Resource Unavailability PDB

1
Resource Unavailability PDB
draft-karagiannis-ru-pdb-03.txt

• Georgios Karagiannis, Attila Bader, Hannes
  Tschofenig, Lars Westberg

2
Problem to be solved

• Without using signaling, how can a diffserv node
  at outside Diffserv domain(s) detect resource
  unavailability in Diffserv domain(s)?

3
Goal of the draft

• Draft provides a Per Domain Behavior (PDB)
  solution for this problem by using throughput
  (bandwidth) measurements and DSCP remarking in
  Diffserv domain(s) to notify resource
  unavailability to nodes outside the Diffserv
  domain(s)

4
Changes based on received comments

• RU-PDB Applicability section clarified
• Traffic Conditioning Specification (TCS) and PHB
  configuration clarified
• Editorial changes

5
RU-PDB Applicability

• Resource unavailability on Diffserv nodes
  detected using excess traffic measurements and
  DSCP remarking approach where remarking
  proportional to amount of unavailable resources
• Diffserv nodes located outside Diffserv domain(s)
  can use remarked DSCP packets to calculate the
  throughput (or bandwidth) that exceeds the
  loosely agreed SLS throughput bound
• Applied in combination with
• Single Diffserv domain
• Multiple neighboring Diffserv domains when SLA
  agreements exist between operator(s) of the
  domain(s)

6
RU-PDB Applicability

• Sender of traffic blocks new requests until the
  loosely agreed SLS throughput bound is satisfied
  this operation is outside scope of this draft
• Applied in any IP based Diffserv scenario that
  supports real-time applications and use media
  codecs that do not have the benefit of being
  totally free in selecting/producing bit-rates
• minimum bandwidth allocated by using the DSCP
  based RU-PDB mechanism in combination with
  admission control functionality
• when Diffserv network has more capacity it can
  utilize that and give to end user higher quality
  with better end user experience
• end to end ECN method (DCCP) can be used to
  monitor whether the network has more available
  capacity.

7
TCS and PHB Configuration

• Clarified how (1) Meter (2) Marking (3)
  Packet Classification (4) Scheduling work

8
Conclusions and future work

• Without using signaling but using throughput
  (bandwidth) measurements and DSCP remarking in
  Diffserv domain(s) able to notify resource
  unavailability to nodes outside Diffserv domains
• It can be used in combination with end-to-end ECN
  (DCCP) when QoS support for real time
  applications that use media codecs is required.