Inside the Social Network's (Datacenter) Network

By Arjun Roy, Hongyi Zeng, Jasmeet Bagga, George Porter, Alex C. Snoeren. Computer communication review, 2015

1,133 citations recorded by the source corpus.

Large cloud service providers have invested in increasingly larger datacenters to house the computing infrastructure required to support their services. Accordingly, researchers and industry practitioners alike have focused a great deal of effort designing network fabrics to efficiently interconnect and manage the traffic within these datacenters in performant yet efficient fashions. Unfortunately, datacenter operators are generally reticent to share the actual requirements of their applications, making it challenging to evaluate the practicality of any particular design. Moreover, the limited large-scale workload information available in the literature has, for better or worse, heretofore largely been provided by a single datacenter operator whose use cases may not be widespread. In this work, we report upon the network traffic observed in some of Facebook's datacenters. While Facebook operates a number of traditional datacenter services like Hadoop, its core Web service and supporting cache infrastructure exhibit a number of behaviors that contrast with those reported in the literature. We report on the contrasting locality, stability, and predictability of network traffic in Facebook's datacenters, and comment on their implications for network architecture, traffic engineering, and switch design.

Subfields: Computer Networks and Communications, Information Systems

View Paper: https://www.semanticscholar.org/paper/28fed5a9e3f5468c18791e4afd09da3f4e3bc882

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

Homa: a receiver-driven low-latency transport protocol using network priorities (2018, neutral): “W4 Hadoop cluster at Facebook [28].”
HPCC: high precision congestion control (2019, methodology): “We use widely accepted and public available data center traffic traces, WebSearch [43] and FB_Hadoop [37] in both testbed experiments and simulations.”
DRILL: Micro Load Balancing for Low-latency Data Center Networks (2017, methodology): “arrival times from [62], and scale the interarrival times to emulate various degrees of network load.”
FlowRadar: A Better NetFlow for Data Centers (2016, background): “In Facebook data center and traffic setting [35], a rack switch connects to 44 hosts with 10Gbps links, where each host send at most 100s to 1000s of concurrent flows in 5ms.”
ProjecToR: Agile Reconfigurable Data Center Interconnect (2016, background): “Roy et al. made a similar observation for trafﬁc matrices in Facebook’s data centers [35].”
Re-architecting datacenter networks and stacks for low latency and high performance (2017, methodology): “We use measurements from Facebook’s network to setup our experiment [34].”
Packet Transactions: High-Level Programming for Line-Rate Switches (2015, background): “This rigidity affects many stakeholders adversely: vendors [7, 9, 4] building network switches with merchant-silicon chips [13, 14, 21], network operators deploying switches [85, 80, 58], and researchers developing new switch algorithms [64, 74, 96, 99, 60].”
Elastic sketch: adaptive and fast network-wide measurements (2018, background): “It can happen frequently within a single second [19] and be as large as more than half of the network bandwidth [9].”
PINT: Probabilistic In-band Network Telemetry (2020, background): “, up to 100K [62]), in the case of path tracing, is challenging for limited switch space (e.”
Datacenter RPCs can be General and Fast (2018, background): “[55] report that 99% of all datacenter links are less than 10% utilized at one-minute timescales.”