Martin Casado and David Erickson and Igor Anatolyevich Ganichev and Rean Griffith and Brandon Heller and Nick Mckeown and Daekyeong Moon and Teemu Koponen and Scott Shenker and Kyriakos Zarifis

EECS Department, University of California, Berkeley

Technical Report No. UCB/EECS-2010-93

June 7, 2010

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2010/EECS-2010-93.pdf

Data centers present many interesting challenges, such as extreme scalability, location independence of workload, fault-tolerant operation, and server migration. While many data center network architectures have been proposed, there has been no systematic way to compare and evaluate them— apples-to-apples—in a meaningful or realistic way. In this paper, we present Ripcord, a platform for rapidly prototyping, testing, and comparing different data center networks. Ripcord provides a common infrastructure, and a set of libraries to allow quick prototyping of new schemes. We built a prototype of Ripcord and evaluated it in software and running on a real network of commodity switches. To evaluate Ripcord, we implemented and deployed several schemes, including VL2 and PortLand. A key feature of Ripcord is its ability to run multiple routing applications, side-by-side on the same physical network. Although our prototype implementation is not production quality, we believe that Ripcord provides a framework for both researchers and data center operators to implement, evaluate, and (eventually) deploy new ideas.

---

BibTeX citation:

@techreport{Casado:EECS-2010-93,
    Author= {Casado, Martin and Erickson, David and Ganichev, Igor Anatolyevich and Griffith, Rean and Heller, Brandon and Mckeown, Nick and Moon, Daekyeong and Koponen, Teemu and Shenker, Scott and Zarifis, Kyriakos},
    Title= {Ripcord: A Modular Platform for Data Center Networking},
    Year= {2010},
    Month= {Jun},
    Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2010/EECS-2010-93.html},
    Number= {UCB/EECS-2010-93},
    Abstract= {Data centers present many interesting challenges, such as
extreme scalability, location independence of workload,
fault-tolerant operation, and server migration. While many
data center network architectures have been proposed, there
has been no systematic way to compare and evaluate them—
apples-to-apples—in a meaningful or realistic way. In
this paper, we present Ripcord, a platform for rapidly
prototyping, testing, and comparing different data center
networks. Ripcord provides a common infrastructure, and
a set of libraries to allow quick prototyping of new schemes.
We built a prototype of Ripcord and evaluated it in
software and running on a real network of commodity
switches. To evaluate Ripcord, we implemented and deployed
several schemes, including VL2 and PortLand. A
key feature of Ripcord is its ability to run multiple routing applications, side-by-side on the same physical network. Although our prototype implementation is not production quality, we believe that Ripcord provides a framework for both researchers and data center operators to implement, evaluate, and (eventually) deploy new ideas.},
}

EndNote citation:

%0 Report
%A Casado, Martin 
%A Erickson, David 
%A Ganichev, Igor Anatolyevich 
%A Griffith, Rean 
%A Heller, Brandon 
%A Mckeown, Nick 
%A Moon, Daekyeong 
%A Koponen, Teemu 
%A Shenker, Scott 
%A Zarifis, Kyriakos 
%T Ripcord: A Modular Platform for Data Center Networking
%I EECS Department, University of California, Berkeley
%D 2010
%8 June 7
%@ UCB/EECS-2010-93
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2010/EECS-2010-93.html
%F Casado:EECS-2010-93