TCP incast solutions in data center networks: A classification and survey

• Main Authors: Alipio, Melchizedek Ibarrientos, Tiglao, Nestor Michael, Bokhari, F., Khalid, Salman
• Format: text
• Published: Animo Repository 2019
• Subjects: TCP/IP (Computer network protocol); Queuing networks (Data transmission); Cloud computing; Data centers; Electrical and Electronics; Systems and Communications
• Online Access: https://animorepository.dlsu.edu.ph/faculty_research/1361; https://animorepository.dlsu.edu.ph/context/faculty_research/article/2360/type/native/viewcontent
• Institution: De La Salle University

In recent years, Data Centers Networks (DCNs) have been deployed to serve as the backbone to support the extensive variety of services offered through the Internet like social networking, web hosting, and e-commerce. The Transmission Control Protocol (TCP) is at present, the most widely used transmission protocol in DCNs and the Internet. When several senders simultaneously send data to a single receiver in a DCN, congestion occurs at the switches' buffer at the receiver's end. This phenomenon of throughput collapse, particularly in data centers, is termed as the TCP incast problem. In this work, we survey and develop a classification scheme of TCP incast solutions in DCNs. We classify the TCP incast solutions on a multi-level criteria based on the TCP/IP protocol stack either TCP or non-TCP approach from our comprehensive survey. Each level includes sublevel classification such as delay-based congestion control schemes, active queue management based solutions, or probabilistic schemes. We also discuss the performance of each existing solutions in terms of its ability to alleviate the network during incast conditions. Furthermore, we present the strengths and weaknesses of each TCP incast techniques. Finally, we outline the open challenges and issues in mitigating the TCP incast in DCNs. We believe that the result of our survey can serve as a design guide to researchers and engineers in designing future DCN protocols. © 2019 Elsevier Ltd