Network traffic characteristics of data centers in the wild

By Theophilus A. Benson, Aditya Akella, David A. Maltz. ACM/SIGCOMM Internet Measurement Conference, 2010

2,458 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/f7c3b5a070e3cb73708964d4d2199c1c146a5527

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

A survey: Control plane scalability issues and approaches in Software-Defined Networking (SDN) (2017, background): “However, that number may not be enough for some network settings, such as data centers, depending on the number of servers and the switches [10,46] .”
CONGA: distributed congestion-aware load balancing for datacenters (2015, background): “First, datacenter traffic is very bursty and unpredictable [18, 28, 8].”
DevoFlow: scaling flow management for high-performance networks (2011, background): “Recent measurement studies have shown data center ﬂow sizes do not follow a Pareto distribution [9,24].”
Presto: Edge-based Load Balancing for Fast Datacenter Networks (2015, background): “Second, an overwhelming fraction of mice ﬂows are less than 64 KB in size and thus do not have to worry about reordering [11, 23, 33].”
Towards an elastic distributed SDN controller (2013, background): “…distributed SDN Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for pro.t or commercial advantage and that copies bear this notice and the full citation on the .rst page.”
Finishing flows quickly with preemptive scheduling (2012, methodology): “Assuming a fluid traffic model (infinitesimal units of transmission), the result given by fair sharing is shown in Figure 1b: [fA,fB ,fC ] finish at time [3,5,6], and the average flow completion time is 3+5+6 3 = 4.”
A Survey on Optical Interconnects for Data Centers (2012, background): “There are several research papers that have investigated the data center traffic such as the ones presented by Microsoft Research ([23],[24],[25]).”
Understanding network failures in data centers: measurement, analysis, and implications (2011, background): “Similar to prior studies of data center network traffic [5], we observe higher utilization at upper layers of the topology as a result of aggregation and high bandwidth oversubscription [12] .”
A roadmap for traffic engineering in SDN-OpenFlow networks (2014, background): “Such difference in traffic burstiness leads to high packet loss rates of the underutilized edge and aggregation links, and low packet loss rates of highly utilized core links [66,67].”
OpenNF (2014, methodology): “W e use a combination of replayed university-to-cloud [25] and datacenter [21] network traffic traces, along with synthetic wor kl ads.”
DRILL: Micro Load Balancing for Low-latency Data Center Networks (2017, background): “[3, 13, 14, 23], as even the fastest ones have control loops with 10s of millisecond to a few second delays [19, 25, 61].”
A Survey of Machine Learning Techniques Applied to Software Defined Networking (SDN): Research Issues and Challenges (2019, background): “However, the majority of bytes are carried in elephant flows [142].”
On Controller Performance in Software-Defined Networks (2012, background): “[2] show that a network with 100 switches can have spikes of 10M flows arrivals per second in the worst case.”
Mirror mirror on the ceiling: flexible wireless links for data centers (2012, background): “…center computing is likely to continue to require wired network deployments, in this paper we focus on the subset of applications, andperhapsthemajority of applications asborneoutby recent measurement studies[15,16,25], with more modest or Copyright2012ACM978-1-4503-1419-0/12/08 ...$15.00.”
Inside the Social Network's (Datacenter) Network (2015, background): “note that for cloud datacenters “a majority of traffic originated by servers (80%) stays within the rack” [12].”
Elastic sketch: adaptive and fast network-wide measurements (2018, background): ““Skewed” means most flows are mouse flows [34], while a few flows are elephant flows [4, 22, 35].”
Language-Directed Hardware Design for Network Performance Monitoring (2017, background): “5 hour university data-center trace (~100M packets) from 2010 [32].”