Inho's work on building a high-resolution latency debugger that can detect microsecond-level latency events was accepted to NSDI 2024.
Eric presented his poster at SIGCOMM 2023, winning the 1st of the Student Research Competition (SRC) in the undergraduate category.
I am the PI on a $40k award from Spec Tech to improve the scalability of routing and placement algorithms for synthesizing circuits using nanomodular electronics.
Vaibhav's work on estimating round trip times in LEO satellite networks was accepted at the workshop on LEO networking and communication that's held in conjunction with MobiCom 2023.
Eric's work on the interaction between overload controllers and core allocators accepted as a poster at SIGCOMM 2023.
Work on understanding variability in LEO satellite networks invited to IETF 116 by MAPRG.
The paper offers a thorough characterization of latency variability in LEO satellite networks, exploring its fundamental causes and structure and possible remedies.
Our poster on improving the performance and resilience of satellite networks was selected among the top five posters presented by DARPA Risers. DARPA Risers are up-and-coming standouts in their fields, whose research is related to national security and demonstrates the potential to lead to technological surprise — the heart of DARPA's mission.
I am the PI on a $1.19M award from NSF to improve the scalability of edge servers from the hardware to the applications looking at problems in overload control, network stack architecture, and hardware offloads.
I am the co-PI on a $1.19M award from NSF to develop new models and algorithms to efficiently apply formal methods to analytically identify bounds the performance of resource allocation heuristics used to manage network and CPU resources.
Protego is an overload control system that allows applications to maintain their SLOs even in the presence of data-dependent lock contention.
I received $80,000 from Google to support my work on scalable systems at the edge.
CCAC is a new approach to understanding the performance of congestion control algorithms through formal methods.
Mobius allows operators of large fleets of vehicles to navigate the tradeoffs between task fulfillment throughput and fairness to their customers.
In this paper, we provide a thorough analysis of the bottlenecks that arise in the Linux kernel running on a server as we increase the number of TCP connections handled by the server.
Breakwater is an overload control system that allows microsecond-scale RPCs to maintain their Service Level Objectives (SLOs) even when their demand is much larger than available capacity.
We identify a new problem in WAN congestion control when WAN traffic competes for bandwidth with datacenter traffic inside the datacenter network. Annulus solves the problem by allowing WAN traffic to react to congestion as fast as datacenter traffic using explicit and direct congestion notification messages.
Ahmed Saeed is an Assistant Professor in the School of Computer Science at Georgia Tech. He was a Postdoctoral Associate at MIT working with Prof. Mohammad Alizadeh. He completed his PhD in August 2019 at Georgia Tech, where he was advised by Prof. Mostafa Ammar and Prof. Ellen Zegura. During his PhD, he interned several times at Google, where he collaborated with Nandita Dukkipati and Amin Vahdat. He received his bachelor's degree in Computer and Systems Engineering from Alexandria University in 2010.
Theory, design, and implementation of scalable computer networks and computer systems, including resource scheduling, congestion control, operating systems, and formal methods.
I am looking for highly motivated PhD, master's, and undergraduate students. Reach out to me if you are interested in building large scale systems and/or using formal methods to understand them. For more information, check out this page.
Active projects (more details here):
The goal of the project is develop end-host networking stacks that can scale, not only in terms of their messages per second capacity, but in terms of the number of network connections they can handle. To that end, we tackle problems in schedulers, the VM-hypervisor API, and the transport layer.
Wide area networks (and the Internet at large) are getting more heterogeneous and programmable, proving to be a challenging environment to manage while providing many new knobs to provide better performance. To that end, we explore new tools to better characterize the challenges and new algorithms to address them.
This project attempts to provide concrete analytical tools to understand the performance of heuristics used in resource allocation and management of distributed system. The project leverages progress made in formal verification tools (e.g., Z3) that can efficiently search through all potential scenarios that can encounter a heuristic. We use such tools to identify scenarios where systems underperform, helping system designers avoid them or plan for them.
LDB: An Efficient Latency Debugging Tool for Datacenter Applications
Inho Cho, Seo Jin Park, Ahmed Saeed, Mohammad Alizadeh, Adam Belay
USENIX NSDI 2024 [PDF]
Protego: Overload Control for Applications with Unpredictable Lock Contention
Inho Cho, Ahmed Saeed, Seo Jin Park, Mohammad Alizadeh, Adam Belay
USENIX NSDI 2023 [PDF]
A Characterization of Route Variability in LEO Satellite Networks
Vaibhav Bhosale, Ahmed Saeed, Ketan Bhardwaj, Ada Gavrilovska
PAM 2023 [PDF]
Annulus: A Dual Congestion Control Loop for Datacenter and WAN Traffic Aggregates
Ahmed Saeed, Varun Gupta, Prateesh Goyal, Milad Sharif, Rong Pan, Mostafa Ammar, Ellen Zegura, Keon Jang, Mohammad Alizadeh, Abdul Kabbani, and Amin Vahdat
ACM SIGCOMM 2020 [PDF] [Slides] [Video]
Eiffel: Efficient and Flexible Software Packet Scheduling
Ahmed Saeed, Yimeng Zhao, Nandita Dukkipati, Ellen Zegura, Mostafa Ammar, Khaled Harras, and Amin Vahdat
USENIX NSDI 2019 [PDF] [Extended Version] [Slides] [Video] [Code]