A Resource Delegation Framework for Software-Defined Networks Ilya - - PowerPoint PPT Presentation

a resource delegation framework for software defined
SMART_READER_LITE
LIVE PREVIEW

A Resource Delegation Framework for Software-Defined Networks Ilya - - PowerPoint PPT Presentation

A Resource Delegation Framework for Software-Defined Networks Ilya Baldin ibaldin@renci.org Shu Huang shuang@renci.org Rajesh Gopidi rajesh@renci.org The Problem B A E F D C Offering network services in a


slide-1
SLIDE 1

A Resource Delegation Framework for Software-Defined Networks

Ilya Baldin ibaldin@renci.org Shu Huang shuang@renci.org Rajesh Gopidi rajesh@renci.org

slide-2
SLIDE 2

The Problem

2

  • Offering network services in a multi-domain environment
  • In SDN environment primarily boils down to

– Controller coordination for provisioning (e.g. DISCO) – ‘Big-switch’ abstractions

  • What we would like

– Offer heterogeneous network services (QoS, resiliency, virtual networks) across many providers – Allow providers to trust ‘alien’ controllers – Support nested virtualization – Separate resource management from provisioning

  • Split off coordination of resource allocation on long term scale from short-term

provisioning

This ¡work ¡is ¡funded ¡by ¡DOE ¡ASCR ¡award ¡DE-­‑SC0007425 ¡and ¡NSF ¡grant ¡ ¡CNS-­‑1111256 ¡ ¡

A B C E F D

slide-3
SLIDE 3

The Overview

  • Multi-domain environment where customers use

‘services’ (connections or virtual networks)

– Different QoS requirements (latency, bandwidth, resiliency, isolation) – E.g. the Internet can be one of the overlays

Presentation title goes here 3

A B C E F D

slide-4
SLIDE 4

The Problem Formulation

  • Label

– Labels can be translated in some points in the network

  • Technology level

– Binds to a label offset or type. – Technology levels have a partial order where A < B indicates B can enclose A and carry its traffic – E.g. 802.3 > IP (not necessarily OSI-compliant)

  • Label extents within technology levels are defined per port, per

network element along with other consumable resources

– Outgoing bandwidth, Buffer space, Flow table space

  • Represent multi-dimensional spaces that can be tested for

inclusion/intersection

  • A portion of such volume represents a delegation of resources

– Can be used by a controller to form forwarding rules managing traffic – Can be further subdivided to create nested delegations

4

slide-5
SLIDE 5

Example

  • Constraints are needed to describe delegations

– Volume inclusion – Path/label continuity – Label/bandwidth accounting

  • ILP formulation in the paper

5

A D E C B 1,2,3 1,5,6 A E

Dt d Dt c

1 1 1 1 2 2 3 3 5 5 6 6

X Y

Customer ¡View ¡ Provider ¡View ¡

slide-6
SLIDE 6

The Architecture

  • Form and describe

delegations

– Resource manager – Coordinated by customer, directly or via broker

  • Inform controllers

– Controllers know the constraints

  • Enforce delegations on

behalf of providers at switch level

– PDP vs. PEP – Flow space manager

  • Pervasive Authz

6

RESOURCE MANAGER

CONTROLLER

RESOURCE MANAGER

CONTROLLER

AUTHORIZATION FRAMEWORK

RESOURCE MANAGER

CONTROLLER

Flow Space Manager Network Element

slide-7
SLIDE 7

The Prototype

  • Delegation framework

– GUI tool to form reservations and describe using GraphDB – Floodlight module to accept these descriptions

  • Sample multi-domain application

– Virtual transport provider built out of 3 other providers (virtual or physical) – Provides transport path-based services using a portion of L2 MAC address field delegated to it for path identification

  • Tested in a GENI slice

7

slide-8
SLIDE 8

The Experiment

8

1 2 3 4 5 6 7 8 9

A B C

1 2 3 4 6 7 8 9

Transport Provider D

slide-9
SLIDE 9

The Experiment (GENI Slice)

9

slide-10
SLIDE 10

The Prototype Modules in Floodlight

  • Topology delegation

– Accepts constrained delegation descriptions in GraphDB

  • Topology verification

– Uses stochastic probing across delegated label space to verify connectivity – LLDP not suitable for this purpose

  • ARP Resolution

– Listens for client ARP requests – Performs substitution of MAC address with Path IDs

  • Circuit computation

– Computes paths and assigns path IDs

10

slide-11
SLIDE 11

Outcomes

  • Direct control over provider equipment with

verifiable constraints

  • Explicit communication of constraints to

controllers

  • Nested virtualization
  • Efficient use of label spaces
  • Dynamic resource allocation
  • Multiple approaches in one architecture
  • Support for an economy

11