In this talk, we will cover the topics of load balancing in a bottom-up fashion. We will start from the relatively new problem of dispatching the exponentially growing number of packets received by Network Interface Controllers to an increasingly large number of CPU cores.
Our new approach called RSS++ aims to answer the key question in this domain: which CPU core should get an incoming packet? RSS++ achieves very good load balancing over multiple CPU cores by exploiting opportunistic and controlled flow migration (utilizing a new design that enables lockless and zero-copy migration of state between CPU cores).
We will then study how to implement load-balancers that will be able to dispatch packets to thousands of servers with hundred-gigabit connectivity. The challenge is to maintain uniform load distribution of the incoming connections across the servers while maintaining per-connection-consistency (PCC), i.e., the ability to map packets belonging to the same connection to the same server even in the presence of changes in the number of active servers and load balancers. We will detail the design of a new load balancer that supports uniform load distribution and PCC while being scalable, memory efficient, resilient to clogging attacks, and fast at processing packets.
Tom Barbette, is a Post-Doc in Networked Systems at KTH, Sweden. I defended my PhD thesis in 2018 at the Université de Liège in Belgium. His main topics of interest are high-speed networking, network function virtualization, load-balancing, operating systems and high-performance programming.