VL2: a scalable and flexible data center network

By Albert G. Greenberg, James R. Hamilton, Navendu Jain, Srikanth Kandula, Changhoon Kim, Parantap Lahiri, David A. Maltz, Parveen Patel, Sudipta Sengupta. Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication, 2009

2,812 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Information Systems

View Paper: https://www.semanticscholar.org/paper/18865de480ed9da8d67015f68458ce03cdbae40e

How researchers see this paper

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

In Computer Networks and Communications

In Information Systems

Citations in the corpus

B4: experience with a globally-deployed software defined wan (2013, background): “Load balancing and multipath solutions have largely focused on data center architectures [1, 18, 20], though at least one eort recently targets the WAN [22].”
XORing Elephants: Novel Erasure Codes for Big Data (2013, background): “However, it is consistently reported that the network is typically the bottleneck for large-scale MapReduce tasks [5, 14, 15].”
Onix: A Distributed Control Platform for Large-scale Production Networks (2010, methodology): “Similarly, Onix can be the platform for flexible data center network architectures such as SEATTLE [21], VL2 [14] and Portland [25] to manage large data centers.”
A Survey on Network Methodologies for Real-Time Analytics of Massive IoT Data and Open Research Issues (2017, background): “different paths of the trees is typically oversubscribed by factors of 1:5 [163] or even more.”
Copa: Practical Delay-Based Congestion Control for the Internet (2018, methodology): “In datacenter network simulations, on a web search workload trace drawn from datacenter network [16], Copa achieved ≈ 4× reduction in”
Improving the Scalability of Data Center Networks with Traffic-aware Virtual Machine Placement (2010, background): “While multi-level network architectures, such as BCube [5] , reap the most benefit, architectures that have been designed with uniform traffic patterns in mind, such as VL2 [3], see little benefit in optimally solving the TVMPP.”
An adaptive hybrid elasticity controller for cloud infrastructures (2012, background): “As the workload observed in data centers is dynamic [8], setting an”
Inter-datacenter bulk transfers with netstitcher (2011, background): “They also report that wide-area transit bandwidth costs more than building and maintaining the internal network of a datacenter, a topic that has recently received much attention [26].”
NetworkCloudSim: Modelling Parallel Applications in Cloud Simulations (2011, neutral): “The communication links are generally shared and interconnected through fat-tree type of network architecture [12].”
Composing Software Defined Networks (2013, neutral): “Later, the discovery module could be replaced with one implementing a directory mapping between MAC and IP addresses as in VL2 [5], without having to change the other module.”
Fastpass (2014, background): “Tunneling (e.g. IP-in-IP or GRE) entails a small throughput overhead, but is supported in hardware by many switch chipsets making it a viable option [18].”
RDMA over Commodity Ethernet at Scale (2016, methodology): “Our data center network is an Ethernet-based multi-layer Clos network [1, 3, 19, 31] as shown in Figure 1.”
The only constant is change: incorporating time-varying network reservations in data centers (2012, background): “By more precisely capturing the networking requirement of applications, TIVC enables the cloud provider to schedule more jobs to run concurrently not only in over-subscribed networks, but also in networks with full-bisection bandwidth, such as fat-trees [7, 19].”
Ananta: cloud scale load balancing (2013, background): “This two-level Clos network architecture [11] typically has an over-subscription ratio of 1 : 4 at the spine layer.”
The nature of data center traffic: measurements & analysis (2009, methodology): “Greenberg etal.[ 18 ] reportdatacentertra�ccharacteristics–variabilityatsmalltimescale sandstatistics on �ow sizes and concurrent �ows, and use these to guide network design.”
Cloud computing: state-of-the-art and research challenges (2010, methodology): “[21] report data center traffic characteristics on flow sizes and concurrent flows, and use these to guide network infrastructure design.”
A Survey on Optical Interconnects for Data Centers (2012, background): “While there are several research efforts that try to increase the required bandwidth of the data centers that are based on commodity switches (e.g. using modified TCP or Ethernet enhancements [22]), the overall improvements are constraints by the bottlenecks of the current technology.”
Finishing flows quickly with preemptive scheduling (2012, background): “[12] demonstrated that, under a production data center of a large scale cloud service, the number of concurrent flows going in and out of a machine is almost never more than 100.”
Understanding network failures in data centers: measurement, analysis, and implications (2011, methodology): “Figure 1: A conventional data center network architecture adapted from figure by Cisco [12].”