Mitigating egregious ACK delays in cellular data networks by eliminating TCP ACK clocking

It is not uncommon for the uplink buffers of cellular data networks to be saturated when the uplink bandwidths are low. This can cause the ACK packets for a downlink TCP flow to be severely delayed. Since existing TCP implementations are ACK-clocked, the downstream flow will suffer significant degradation, causing the downlink to be under-utilized. We present a new TCP variant, called TCP Receiver-Rate Estimation (TCP-RRE), that addresses this problem directly by eliminating ACK clocking. Instead, it uses TCP timestamps to estimate the receiving rate at the receiver, which it then uses to determine the sending rate. We show that TCP-RRE is able to improve download speeds by 2 to 4 times compared to existing TCP variants in both simulation and on real commercial cellular data networks. Our solution is practical because it is compatible with existing TCP implementations, requires no modifications to existing mobile devices, and is thus immediately deployable in existing ISP proxies.

[1]  Andrzej Duda,et al.  Two-way TCP connections: old problem, new insight , 2011, CCRV.

[2]  Carey L. Williamson,et al.  Towards a rate-based TCP protocol for the Web , 2000, Proceedings 8th International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems (Cat. No.PR00728).

[3]  Lili Wang,et al.  A Conservative Selective Acknowledgment (SACK)-based Loss Recovery Algorithm for TCP , 2003, RFC.

[4]  Paramvir Bahl,et al.  Anatomizing application performance differences on smartphones , 2010, MobiSys '10.

[5]  Ren Wang,et al.  TCP westwood: Bandwidth estimation for enhanced transport over wireless links , 2001, MobiCom '01.

[6]  Yin Xu,et al.  Dynamic regulation of mobile 3G/HSPA uplink buffer with Receiver-side Flow Control , 2012, 2012 20th IEEE International Conference on Network Protocols (ICNP).

[7]  Vikram Visweswaraiah,et al.  Rate Based Pacing for TCP , 1997 .

[8]  V. Jacobson,et al.  Congestion avoidance and control , 1988, CCRV.

[9]  K. K. Ramakrishnan,et al.  Improving TCP throughput over two-way asymmetric links: analysis and solutions , 1998, SIGMETRICS '98/PERFORMANCE '98.

[10]  Van Jacobson,et al.  Controlling Queue Delay , 2012, ACM Queue.

[11]  Raj Jain,et al.  A Quantitative Measure Of Fairness And Discrimination For Resource Allocation In Shared Computer Systems , 1998, ArXiv.

[12]  Srinivasan Keshav.  A control-theoretic approach to flow control , 1991, SIGCOMM 1991.

[13]  Anurag Kumar,et al.  Performance of TCP congestion control with explicit rate feedback: rate adaptive TCP (RATCP) , 2000, Globecom '00 - IEEE. Global Telecommunications Conference. Conference Record (Cat. No.00CH37137).

[14]  Mark Handley,et al.  Equation-based congestion control for unicast applications , 2000, SIGCOMM.

[15]  Monia Ghobadi,et al.  Proportional rate reduction for TCP , 2011, IMC '11.

[16]  Amit Agarwal,et al.  An argument for increasing TCP's initial congestion window , 2010, CCRV.

[17]  Larry Peterson,et al.  TCP Vegas: new techniques for congestion detection and avoidance , 1994, SIGCOMM 1994.

[18]  Injong Rhee,et al.  Tackling bufferbloat in 3G/4G networks , 2012, Internet Measurement Conference.

[19]  Guido Appenzeller,et al.  Sizing router buffers , 2004, SIGCOMM '04.

[20]  Injong Rhee,et al.  CUBIC: a new TCP-friendly high-speed TCP variant , 2008, OPSR.