At Rutgers Systems Research Lab (RSRL), our goal is to build scalable, efficient, secure systems for handling storage and memory heterogeneity. Given the multi-faceted dierences among these technologies, a critical step toward maximizing their impact is to shield applications and upper-level software stacks from the consequent complexity. We work on Operating Systems (OS) for large-scale datacenter and mobile systems. and their implications on Computer Architecture, Distributed Systems, and High-performance Computing systems.


Research Projects

Storage Offloading and Scalability - We are exploring how to provide direct access to storage access without compromising performance, crash-consistency, integrity, and security of the data. [Related Paper]
Memory Heterogeneity - Compute heterogeneity (CPU, GPU, TPU, FPGA) as well as memory heterogeneity (DRAM, NVM, Stacked 3D Memory) is increasing to support Machine Learning and other data-intensive applications. We are exploring how to design and build OS for modern heterogeneous systems. [Related Paper]
Theorectical Innovations for Architecture Redesgin We are exploring how to rethink OS and runtimes to adapt to modern theoretical innovations, which includes data structures.
Distributed Datacenter Heterogeneity - We are exploring the impact of resource heterogeneity for large scale distributed systems.


Team

Yujie Ren - PhD Student, Projects: File System Scalability, Virtual Memory
David Domingo - PhD Student (Pre-Quals) - File System Correctness
Shaleen Garg (with Manish Parashar) - PhD Student (Pre-Quals) - HPC Systems
Jian Zheng
Sudarsun Kannan - Faculty

Alumini

Avin Abraham - Undergraduate - Memory Heterogeneity
Jae Woo Joo - MS (Now at Amazon)
Kyle Straton - Undergraduate


Recent Publications

The Need for Precise and Efficient Memory Capacity Budgeting
Shaleen Garg, Sudarsun Kannan, Manish Parashar
MEMSYS 2020 (To appear)

CompoundFS: Compounding I/O Operations in Firmware File Systems
Yujie Ren, Jian Zhang, Sudarsun Kannan
USENIX HotStorage 2020 [Paper] [Slides] [Video]

An Integrated Micro-Metrics Monitoring Framework for Tackling Distributed Heterogeneity
Babar Khalid, Nolan Rudolph, Ramakrishnan Durairajan, Sudarsun Kannan
USENIX HotStorage 2020 [Paper] [Slides] [Video]

Durable Transactional Memory Can Scale with TimeStone
R.Madhava Krishnan, Jaeho Kim, Ajit Mathew, Xinwei Fu, Anthony Demeri, Changwoo Min, Sudarsun Kannan
NVMW 2020

Accelerating Filesystem Checking and Repair with pFSCK
David Domingo, Kyle Stratton, Sudarsun Kannan
USENIX VAULT 2020 (Linux Storage and Filesystems Conference)

Read as Needed: Building WiSER, a Flash-Optimized Search Engine
Jun He, Kan Wu, Sudarsun Kannan, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau
USENIX FAST 2020 [Paper] [Code] [Slides] [Video]

Durable Transactional Memory Can Scale with TimeStone
R.Madhava Krishnan, Jaeho Kim, Ajit Mathew, Anthony Demeri, Xinwei Fu, Changwoo Min, Sudarsun Kannan
ASPLOS 2020 [PAPER]

Can We Containerize Internet Measurements?
Christopher Misa, Sudarsun Kannan, Ramakrishnan Durairajan
Applied Networking Research Workshop (ANRW'19)

File Systems as Processes.
Jing Liu, Andrea Arpaci-Dusseau, Remzi Arpaci-Dusseau, Sudarsun Kannan
HotStorage 2019

HeteroOS: OS Design for Heterogeneous Memory Management in Datacenters.
Sudarsun Kannan, Ada Gavrilovska, Vishal Gupta, Karsten Schwan
ACM SIGOPS Operating Systems Review - Special Topics, 2019 [PAPER]

Redesigning LSMs for Nonvolatile Memory with NoveLSM.
Sudarsun Kannan, Nitish Bhat, Ada Gavrilovska, Andrea Arpaci-Dusseau, Remzi Arpaci-Dusseau
USENIX ATC 2018

Research Support

NSFLogo