The unfairness of UDP traffic in routers with different buffer units
- PDF / 2,880,628 Bytes
- 13 Pages / 595.276 x 790.866 pts Page_size
- 27 Downloads / 152 Views
ORIGINAL RESEARCH
The unfairness of UDP traffic in routers with different buffer units Bo Zhang1 · Qian Li1 · Jinyao Yan2 Received: 21 June 2018 / Accepted: 31 August 2018 © Springer-Verlag GmbH Germany, part of Springer Nature 2018
Abstract In recent years, router buffer sizing becomes one of hot research topics. There were a variety of different views of buffer sizing presented in the literature. However, all the previous studies ignored whether buffer unit was structured in terms of packet or byte. In this paper, we find that the loss performance of routers with very small buffers is impacted by the buffer unit structure. When buffer unit is structured in terms of byte, the UDP packet loss rate is lower than the TCP packet loss rate. However, when buffer unit is structured in terms of packet, the UDP packet loss rate is higher than the TCP packet loss rate. We draw the conclusion that the different buffer units would create an unfair impact on UDP traffic. We further analyze this unexpected unfairness for UDP traffic, and present two models to explain the unfairness created by different buffer units in routers. The simulation results reveal that a smaller proportion of UDP traffic ensures low packet loss rate for TCP traffic when buffer unit is byte. When buffer unit is packet, a smaller proportion of UDP traffic ensures low packet loss rate for both TCP traffic and UDP traffic. Keywords Router buffer sizing · Very small buffers · Buffet unit · Unfairness · Models
1 Introduction Internet routers need buffers to absorb the transient bursts that naturally occur in networks and reduce the packet drops. At the same time, the buffers can also keep the output links fully utilized during the period of congestion. However, buffers introduce delay and jitter, and increase the router cost and power dissipation. A basic and important question that should be concerned is how much buffer routers need (Appenzeller et al. 2004; Prasad et al. 2007). Buffer overflow causes packet loss, adversely affecting application performance. On the other hand, an underflow leads to the wastage of link bandwidth and idling, thereby degrades the network throughput. The issue of router buffer sizing has generated a * Bo Zhang [email protected] Qian Li [email protected] Jinyao Yan [email protected] 1
School of Information Engineering, Communication University of China, Beijing 100024, China
Key Laboratory of Media Audio and Video, Communication University of China, Beijing 100024, China
2
lot of debates in the last 20 years (Villamizar and Song 1994; Appenzeller et al. 2004; Dhamdhere et al. 2005; Stanojevic and Shorten 2007; Yfoulis and Xanthopoulos 2010). Nowadays the router buffers are sized based on the ruleof-thumb (Villamizar and Song 1994). The rule claims that in order to make full utilization of the bottleneck link, a router needs a buffer of size B = RTT × C , where C is the capacity of bottleneck link, and RTT denotes the average round-trip time of a TCP flow passing across the link. If the network operators foll
Data Loading...
