52% throughput gain, sustaining line-rate processing.

Memory Systems Beyond DRAM and Flash with CXL

I am exploring how Compute Express Link (CXL) can integrate new memory technologies such as ReRAM, MRAM, and PCM into mainstream systems. Unlike today’s incremental capacity extensions, CXL enables the OS to optimize based on each device’s unique latency, endurance, and persistence profile. This approach expands the memory hierarchy beyond the binary DRAM/flash divide, creating a flexible, multi-dimensional design space where emerging memories become first-class system resources.

Publications and Achievements

  • STM: Improving the Reliability of SSD through Selective Trim Manager (Samsung Best Paper Contest, 2022, first author).
  • CES Innovation Award (2022) for the world’s first PCIe Gen5 enterprise SSD.
  • Samsung President’s Award (2019) for the world’s first 32TB SAS enterprise SSD.
  • Multiple patents in SSD firmware and memory reliability.