Redesigning networking stacks for scalable mobile data communication

Abstract

Rapid development of today’s mobile applications drives the demand for high-bandwidth, low-latency mobile data communication. Unfortunately, today’s network architecture poses several critical challenges to achieve scalable mobile data communication. On the client side with mobile devices, it is unclear whether the cellular network alone can meet the future demand of mobile data communication, due to the limited cellular network coverage and bandwidth. On the other hand, on the server side, CPU resources become a key limiting factor that determines the performance of network applications, but CPU cycles are often wasted for repetitive, mechanical operations merely to meet protocol requirements. In this dissertation, we claim that the networking stacks should be redesigned with new principles to achieve scalability in mobile data communication in two directions. First, we propose Cedos, a networking stack that can easily support offloading mobile data traffic to Wi-Fi, in order to overcome the limitation on cellular network bandwidth. Cedos allows users to offload as much mobile data traffic to Wi-Fi as possible, taking advantage of the user's latency tolerance, while transparently masking network disruptions and delays from applications. We find that up to 92.4% of the podcast traffic is offloaded to Wi-Fi, and one can watch a streaming video in a moving train while offloading 48% of the content to Wi-Fi without a single pause. Second, we propose AccelTCP, which offload complex TCP operations such as connection setup and teardown completely to NIC, which simplifies the host stack operations and frees a significant amount of CPU cycles for application processing. In addition, it supports running connection splicing on NIC so that the NIC relays all packets of the spliced connections with zero DMA overhead. Our evaluation shows that AccelTCP enables short-lived connections to perform comparably to persistent connections. It also improves the performance of Redis, a popular in-memory key-value store, and HAProxy, a widely-used layer-7 load balancer, by 2.3x and 11.9x, respectively.