Guest

Some Areas of Interest

QoS Frameworks

Goals and Objectives

This proposal aims at creating and validating a network infrastructure for the delivery of real-time sensitive traffic over the Internet.

Languages

PDF document View PDF version

Overview

As new applications start using the Internet, the requirements that they impose to the IP network are not always fulfilled. In particular time-sensitive applications such as videoconferencing, video distribution and voice have strict requirements in terms of packet loss, latency, jitter and bandwidth. This new scenario forces the network to evolve towards a model that:

  • Creates a measurement infrastructure
  • Defines QoS requirements
  • Dynamically selects an end-to-end optimal path based on the applications' QoS requirements
  • If the requested QoS level can not be achieved, then it mediates among application demands using distributed Admission Control
  • Allows applications to specify their QoS requirements and automatically maps them to a redefined Service Level Agreement
  • Provides secure identification of the sources of resource requests and the reservation of resources.

Distributed Measurement Framework


Historically Ethernet has not been well suited for applications requiring a common precise time reference because of its non-deterministic nature. Recently introduced time synchronization protocols such as IEE 1588 have helped mitigate the latency and jitter issues through precise hardware assisted time-stamping function at the physical layer of the network by using an Ethernet packet to carry the time reference to all the involved clocks. Network devices can be synchronized within accurateness in the 100s of nanoseconds range, thereby, facilitating real-time applications.

Monitoring and reporting tools for real-time traffic properties should be based upon a well defined measurement framework including:

  • A method for collecting measurement statistics
  • A method for processing the collected statistics
  • A method for dynamically visualizing the updated properties of the network.

The results can then be reused in multiple other blocks such as QoS routing or the SLA management.

QoS Routing


QoS-aware routing algorithms take both a set of constraints and some basic knowledge of network conditions and availability of resources to determine an optimized path for a flow. However, this class of algorithm usually takes into consideration a static view of conditions and network resources to selects the path.

A more dynamic algorithm is sought. In order to be efficient, such an algorithm needs to be tightly integrated with resources allocation, traffic engineering and admission control.

Distributed Admission Control (DAC) Mechanism


  • Admission Control refers to the network’s ability to accept or reject a new session based upon session requirements and the resources available along the path which will be traversed
  • A DAC mechanism distributes the admission control without the involvement of a central entity to keeps track of the global network resources
  • A DAC may be implemented through resource reservation protocols such as RSVP.  RSVP is sometimes deemed too heavy and un-scalable, making it unattractive for the QoS framework. Therefore a lightweight DAC is a research objective.

Automatic SLA Definition


  • Definition of SLA represented with multiple independent parameters
  • SLA enforcement and verification techniques
  • Application identification for SLA assignment
  • Research on application impact of SLAs.

RFP: Application Flow Management and Service Assurance

Flow Based Framework & Traffic Steering


As we move towards a "connection based routing" we need to be aware of flows and connections. Scaling issues may restrict the ability to track individual flows to the edge of the network while aggregated flows are all that can be monitored in the core.
Solutions should be structured at both the control-plane and the data-plane level with a particular emphasis on integration challenges with existing protocols.

Trust and Identity


The identification of the sources of resource requests, the mediators amongst the various requests, and the network control plane must be mutually trusted to avoid attacks against the applications as well as the shared network transport mechanisms. The set of trusted identities must span:

  • Across service provider domains
  • Among applications and to the network
  • Among parties involved in Admission Control
  • Among parties associated with distributed protocols.

If your research topic doesn't directly address any of the proposed RFPs please submit it as a generic proposal for this area.

If you have comments, questions or feedbacks related to this area please don't hesitate to contact us at research-overlay@cisco.com

Return to Top