wiki:sol4/MultidomainShakedown

Version 4 (modified by nriga@bbn.com, 10 years ago) (diff)

--

Project Number

1934

Project Title

GENI Science Shakedown

Technical Contacts

PI S. J. Ben Yoo sbyoo at ucdavis.edu
Co-PI Matt Bishop bishop at ucdavis dot edu
Co-PI Chen-Nee Chuah chuah at ece.ucdavis.edu

Participating Organizations

UCDavis, CENIC, ESnet, ICAIR, Cal-IT2, KDDI, U Bristol, Rutgers U, Stanford, ICSI, LBL, SDSC

GPO Liaison System Engineer

Niky Riga nriga at bbn.com

Scope

We will conduct advanced networking experiments aiming at testing the scalability and interoperability of programmable networks running OpenFlow across multiple domains. In particular, we will conduct a number of key experiments testing a new cross-layer and multi-domain OpenFlow control and management mechanism across multiple GENI testbeds while paying special attentions to configurability, security, and monitoring.

Through experimentations, we will test the scalability of GENI network stitching, GENI’s SDN capabilities across multiple domains and the usability of GENI tools in order to setup multi-domain experiments. We deploy experiments from the list below or defined in consultation with the GPO. Detailed description of the experiments is included in the end. The experiment designs and setup will be made available to the GENI community.

  • Experiment #A: Application aware Big Data experiment across multi-domain, heterogeneous network so that the nature of the Big Data applications should utilize the best resources. (Testing of GENI Tools including: Stitching, OpenFlow, ORCA/ExoGENI, and other emerging tools).
  • Experiment #B: Distributed iceberg detection: We evaluate our close-loop analysis and programmable measurement approach by injecting varying degrees of high volumetric flows in CAIDA Backscatter data traces.
  • Experiment #C: Attacking OpenFlow Controller: Given an OpenFlow controller server, how can an attacker gain access to that server, and how can she reconfigure it to give her desired control over the OpenFlow controller?
  • Experiment #D: Intelligent traffic inference: We will implement a prototype of an intelligent SDN based traffic (de)aggregation and measurement paradigm (iSTAMP), which leverages OpenFlow to dynamically partition TCAM entries of a switch/router into two parts. In the first part, a set of incoming flows are optimally aggregated to provide well-compressed aggregated flow measurements that can lead to the best estimation accuracy via network inference process. The second portion of TCAM entries are dedicated to track/measure the most rewarding flows (defined as flows with the highest impact on the ultimate monitoring application performance) to provide accurate per-flow measurements. These flows are selected and "stamped" as important (or rewarding from monitor's perspective) using an intelligent Multi-Armed Bandit (MAB) based algorithm.

Through experimentation we will also test GENI’s capabilities for running multi-domain experiments and will identify limitations of the current deployment. We will develop prototype services that will demonstrate benefits of enhancing GENI to address identified limitations.

We will also provide feedback to the GENI community by reporting on experiences with porting and running multi-domain experiments on GENI, including performance studies.

Depending on the results, we would try to publish our findings to a major Conference or Journal.

Current Capabilities

Milestones

GEC18 Milestone
GEC19 Milestone
GEC20 Milestone

Spiral 5

Project Technical Documents

Quarterly Status Reports

Related Projects

Attachments (13)