Juncheng Yang

I am an Assistant Professor in School of Applied Science and Engineering, Harvard University.

I am broadly interested in storage systems, data management and machine learning systems with particular interests on workload analysis, efficient storage, and sustainable system design. I like in-depth measurement and analysis to get deep understanding of systems and algorithms in the real world.

My works have received best-paper awards at NSDI'24, NSDI'21, SOSP'21, and SYSTOR'16 and have been deployed in production at Google, VMware, Twitter, Redpanda, Momento with many open-source libraries contributed by the community. My research has been sponsored by Meta, Google Cloud, and AWS. I am a 2020 Meta Fellow, a 2023 Google Cloud Research Innovator, and a 2023 Rising Star in Machine Learning and Systems.

---

---

News [all news]

• 04/2024 - SIEVE received a community award at NSDI'24, was featured on TLDR newletter and discussed in a a blog post by Marc. Many open-source libraries (in more than 12 languges) are available on GitHub.
• 12/2023 - S3-FIFO was discussed in Aleksey's Online Reading Group, it has been covered by blogs in English, Chinese, Korean, and Japanese, used in course reading materials at UIUC CS525, and implemented at Google, VMware, Redpanda and in many open-source libraries and systems.
• 08/2023 - Recognized as a Rising Star in machine learning and systems.
• 08/2023 - 10/2023 Invited to talk about S3-FIFO at Cloudflare, Alluxio, VMware, Kuaishou, MSRA, USTC, Tsinghua University, WOS conference.
• 05/2023 - Honored to be a Google Cloud Research Innovator with funding from Google Cloud.
• 10/2022 - 02/2023 - Invited to talk about LESSCache at VMware, Meta and Oracle.
• 10/2022 - Invited speaker to talk about ubiquitous caching at QCon SF .
• 09/2020 - Gave a keynote talk together with Yao Yue at SNIA SDC on using PMEM for caching

---

Research Areas and Interests

Storage systems and machine learning systems with a focus on efficiency, scalability and robustness:

• Efficient and scalable cache management systems
• Measurement studies [OSDI'20, TOS'21][HotOS'23]
• New cache system designs [NSDI'21][SOSP'21, TOS'22][EuroSys'23]
• New replacement algorithm designs [SOSP'23][NSDI'24]
• Robust and reliable cache/storage management and machine learning systems [OSDI'20][NSDI'22][VLDB'23]
• New approaches to make machine learning practical for storage systems (machine learning for systems) [FAST'23][SOCC'17]
• Performance optimization and sustainability of microservices and serverless architecture [SOCC'23]
• Reliable large model inference on wimpy hardware (system for machine learning)

---

Research Highlights

• SIEVE (NSDI'24): the first cache eviction algorithm simpler than LRU but yet more effective than state-of-the-art algorithms for web caches. Implemented in many open-source libraries, e.g., Golang, Python, JavaScript, Rust, Java, Swift, Ruby, Nim, and Zig. Find more details here.
• S3-FIFO (SOSP'23): a simple and scalable cache eviction algorithm composed of only FIFO queues. Implemented or deployed at companies including Google, VMware and Redpanda, and many open-source libraries. Find more details here.
• GL-Cache (FAST'23): my exploration on bringing low-overhead machine learning to caching.
• Segcache (NSDI'21) received a best-paper award, and deployed at Twitter and Momento.
• Kangaroo (SOSP'21) received a best-paper award at SOSP'21.

---

I have been very fortunate to work with many talented students. If you have worked with me, but not showing up on this page, please feel free to let me know.

---

Selected Publications [Google Scholar]

• SIEVE: Cache eviction can be simple, effective, and scalable.
  Juncheng Yang, Yazhuo Zhang, Yao Yue, Ymir Vigfusson, K. V. Rashmi.
  USENIX ;login: , 2024
  

• SIEVE is Simpler than LRU: an Efficient Turn-Key Eviction Algorithm for Web Caches. [website] [blog] [pdf] [slides]
  Yazhuo Zhang*, Juncheng Yang* (corresponding author), Yao Yue, Ymir Vigfusson, K. V. Rashmi.
  The 20th USENIX Symposium on Networked Systems Design and Implementation (NSDI), 2024
  SIEVE receives the community (best paper) award
  Featured on TLDR newletter, blog coverage: Marc from AWS, blog post in Korean.
  Independent implementations and evaluation: golang-fifo, use in DNSCrypt.
  Open-source libraries (not an extensive list): Golang, Python, JavaScript, Rust, Java, Swift, Ruby, Nim, Zig.

• FIFO queues are all you need for cache eviction. [website] [blog] [pdf] [slides] [video]
  Juncheng Yang, Yazhuo Zhang, Ziyue Qiu, Yao Yue, K. V. Rashmi.
  The 29th ACM Symposium on Operating Systems Principles (SOSP), 2023
  Will be discussed in Aleksey's Online Reading Group. Covered in blog [1], [2], [3] in Korean, [4] in Chinese, [5] in Japanese, newsletters [1], [2] .
  Open-source libraries (not an extensive list): Rust, Golang, JavaScript, Python, C++.

• FIFO Can be Better than LRU: the Power of Lazy Promotion and Quick Demotion. [pdf] [slides]
  Juncheng Yang, Ziyue Qiu, Yazhuo Zhang, Yao Yue, K. V. Rashmi.
  The 19th Workshop on Hot Topics in Operating Systems (HotOS), 2023

• GL-Cache: Group-level learning for efficient and high-performance caching. [pdf] [slides]
  Juncheng Yang, Ziming Mao, Yao Yue, K. V. Rashmi.
  The 21st USENIX Conference on File and Storage Technologies (FAST), 2023

• FrozenHot Cache: Rethinking Cache Management for Modern Hardware. [pdf] [slides]
  Ziyue Qiu, Juncheng Yang, Juncheng Zhang, Cheng Li, Xiaosong Ma, Qi Chen, Mao Yang, Yinlong Xu.
  The European Conference on Computer Systems (EuroSys), 2023

• Efficient Fault Tolerance for Recommendation Model Training via Erasure Coding. [pdf]
  Tianyu Zhang, Kaige Liu, Jack Kosaian, Juncheng Yang, K. V. Rashmi.
  49th International Conference on Very Large Databases (VLDB), 2023

• Latenseer: Causal Modeling of End-to-End Latency Distributions by Harnessing Distributed Tracing. [pdf]
  Yazhuo Zhang, Rebecca Isaacs, Yao Yue, Juncheng Yang, Lei Zhang, Ymir Vigfusson.
  ACM Symposium on Cloud Computing (SoCC), 2023

• C2DN: How to Harness Erasure Codes at the Edge for Efficient Content Delivery. [pdf] [slides]
  Juncheng Yang, Anirudh Sabnis, Daniel S. Berger, K. V. Rashmi, Ramesh K. Sitaraman
  The 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI), 2022

• Segcache: memory-efficient and high-throughput DRAM cache for small objects. [pdf] [slides]
  Juncheng Yang, Yao Yue, K. V. Rashmi.
  The 18th USENIX Symposium on Networked Systems Design and Implementation (NSDI), 2021
  Segcache receives the community award (one of the best papers)
  This work has been adopted for production at Twitter. See a short summary of the work.
  Open-source systems and libraries: Pelikan, Rust crate,

• Kangaroo: Caching Billions of Tiny Objects on Flash. [pdf] [slides]
  Sara McAllister, Benjamin Berg, Julian Tutuncu-Macias, Juncheng Yang, Sathya Gunasekar, Jimmy Lu, Nathan Beckmann, Gregory R. Ganger.
  28th ACM Symposium on Operating Systems Principles (SOSP), 2021
  Extended version (invited submission) - ACM Transaction on storage (TOS) 2022
  Kangaroo receives the best paper award

• A Large Scale Analysis of Hundreds of In-memory Cache Clusters at Twitter. [pdf] [slides]
  Juncheng Yang, Yao Yue, K. V. Rashmi.
  The 14th USENIX Symposium on Operating Systems Design and Implementation (OSDI), 2020
  Recognized as one of the best storage papers and invited for submission to ACM Transactions on Storage.
  Discussed in Aleksey's Online Reading Group.

• PACEMAKER: Avoiding HeART Attacks in Storage Clusters with Disk-adaptive Redundancy. [pdf][slides]
  Saurabh Kadekodi, Francisco Maturana, Suhas Jayaram Subramanya, Juncheng Yang, K. V. Rashmi, Gregory R. Ganger.
  14th USENIX Symposium on Operating Systems Design and Implementation (OSDI), 2020

• Mutant: Balancing Storage Cost and Latency in LSM-Tree Data Stores. [pdf]
  Hobin Yoon, Juncheng Yang, Sveinn F. Kristjansson, Steinn E. Sigurdarson, Ymir Vigfusson, Ada Gavrilovska.
  ACM Symposium on Cloud Computing (SoCC), 2018

• Skyline Diagram: Finding the Voronoi Counterpart for kyline Queries [pdf]
  Jinfei Liu, Juncheng Yang, Li Xiong, Jian Pei, Jun Luo.
  IEEE International Conference on Data Engineering (ICDE), 2018.
  Extended version - IEEE Transactions on Knowledge and Data Engineering (TKDE), 2019

• MITHRIL: Mining Sporadic Associations for Cache Prefetching. [pdf]
  Juncheng Yang, Reza Karimi, Trausti Saemundsson, Avani Wildani, Ymir Vigfusson.
  ACM Symposium on Cloud Computing (SoCC), 2017

• Secure Skyline Queries on Cloud Platform [pdf]
  Jinfei Liu, Juncheng Yang, Li Xiong, Jian Pei.
  IEEE International Conference on Data Engineering (ICDE), 2017.
  Extended version - IEEE Transactions on Knowledge and Data Engineering (TKDE), 2018.

• Enabling Space Elasticity in Storage Systems. [pdf]
  Helgi Sigurbjarnarson, Petur Orri Ragnarsson, Juncheng Yang, Ymir Vigfusson, Mahesh Balakrishnan.
  ACM International Systems and Storage Conference (SYSTOR), 2016
  Best student paper

---

Teaching

---

Services

Conference

• FAST Artifact Evaluation Chair: 2025, 2026
• FAST Reivewer: 2025, 2026
• MLSys Reviewer: 2025
• ICDCS Reviewer: 2025

Workshop

• New England Systems Day: 2025
• HotStorage Reviewer: 2025

Journal review

• ACM Transactions on Storage (TOS): 2023
• ACM Computing Survey: 2025
• Springer Journal of Supercomputing: 2024
• IEEE Transaction on Computers (TOC): 2023, 2024
• IEEE Transaction on Reliability (TOR): 2024
• IEEE Transactions on Cloud Computing (TCC): 2023
• IEEE ACCESS: 2023
• IEEE Transactions on Parallel and Distributed Systems (TPDS): 2023, 2024
• IEEE Transactions on Mobile Computing (TMC): 2023
• IEEE Transactions on Knowledge and Data Engineering (TKDE): 2023

---

Awards

• 2023 Machine learning and System Rising Star
• 2023 Google Cloud Innovator
• 2020-22, Facebook Fellowship
• 2018, AWS research grant
• 2013, Emerson Fellowship
• 2013, Best Thesis Award (5/3000)
• 2012, "Person of the Year" Nomination
• 2012, Third Place in Green Tech International Competition, Taiwan
• 2009, First Award in National Chemistry Olympiad

---

Talks

• Simple Scalable Caching with Three Static FIFO Queues.
  • [11/2023] WOS conference
  • [09/2023] VMware
  • [09/2023] Kuaishou
  • [09/2023] Microsoft Research Asia
  • [09/2023] Tsinghua University
  • [08/2023] Cloudflare
  • [08/2023] USTC
  • [07/2023] Alluxio
• LESSCache: LEarned Segment-Structured Cache
  • [02/2023] Meta
  • [10/2022] VMware vSAN and VMware Research
• Ubiquitous Caching: A Journey of Building Efficient Distributed and Process Caches.
  • [10/2022] QCon SF
  • [08/2022] Alluxio
• Segcache: a memory-efficient and scalable in-memory key-value cache for small objects
  • [01/2023] Oracle
  • [2022] Shopify

---

Open-source Software & Data

• Cache trace datasets
• libCacheSim: a high-performance cache library and simulator
• fastscp: a tool for transfering files with two nodes
• distComp: a tool for running computation jobs in parallel on many nodes

---

Bio

Juncheng Yang is an Assistant Professor in the School of Engineering and Applied Science at Harvard University. He received his Ph.D. in Computer Science from Carnegie Mellon University in 2024, His research interests broadly cover the efficiency, performance, reliability, and sustainability of large-scale data systems.

Juncheng's works have received best paper awards at NSDI'24, NSDI'21, SOSP'21, and SYSTOR'16. His OSDI'20 paper was recognized as one of the best storage papers at the conference and invited to ACM TOS'21. Juncheng received a Facebook Ph.D. Fellowship in 2020, was recognized as a Rising Star in machine learning and systems in 2023, and a Google Cloud Research Innovator in 2023.

His work, Segcache, has been adopted for production at Twitter and Momento. The two eviction algorithms he designed (S3-FIFO, SIEVE) have been adopted for production at Google, VMware, Redpanda, and several others, with over 20 open-source libraries available on GitHub. Moreover, the open-source cache simulation library he created, libCacheSim, has been used by almost 100 research institutes and companies.

---