Queue length feedback-based solution of TCP Incast in data center networks

The Internet offers a large number of applications and services that we use on a daily basis. These widelyused applications are hosted on large-scale, high-performance computing systems called data centers. The performance ofTCP is inefficient in many-to-one communication, which is a common traffic pattern in data center networks. This manyto-one communication causes significant packet losses followed by timeouts, which consequently results in throughputcollapse in data center networks; this problem is known as TCP Incast. In this paper, we present a queue length feedbackbased solution to mitigate TCP Incast. The scheme has two parts: i) a novel queue length-based congestion parameter,which accurately measures congestion along the path from source to destination, and ii) a congestion control schemethat effectively uses the new congestion parameter to prevent throughput collapse due to Incast traffic patterns. Resultsare compared with TCP and DCTCP, the two most common transport protocols deployed in data center networks. Theresults show that the proposed scheme minimizes packet drops and achieves high utilization and burst tolerance.

PDF

___

[1] Al-Fares M, Loukissas A, Vahdat A. A scalable, commodity data center network architecture. ACM Sigcomm Comp Com 2008; 38: 63-74.
[2] Greenberg A, Hamilton J, Maltz DA, Patel P. The cost of a cloud: research problems in data center networks. ACM Sigcomm Comp Com 2008; 39: 68-73.
[3] Guo C, Wu H, Tan K, Shi L, Zhang Y. Dcell: a scalable and fault-tolerant network structure for data centers. ACM Sigcomm Comp Com 2008; 38: 75-86.
[4] Abts D, Felderman B. A guided tour of data-center networking. Commun ACM 2012; 55: 44-51.
[5] Zhang Y, Ansari N. On architecture design congestion notification TCP Incast and power consumption in data centers. IEEE Commun Surv Tut 2013; 15: 39-64.
[6] Zhang J, Ren F, Lin C. Survey on transport control in data center networks. IEEE Network 2013; 27: 22-26.
[7] Ousterhout J, Agrawal P, Erickson D, Kozyrakis C, Leverich J, Mazières D, Mitra S, Narayanan A, Ongaro D, Parulkar G et al. The case for RAMCloud. Commun ACM 2011; 54: 121-130.
[8] Bilal K, Khan SU, Zhang L, Li H, Hayat K, Madani SA, Min-Allah N, Wang L, Chen D, Iqbal M et al. Quantitative comparisons of the state-of-the-art data center architectures. Concurr Comp-Pract E 2013; 25: 1771-1783.
[9] Zhang J, Ren F, Lin C. Modeling and understanding TCP Incast in data center networks. In: Proceedings of IEEE INFOCOM; 10–15 April 2011; Shanghai, China. New York, NY, USA: IEEE. pp. 1377-1385.
[10] Ren Y, Zhao Y, Liu P, Dou K, Li J. A survey on TCP Incast in data center networks. Int J Commun Syst 2014; 27: 1160-1172.
[11] Chen Y, Griffith R, Liu J, Katz RH, Joseph AD. Understanding TCP Incast throughput collapse in datacenter networks. In: Proceedings of the 1st ACM Workshop on Research on Enterprise Networking; 16–21 August 2009; Barcelona, Spain. New York, NY, USA: ACM. pp. 73-82.
[12] Alizadeh M, Greenberg A, Maltz DA, Padhye J, Patel P, Prabhakar B, Sengupta S. Data center TCP (DCTCP). ACM Sigcomm Comp Com 2010; 40: 63-74.
[13] Sreekumari P, Jung JI, Lee M. A simple and efficient approach for reducing TCP timeouts due to lack of duplicate acknowledgments in data center networks. Cluster Comput 2016; 19: 633-645.
[14] Xu L, Xu K, Jiang Y, Ren F, Wang H. Throughput optimization of TCP Incast congestion control in large-scale datacenter networks. Comput Netw 2017; 124: 46-60.
[15] Huang J, He T, Huang Y, Wang J. ARS: Cross-layer adaptive request scheduling to mitigate TCP Incast in data center networks. In: 35th Annual IEEE International Conference on Computer Communications; 10–14 April 2016; San Francisco, CA, USA. New York, NY, USA: IEEE. pp. 1-9.
[16] Li S, Li D, Du Z. Adaptive rate control for TCP Incast based on selective ECN-marking. In: 7th IEEE International Conference on Software Engineering and Service Science; 26–28 August 2016; Beijing, China. New York, NY, USA: IEEE. pp. 353-356.
[17] Tseng HW, Chang WC, Peng I, Chen PS. A cross-layer flow schedule with dynamical grouping for avoiding TCP Incast problem in data center networks. In: Proceedings of the International Conference on Research in Adaptive and Convergent Systems; 2016; Odense, Denmark. New York, NY, USA: ACM. pp. 91-96.
[18] Zhang J, Ren F, Yue X, Shu R, Lin C. Sharing bandwidth by allocating switch buffer in data center networks. IEEE J Sel Area Comm 2014; 32: 39-51.
[19] Zou S, Huang J, Zhou Y, Wang J, He T. Flow-aware adaptive pacing to mitigate TCP Incast in data center networks. In: 37th IEEE International Conference on Distributed Computing Systems; 5–8 June 2017; Atlanta, GA, USA. New York, NY, USA: IEEE. pp. 2119-2124.
[20] Adesanmi A, Mhamdi L. M21TCP: Overcoming TCP Incast congestion in data centres. In: IEEE 4th International Conference on Cloud Networking; 5–7 October 2015; Niagara Falls, Canada. New York, NY, USA: IEEE. pp. 20-25.
[21] Abdelmoniem AM, Bensaou B, Abu AJ. Mitigating Incast-TCP congestion in data centers with SDN. Ann Telecommun 2018; 73: 263-277.
[22] Hafeez T, Ahmed N, Ahmed B, Malik AW. Detection and mitigation of congestion in SDN enabled data center networks: a survey. IEEE Access 2018; 6: 1730-1740.
[23] Huang J, Huang Y, Wang J, He T. Adjusting packet size to mitigate TCP Incast in data center networks with COTS switches. IEEE T Cloud Comput 2018; 1: 1-1.
[24] Shukla S, Chan S, Tam ASW, Gupta A, Xu Y, Chao HJ. TCP PLATO: Packet labelling to alleviate time-out. IEEE J Sel Area Comm 2014; 32: 65-76.
[25] Floyd S. TCP and explicit congestion notification. ACM Sigcomm Comp Com 1994; 24: 8-23.