Operating Systems

Kernel design and implementation

Solution: abstracting the underlying hardware and providing a secure, managed, and efficient environment for user-level applications to run concurrently and share system resources

Process and thread management

Solution: ensuring fair resource allocation, smooth multitasking, and proper coordination between processes and threads to maximize system performance and responsiveness.

GPU Checkpoint and Restore

Challenge: Saving and restoring the complete state for GPUs is difficult without causing long application stalls

Year	Venue	Authors	Title	Tags	P	E	N
2025	SOSP	SJTU	PHOENIXOS: Concurrent OS-level GPU Checkpoint and Restore with Validated Speculation	validated speculation; concurrent GPU checkpoint and restore; soft copy-on-write; GPU context pool	4	4	4

Memory management

Memory management is the basic of system design.

Memory Partitioning & Mapping

Solution: efficiently and securely divide physical memory among multiple processes or systems

Challenge: ensuring processes get enough memory without waste, while preventing interference and maintaining system performance and security.

Year	Venue	Authors	Title	Tags	P	E	N
2012	HPCA	Georgia Institute of Technology	TAP: A TLP-Aware Cache Management Policy for a CPU-GPU Heterogeneous Architecture	thread-level parallelism; core sampling for cache effort indentification; cache block lifetime normalization
2024	arXiv	UMich	Mercury: QoS-Aware Tiered Memory System	contend for local memory; priority inversion; intra- and inter-tier interference; per-tier page reclaimation
2025	HPCA	Seoul National	FACIL: Flexible DRAM Address Mapping for SoC-PIM Cooperative On-device LLM Inference	flexible PIM address mapping and remapping; OS paging mechanism extension for MapID; user-level mapping selector	3	2	3

Page Migration

Challenge: the trade-off between migration overhead and potential performance gains, complicated by prediction accuracy and hardware complexities

Year	Venue	Authors	Title	Tags	P	E	N
2024	SC	THU	Hydrogen: Contention-Aware Hybrid Memory for Heterogeneous CPU-GPU Architectures	fast memory decoupled partitioning; token-based slow memory migration; epoch-based sampling method; consistent hashing based reconfiguration
2024	OSDI	UT Arlington	Nomad: Non-Exclusive Memory Tiering via Transactional Page Migration	Memory allocator for hardware tiering to mitigate outliers
2025	ISCA	XMU	ArtMem: Adaptive Migration in Reinforcement Learning-Enabled Tiered Memory	reinforcement learning for tiered memory management; dynamic adjustment of migration scope	2	4	3

Memory Pooling

Solution: Memory Pooling reduces the overhead of frequent memory allocation and deallocation by maintaining a pool of pre-allocated memory blocks for reuse.

Year	Venue	Authors	Title	Tags	P	E	N
2023	ASPLOS	Virginia Tech	Pond: CXL-Based Memory Pooling Systems for Cloud Platforms	CXL-based memory pooling; small-pool design for low latency; machine learning model for memory allocation prediction; zero-core virtual NUMA (zNUMA) node for untouched memory

Cache Evict

Challenge: to balance the cache hit ratio and the cache miss penalty.

The cache here not only includes the cache in the processors, but also the software concept in systems.

Year	Venue	Authors	Title	Tags	P	E	N
2024	NSDI	CMU	Sieve is Simpler than LRU: an Efficient Turn-Key Eviction Algorithm for Web Caches	web cache management; fifo-like schedule; recognizing the importance of objects	3	3	3

File systems and storage

Solution: organizing, storing, and retrieving data efficiently and reliably on storage devices, manage disk space, and ensure data integrity and security.

Storage Systems

Solution: eficient, reliable, and secure management of data storage across diverse hardware while achieving balance between performance, durability, efficiency, and scalability .

Out-of-Core Processing

Challenge: performance penalty of slow disk I/O

Year	Venue	Authors	Title	Tags	P	E	N
2018	OSDI	UCLA	RStream: Marrying Relational Algebra with Streaming for Efficient Graph Mining on A Single Machine	single-machine out-of-core graph mining; tuple streaming; relational algebra on graphs; GRAS programming model	4	4	4

SSD Management

Challenge: limited write endurance, garbage collection, wear leveling, and the mismatch between erase and write operations, while optimizing performance and extending device lifespan

Year	Venue	Authors	Title	Tags	P	E	N
2025	EuroSys	Samsung	Towards Efficient Flash Caches with Emerging NVMe Flexible Data Placement SSDs	NVMe Flexible Data Placement (FDP) SSDs for data segregation; targeted data placement for reduced device write amplification; FDP-enabled CacheLib architecture; theoretical DLWA model for CacheLib
2025	arXiv	SDU	Managing Hybrid Solid-State Drives Using Large Language Models	LLM-based auto-tuning framework for hybrid SSD management; hybrid SSD parameter categorization; performance-sensitive parameter selection; prompt engineering for LLM integration; dynamic configuration optimization in hybrid SSDs
2025	arXiv	ETHZ	Harmonia: A Multi-Agent Reinforcement Learning Approach to Data Placement and Migration in Hybrid Storage Systems	dual RL agents for data placement & migration; I/O latency-based reward shaping	3	4	4

Virtualization

Solution: efficiently managing and isolating multiple virtual machines (VMs) or containers on a single physical machine, allowing them to share hardware resources securely and independently while maintaining performance and flexibility.

Synchronization

Solution: coordinating access to shared resources among multiple processes or threads to prevent conflicts, ensure data consistency, and avoid issues like race conditions, deadlocks, and resource starvation.

Year	Venue	Authors	Title	Tags	P	E	N
2019	ASPLOS	THU&USC	pLock: A Fast Lock for Architectures with Explicit Inter-core Message Passing	chaining lock for direct lock transfering; hierarchical lock for non-uniform network	4	4	2