The nature of data center traffic: measurements & analysis

By Srikanth Kandula, Sudipta Sengupta, Albert G. Greenberg, Parveen Patel, Ronnie Chaiken. ACM/SIGCOMM Internet Measurement Conference, 2009

1,177 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Networks and Communications, Information Systems

View Paper: https://www.semanticscholar.org/paper/7a44de587ba362df445327c56bc8be6c7609af31

How researchers see this paper

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

In Artificial Intelligence

In Computer Networks and Communications

In Information Systems

Citations in the corpus

A Survey on Software-Defined Network and OpenFlow: From Concept to Implementation (2014, background): “In the meantime, in [82] authors reported that for a data center with over 1000 servers, it could”
Presto: Edge-based Load Balancing for Fast Datacenter Networks (2015, methodology): “Last, since most bytes in datacenter networks originate from elephant ﬂows [5, 11, 33], this ensures that a signiﬁcant portion of datacenter trafﬁc is routed on uniform sizes.”
Inside the Social Network's (Datacenter) Network (2015, background): “The cluster-wide uniformity outside the local rack accounts for intra-cluster trafﬁc representing over 80% of Hadoop trafﬁc—even though trafﬁc to the local rack dominates any given other rack in isolation.”
On Controller Performance in Software-Defined Networks (2012, background): “[9] found that a 1500-server cluster has a median flow arrival rate of 100k flows per second.”
Towards predictable datacenter networks (2011, methodology): “With a typical communication pattern [20], 99 percentile computation time is 84ms.”
Sharing the Data Center Network (2011, background): “Tasks that move data over congested links suffer collateral damage – they are more likely to experience failures and become stragglers at the job level [6, 22].”
On the impact of packet spraying in data center networks (2013, background): “Further, it has high signaling and connection establishment complexity for short flows, which typically dominate the data center environment [2], [3].”
Joint VM placement and routing for data center traffic engineering (2012, methodology): “[23] reported an empirical observation of flow duration distribution in a 1500-server Microsoft data center over a twomonth period.”
End-to-end Performance Isolation Through Virtual Datacenters (2014, background): “known [15, 39], most workloads exhibit bursty behavior at very fine timescales (below 1 sec); hence, using a very short control interval leads to large estimation errors.”
DeepRecSys: A System for Optimizing End-To-End At-Scale Neural Recommendation Inference (2020, background): “Previous work has shown that, these distributions can lead to different system design optimizations [22], [29].”
Trends in big data analytics (2014, background): “However, a typical data center network environment is radically different from wide area networks [57].”
Hedera: Dynamic Flow Scheduling for Data Center Networks (2010, background): “Recent data center traffic studies [4, 16, 24] show tremendous variation in the communication matrix over space and time; a typical server exhibits many small, transactional-type RPC flows (e.”
Mirror mirror on the ceiling: flexible wireless links for data centers (2012, background): “Second, our scheduler needs to handle short-lived traffic bursts [15, 25], and thus must be online.”
Pingmesh: A Large-Scale System for Data Center Network Latency Measurement and Analysis (2015, background): “[21] and [6] studied traﬃc and ﬂow characteristics of diﬀerent types of data centers, by collecting network traces.”
Augmenting data center networks with multi-gigabit wireless links (2011, result): “We also verify that flow sizes and arrival rates in the DC [6, 14 ] indicate that traffic in the DC lies in a fast-changing collection of medium-sized flows.,We have designed a simple traffic estimation scheme that uses a shim layer (an NDIS filter driver) on servers to collect traffic statistics, ”
DevoFlow: scaling flow management for high-performance networks (2011, methodology): “at Microsoft Research (MSR) [30], of a 1500-server cluster.”
Data center TCP (DCTCP) (2010, background): “The data center environment [ 19 ] is significantly different from wide area networks.”