205, Arthur K. Watson Hall
51 Prospect St
New Haven, CT 06520

anurag [dot] khandelwal [at] yale [dot] edu


I am an Assistant Professor at the Department of Computer Science at Yale University. My research interests span distributed systems, databases, networking, and algorithms.

Prior to starting at Yale, I did a short post-doc at Cornell Tech where I worked with Tom Ristenpart and Rachit Agarwal. I received my PhD from the University of California, Berkeley, at the RISELab, where I was advised by Ion Stoica. I completed my bachelor’s degree (B. Tech. in Computer Science and Engineering) from the Indian Institute of Technology, Kharagpur.

I am looking for highly motivated graduate students. Please send me an email if you are interested.


OS Stack for Serverless Architectures: Cloud services are quickly moving from traditional server-based architecture to a serverless model. Such serverless architectures enable higher scalability and resource utilization by allowing applications to launch short-lived compute tasks that operate on data stored on a remote store. However, today’s serverless stacks cater mainly to stateless (i.e., embarassingly parallel) tasks, there is a tremendous push towards supporting stateful applications on serverless architectures. To this end, we are exploring the ground up design of the serverless OS stack that facilitates stateless and stateful applications.


System stack for emerging hardware: Today’s system stacks were designed to operate with traditional hardware, e.g., with 1 Gbps links and traditional storage media. The next generation of emerging hardware (e.g., 100Gbps links, non-volatile memory) change the many fundamental assumptions made in the design and optimization of these systems. To resolve these challenges, we are revisiting traditional system designs to bridge the gap between hardware capabilities and realizable system properties.


Secure cloud systems: With web applications and services moving from self-owned servers in private data centers to to public cloud platforms, users must now trust the cloud provider who manages the physical infrastructure that their applications run on. Unfortunately, high-profile security breaches in the public cloud indicate that this trust may not always be well placed. We are exploring the vulnerabilities of existing system deployments hosted on the cloud and the design of secure systems that no longer have to trust the cloud provider.

Queries on compressed data: Ensuring low latency and high throughput for user-facing queries is challenging when the volume of data being queried grows larger than the DRAM capacity. Traditionally, storage systems have resorted to spilling over such data to significantly slower secondary storage, resulting in higher query latency and reduced throughput. We have been exploring a fundamentally new approach to resolve this challenge — enabling queries directly on a compressed representation of the data.



CPSC 433/533: Computer Networks