gec14 session sdn in geni
Download
Skip this Video
Download Presentation
GEC14 Session: SDN * in GENI

Loading in 2 Seconds...

play fullscreen
1 / 13

GEC14 Session: SDN * in GENI - PowerPoint PPT Presentation


  • 100 Views
  • Uploaded on

GEC14 Session: SDN * in GENI. Marshall Brinn, GPO July 11, 2012. * Software -Defined Networking. Outline. Overview Guest Speakers: Joe Mambretti , ICAIR Ilia Baldine , RENCI Nick Bastin , Big Switch Discussion. Overview.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about 'GEC14 Session: SDN * in GENI' - monty


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
gec14 session sdn in geni

GEC14 Session:SDN* in GENI

Marshall Brinn, GPO

July 11, 2012

*Software-Defined Networking

outline
Outline
  • Overview
  • Guest Speakers:
    • Joe Mambretti, ICAIR
    • Ilia Baldine, RENCI
    • Nick Bastin, Big Switch
  • Discussion
overview
Overview

The goal of this session is to characterize the programmable network substrate GENI will provide to network researchers.

  • Some top-level questions:
    • How does GENI infrastructure provide a sandbox for experimenters? [“Slicing the network”]
    • How does an experimenter operate within that sandbox? [“Programming the network”]
    • What constraints does GENI impose on the experimenter in terms of operating within the sandbox?
    • What assumptions can the experimenter make about the isolation properties of that sandbox?
slicing the network and programming the sliced network
Slicing the Network and Programming the (Sliced) Network

Sliced Network

Physical Network

Network traffic within a GENI slice is segregated along some fixed dimension (e.g. VLAN ID), providing a sandbox for researcher.

Experimenter may customize network flows or modify frames/packets at provided programmability loci (e.g. OF switches)

Slice ID=1

…..

NB: whatever dimension is used for slicing the network may not be modified by researcher in programming the network.

Slice ID=N

overview 2
Overview [2]
  • Drilling-down on these Questions:
    • What role should OpenFlow play in providing a sandbox for experimenters?
    • What role should OpenFlow play in allowing experimenters to manipulate that sandbox?
    • What software abstraction of the network should GENI provide to experimenters for “deep programming”?
    • What impact does our approach to stitching across campuses have on the experimenter’s ability to program their network substrate?
    • Should experimenters be provided with visibility and control to (unrequested) intermediate nodes?
    • How close can we come to the goal of providing sandboxes without human-in-the-loop?
proposal geni sdn resource
Proposal: GENI-SDN Resource
  • We suggest thinking of an GENI-SDN as a resource (like other resources managed and sliced at an Aggregate Manager).
  • An GENI-SDN is a virtual, sliced network representing some administrative domain
    • A complete topology may need to be stitched across multiple GENI-SDN’s
  • Much like other resources, an GENI-SDN may be programmed as part of the allocation process
    • As well as during run-time
sample physical topology
Sample Physical Topology

Backbone

Regional

Campus

OF Switch

GENI Rack

Data Center

Wireless

Opt-In

User

The campus may have multiple administrative domains: the data center, plus ‘west campus’ and ‘east campus’, say.

The Regional and Backbone may have nodes at many cities in the region that constitute an SDN-enabled topology. Some nodes may have GENI Racks, others don’t.

mapping physical topology to aggregates
Mapping Physical Topology to Aggregates

It may be that the granularity to view and control network topologies is not the single node, nor the whole network, but, because of administrative domains, somewhere in-between.

Backbone

Regional

Campus

OF Switch

GENI Rack

Data Center

Wireless

Opt-In

User

The Regional and Backbone present aggregates for each GENI rack plus a single aggregate presenting the topology as a GENI-SDN resource.

The campus presents three aggregates: one for each ‘campus’ plus the data center. Each aggregate presents its computation/storage resources plus its topology as a GENI-SDN resource.

some principles for this discussion
Some Principles for this Discussion
  • Nothing should be off the table for this discussion
    • Different Encapsulation / Tunneling protocols
    • Different Software and Hardware switches
  • Let us be informed but not constrained by current S/W or H/W limitations
    • If we can describe the way things ought to work, we can develop a plan that approximates this ideal in the near term, and improves over time.
  • OpenFlow is a critical piece of the GENI vision for Software-definable networks
    • But we should not engineer a solution that precludes other SDN approaches
some principles for this discussion 2
Some Principles for this Discussion [2]
  • It is acceptable to impose some limitations on the experimenter’s operations within their sandbox
    • So long as they are well-documented, relatively fixed and relatively narrow
  • We should be mindful of trust and administrative boundaries as we design our sandboxing approach
    • Key materials should only be required to flow between entities that trust one another

Be inspired but not constrained by how these capabilities have been provided in Meso-scale GENI nor by any plans currently on the drawing board

outline1
Outline
  • Overview
  • Guest Speakers:
    • Joe Mambretti, ICAIR
    • Ilia Baldine, RENCI
    • Nick Bastin, Big Switch
  • Discussion
overview1
Overview

The goal of this session is to detail the programmable network substrate GENI will provide to network researchers.

  • Some top-level questions:
    • How does GENI infrastructure provide a sandbox for experimenters? [“Slicing the network”]
    • How does an experimenter operate within that sandbox? [“Programming the network”]
    • What constraints does GENI impose on the experimenter in terms of operating within the sandbox?
    • What assumptions can the experimenter make about the isolation properties of that sandbox?
ad