#Awarded
#Featured
- Crystal sea's glass bottle
- Christmas Q25
- Where to Go
- Korea Computer Congress 2023
- Contributing to llama.cpp: CANN SSM_CONV Operator on Ascend NPU
#Web
#UX/UI
#Lead
- Haejihae
- Christmas Q25
- Where to Go
- init 3rd
- Sungshin Women's University MakeUs Challenge 3rd
- Better Me
#App
#Launched
#Backend
#HugePage
#paper-review
- Caladan: Mitigating Interference at Microsecond Timescales
- Cold Start Influencing Factors in Function as a Service
- Architectural Implications of Function-as-a-Service Computing
- FaaSNet: Scalable and Fast Provisioning of Custom Serverless Container Runtimes at Alibaba Cloud Function Compute
- Parallelizing Packet Processing in Container Overlay Networks
- FlashCube: Fast Provisioning of Serverless Functions with Streamlined Container Runtimes
- Memory Efficient Fork-based Checkpointing Mechanism for In-Memory Database Systems
- Coordinated and Efficient Huge Page Management with Ingens
- MEGA: Overcoming Traditional Problems with OS Huge Page Management
- Reducing Minor Page Fault Overheads through Enhanced Page Walker
- Prebaking Functions to Warm the Serverless Cold Start
- Benchmarking, Analysis, and Optimization of Serverless Function Snapshots
- Replayable Execution Optimized for Page Sharing for a Managed Runtime Environment
- Serverless in the Wild: Characterizing and Optimizing the Serverless Workload at a Large Cloud Provider
- SEUSS: Skip Redundant Paths to Make Serverless Fast
- Shared Address Translation Revisited
- SOCK: Rapid Task Provisioning with Serverless-Optimized Containers
- Tile Size Selection Using Cache Organization and Data Layout
- Extending Applications Safely and Efficiently
- Criticality-Aware Instruction-Centric Bandwidth Partitioning for Data Center Applications
- Disentangling the Dual Role of NIC Receive Rings
- Oasis: Pooling PCIe Devices Over CXL to Boost Utilization
- Spirit: Fair Allocation of Interdependent Resources in Remote Memory Systems
- Robust LLM Training Infrastructure at ByteDance
- Sailor: Automating Distributed Training over Dynamic, Heterogeneous, and Geo-distributed Clusters
- Kinetic Modeling of Data Eviction in Cache
- Mitigating Application Resource Overload with Targeted Task Cancellation
- COpter: Efficient Large-Scale Resource-Allocation via Continual Optimization
- IC-Cache: Efficient Large Language Model Serving via In-context Caching
- PrefillOnly: An Inference Engine for Prefill-only Workloads in Large Language Model Applications
- Jenga: Effective Memory Management for Serving LLM with Heterogeneity
- A large scale analysis of hundreds of in-memory cache clusters at Twitter
- cache_ext: Customizing the Page Cache with eBPF
- Sleeping with One Eye Open:Fast, Sustainable Storage with Sandman
- Tiga: Accelerating Geo-Distributed Transactions with Synchronized Clocks
- SAND: A New Programming Abstraction for Video-based Deep Learning
#OSDI
- Caladan: Mitigating Interference at Microsecond Timescales
- Coordinated and Efficient Huge Page Management with Ingens
- Extending Applications Safely and Efficiently
- Disentangling the Dual Role of NIC Receive Rings
- A large scale analysis of hundreds of in-memory cache clusters at Twitter
#2020
- Caladan: Mitigating Interference at Microsecond Timescales
- SEUSS: Skip Redundant Paths to Make Serverless Fast
- A large scale analysis of hundreds of in-memory cache clusters at Twitter
#UCC
#MICRO
#2019
- Architectural Implications of Function-as-a-Service Computing
- MEGA: Overcoming Traditional Problems with OS Huge Page Management
#ATC
- FaaSNet: Scalable and Fast Provisioning of Custom Serverless Container Runtimes at Alibaba Cloud Function Compute
- Serverless in the Wild: Characterizing and Optimizing the Serverless Workload at a Large Cloud Provider
- SOCK: Rapid Task Provisioning with Serverless-Optimized Containers
- Kinetic Modeling of Data Eviction in Cache
#2021
#EUROSYS
- Parallelizing Packet Processing in Container Overlay Networks
- Replayable Execution Optimized for Page Sharing for a Managed Runtime Environment
- SEUSS: Skip Redundant Paths to Make Serverless Fast
- Shared Address Translation Revisited
#PLOS
#SAC
#2016
- Coordinated and Efficient Huge Page Management with Ingens
- Shared Address Translation Revisited
- Kinetic Modeling of Data Eviction in Cache
#SYSTOR
#Journal
#Middleware
#ASPLOS
#2022
#2018
#PLDI
#2025
- Extending Applications Safely and Efficiently
- Criticality-Aware Instruction-Centric Bandwidth Partitioning for Data Center Applications
- Disentangling the Dual Role of NIC Receive Rings
- Oasis: Pooling PCIe Devices Over CXL to Boost Utilization
- Spirit: Fair Allocation of Interdependent Resources in Remote Memory Systems
- Robust LLM Training Infrastructure at ByteDance
- Sailor: Automating Distributed Training over Dynamic, Heterogeneous, and Geo-distributed Clusters
- Mitigating Application Resource Overload with Targeted Task Cancellation
- COpter: Efficient Large-Scale Resource-Allocation via Continual Optimization
- IC-Cache: Efficient Large Language Model Serving via In-context Caching
- PrefillOnly: An Inference Engine for Prefill-only Workloads in Large Language Model Applications
- Jenga: Effective Memory Management for Serving LLM with Heterogeneity
- cache_ext: Customizing the Page Cache with eBPF
- Sleeping with One Eye Open:Fast, Sustainable Storage with Sandman
- Tiga: Accelerating Geo-Distributed Transactions with Synchronized Clocks
- SAND: A New Programming Abstraction for Video-based Deep Learning
#HPCA
#C++
#LLM
#NPU
#open-source
#SOSP
- Oasis: Pooling PCIe Devices Over CXL to Boost Utilization
- Spirit: Fair Allocation of Interdependent Resources in Remote Memory Systems
- Robust LLM Training Infrastructure at ByteDance
- Sailor: Automating Distributed Training over Dynamic, Heterogeneous, and Geo-distributed Clusters
- Mitigating Application Resource Overload with Targeted Task Cancellation
- COpter: Efficient Large-Scale Resource-Allocation via Continual Optimization
- IC-Cache: Efficient Large Language Model Serving via In-context Caching
- PrefillOnly: An Inference Engine for Prefill-only Workloads in Large Language Model Applications
- Jenga: Effective Memory Management for Serving LLM with Heterogeneity
- cache_ext: Customizing the Page Cache with eBPF
- Sleeping with One Eye Open:Fast, Sustainable Storage with Sandman
- Tiga: Accelerating Geo-Distributed Transactions with Synchronized Clocks
- SAND: A New Programming Abstraction for Video-based Deep Learning
#Python
#Head-of-Line-Blocking
#linux
#kernel
#systems
#caching
#off-loading
#remove-unnecessary
#add-granularity
#survey
#flexlibility
#eBPF