alt text 

Stanford University Networking Seminar


Mohammad Alizadeh (Cisco Systems)
CONGA: Congestion-Aware Load Balancing for Datacenters

12:15pm, Thursday, April 16 2015
Gates 104

About the Talk

Modern datacenter fabrics must provide immense bandwidth — 10s of Tbps across 1000s of ports — to support demanding applications such as big data analytics and large-scale web services. They achieve this by spreading traffic over many paths across dense, multipath topologies. Yet, state-of-the-art multipath load balancing mechanisms in datacenter networks are inefficient and fragile. In this talk, I will present CONGA, a new in-network load balancing solution that we have developed over the last few years as part of Cisco’s Application Centric Infrastructure (ACI) datacenter fabric products. CONGA dynamically balances load based on real-time congestion measurements within the data-path, and seamlessly handles topological asymmetries (e.g., due to link failures), all without needing any end-host or TCP modifications. I will discuss CONGA’s key design decisions, and briefly highlight analysis and testbed experiments that demonstrate CONGA’s effectiveness.

About the Speaker

Mohammad Alizadeh is a principal engineer at Cisco, which he joined through the acquisition of networking startup, Insieme Networks, in December 2013. Starting in the fall, he will join the EECS department at MIT as an assistant professor. Mohammad’s research interests are in the areas of computer networks and systems. His recent projects focus on architectures and algorithms for large-scale datacenters, particularly, predictable high performance networks for real-time web and big data applications. His research on datacenter transport has been implemented in commercial (Windows Server 2012) and open source (Linux) operating systems, and major datacenter fabric products (Cisco ACI). Mohammad is a recipient of several awards, including the SIGCOMM 2014 best paper award and the Numerical Technologies Inc. Prize and Fellowship. He received his PhD in Electrical Engineering from Stanford University in 2013.