PathKatana: Accurate Available Bandwidth Estimation for High-Speed Networks Using Packet Train Consisting Only of Large-Sized Packets

This paper proposes PathKatana, a method for accurately and stably estimating the available bandwidth of high-speed networks using packet train consisting of large-sized packets close to the MTU (Maximum Transmission Unit). Active end-to-end available bandwidth estimation is often intrusive, costly, and inaccurate. However, it is extremely important for designing high-performance network systems, improving network protocols, building distributed systems, and enhancing application performance. PathKatana follows the principle of PathRefiner estimation and calculates estimated values by transmitting and receiving packets near the size of MTU, using a modified theoretical curve. By calculating the absolute delay based on the transmission and reception data of reference packets and fitting it with the modified PathQuick3 theoretical curve, PathKatana estimates the available bandwidth. When implemented on an actual device and evaluated on a 1 Gbps high-speed line, PathKatana demonstrated superior estimation accuracy compared to conventional methods. It was confirmed that PathKatana can estimate the available bandwidth of high-speed lines with high accuracy.