Data center TCP (DCTCP)

Abstract

Cloud data centers host diverse applications, mixing workloads that require small predictable latency with others requiring large sustained throughput. In this environment, today's state-of-the-art TCP protocol falls short. We present measurements of a 6000 server production cluster and reveal impairments that lead to high application latencies, rooted in TCP's demands on the limited buffer space available in data center switches. For example, bandwidth hungry "background" flows build up queues at the switches, and thus impact the performance of latency sensitive "foreground" traffic. To address these problems, we propose DCTCP, a TCP-like protocol for data center networks. DCTCP leverages Explicit Congestion Notification (ECN) in the network to provide multi-bit feedback to the end hosts. We evaluate DCTCP at 1 and 10Gbps speeds using commodity, shallow buffered switches. We find DCTCP delivers the same or better throughput than TCP, while using 90% less buffer space. Unlike TCP, DCTCP also provides high burst tolerance and low latency for short flows. In handling workloads derived from operational measurements, we found DCTCP enables the applications to handle 10X the current background traffic, without impacting foreground traffic. Further, a 10X increase in foreground traffic does not cause any timeouts, thus largely eliminating incast problems.

Multiple Congestion Points and Congestion Reaction Mechanisms for Improving DCTCP Performance in Data Center Networks

“…The buffer size was set to 128 KB for Switch 1 and 512 KB for Switches 2, 3, and 4. The marking threshold value 'K' for DCTCP was set according to [10,21] …”

“…Recently, a few solutions [10][11][12][13][14][15][16][17][18] have been proposed to increase TCP performance in a DCN. Among the existing solutions, Data Center TCP (DCTCP) has gained more popularity in academic as well as industry areas due to its better performance in terms of throughput and latency.…”

“…Among the existing solutions, Data Center TCP (DCTCP) has gained more popularity in academic as well as industry areas due to its better performance in terms of throughput and latency. DCTCP [10] uses a very small buffer space compared with other existing solutions. However, in recent research [19][20][21][22][23][24][25] about the performance of DCTCP, the authors found that most times the sender's congestion window reduces to one segment, which results in timeouts.…”

CDFS: A Centralized Dynamic Flow Scheduling Mechanism in Data Centers

“…Several studies [1][2][3][4][5] indicate that many applications in datacenters fall into two broad categories: bandwidth-consuming, and latency-sensitive. Bandwidth-consuming applications (e.g., file transfer, email delivery and big data analysis) generate long-lived and high-throughput flows (known as "elephant" flows).…”

“…Bandwidth-consuming applications (e.g., file transfer, email delivery and big data analysis) generate long-lived and high-throughput flows (known as "elephant" flows). While latency-sensitive applications (including web browsing, ICMP service and RPC) produce "mice" flows with relatively small size (typically smaller than 1MB [2]). Elephant flows comprise most of the network throughput, whereas a majority of flows are mice ones.…”