Yongsoo Joo

I/O Completion Optimization & Storage Security 2023 – present

DPAS: Dynamic I/O Completion Method for SSDs

Dynamically switches among polling, interrupts, and an adaptive hybrid polling method (PAS) based on CPU contention levels, achieving both prompt and CPU-efficient I/O completion on modern SSDs.

9% YCSB performance improvement on 3D XPoint SSD [USENIX FAST 2026]

Virtualized I/O Performance Optimization

Expands the polled I/O path in the Linux kernel to improve I/O performance in virtualized environments, enabling efficient storage access for virtual machines.

Polled I/O path expansion for Linux virtualization [ACM HotStorage 2024]

Ransomware-Proof SSDs

A memory-efficient overwrite detection method that enables SSDs to detect and prevent ransomware encryption attacks at the firmware level.

Memory-efficient overwrite detection at SSD firmware [IEICE 2023]

I/O Reordering & Mobile Storage 2015 – 2020

I/O Reordering and Interleaving for Application Prefetching

Exploits both I/O reordering and I/O interleaving techniques to optimize application prefetcher design, improving launch performance on storage devices.

Systematic prefetcher optimization via I/O scheduling [ACM TOS 2017]

Mobile Application Loading Optimization

Enlarges I/O request sizes to reduce the number of I/O operations during mobile application loading, accelerating app launch on mobile storage.

Faster mobile app loading via I/O size enlargement [IEEE ESL 2020]

SSD Performance & Nonvolatile Memory 2009 – 2014

FAST (Fast Application STarter)

An SSD-aware application launching scheme that overlaps CPU computation with I/O operations. Works on Linux systems without requiring kernel modifications.

28% reduction of application launch time [USENIX FAST 2011]

SSD Parallelism Exploitation

An optimized application prefetcher for solid-state drives that exploits internal parallelism of SSDs to accelerate application launches.

37% reduction of prefetcher execution time, 18% reduction of launch time

IOLab: Active Storage Device Evaluation Platform

A virtual machine-based system that evaluates intelligence functions such as prefetching and data deduplication. The platform separates intelligence function implementation from specific operating systems using file-backed virtual disk images.

23% improvement over the Windows prefetcher [IEEE TC 2014]

Hybrid Solid-State Storage Architecture

Combines phase change memory (PCM) and flash memory in a hybrid solid-state storage architecture to simultaneously improve performance, energy consumption, and device lifetime.

Performance, energy, and lifetime co-optimization for hybrid PCM/Flash storage [HPCA 2010]

Hybrid PRAM and STT-RAM Cache Architecture

Extends phase change RAM (PRAM) cache lifespan by integrating spin transfer torque RAM (STT-RAM) write buffers through threshold-based ownership management.

39% longer lifetime with only 4% larger die area [IEEE CAL 2013]

Endurance-Aware PCM Cache Design

Addresses phase change memory limitations through write-traffic reduction techniques to improve both energy efficiency and device lifetime.

8% energy saving and 3.8 years of lifetime [DATE 2010]

Hybrid Disk Block Allocation

Optimizes HDD and flash memory allocation for application launches by pinning selective subsequences to minimize seek time.

31% reduction of launch times with only 26% flash capacity [CODES+ISSS 2009]

Energy-Efficient Memory Systems 2002 – 2008

Flash Memory Demand Paging & Energy Optimization

Designed energy- and performance-optimized demand paging for OneNAND flash-based execute-in-place systems, including delayed dual buffering to reduce page faults.

Energy and performance co-optimization for OneNAND flash [IEEE TCAD 2008]

Energy-Aware Data Compression for MLC Flash

Developed energy-optimal prefix coding for multi-level cell flash memory that minimizes programming energy rather than compressed data size, using integer linear programming.

Energy-optimal compression via unequal bit-pattern costs [DAC 2007, ACM TODAES 2008]

Low-Power SDRAM Memory Systems

Explored and reduced energy consumption of off-chip SDRAM memory systems for embedded applications through system-level energy characterization and optimization techniques.

Systematic SDRAM energy exploration and reduction [DAC 2002, ACM TECS 2003]