A Pragmatic Transport for Datacenters

What is it about?

This paper introduces PPT, a pragmatic transport that achieves comparable performance to proactive transports while maintaining good deployability as reactive transports. Our key idea is to run a low-priority control loop to leverage the available bandwidth left by the reactive transports. The main challenge is to send just enough packets to improve performance without harming the primary control loop. We combine two unconventional techniques: an intermittent loop initialization and an exponential window decrease, enabling us to dynamically identify and fill the spare bandwidth. We further complement PPT's design with a buffer-aware flow scheduling scheme to optimize the average FCT of small flows without prior knowledge of flow size information. We have implemented a PPT prototype in the Linux kernel with about 400 lines of code and demonstrated that compared to Homa, it delivers up to 46.3\% lower overall average FCT and even 25\%/55.5\% lower average/tail FCT of small flows in an Memcached workload.

Featured Image

Photo by imgix on Unsplash

Photo by imgix on Unsplash

Why is it important?

PPT is a pragmatic DCN transport that uses a dual-loop rate control design to gracefully utilize the available bandwidth and further complements its design with a buffer-aware flow scheduling to optimize small flow's performance. We have implemented a PPT prototype using commodity hardware, and evaluated it through small-scale testbed experiments and large-scale simulations. Extensive evaluations show that PPT is a viable solution that achieves our design goals.

Perspectives