Packet scheduling with delay and loss differentiation

1
Packet scheduling with delay and loss
differentiation

• A. Striegel and G. Manimaran
• Computer Communication, ELSEVIER
• Vol. 25, issue 1, pp. 21-31, Jan 2002

2
Outline

• Introduction
• Differentiated services
• Proposed differential schedulers
• Simulation studies-per-hop behavior
• Simulation studies-end-to-end behavior
• Conclusions

3
Introduction

• IETF proposes several models to support QoS
• Integrated Services
• Aims to provide an absolute guarantee of QoS for
  each flow across the network
• Provides two broad categories of services
• Guaranteed service
• Controlled load service
• Relies on RSVP for admission control and resource
  reservation

4
Introduction (cont.)

• Has several key weaknesses
• Each router requires a significant amount of
  processing overhead
• Each router is required to maintain state
  information for each flow
• IntServ is impractical for short-lived flows
• Since the connection setup overhead is often
  greater than the transmission of all packets in a
  flow
• Differentiated Services

5
Differentiated services

• The goal of DiffServ was to provide the benefits
  of different levels of QoS
• The DiffServ model accomplishes this by
  aggregating traffic with similar QoS requirements
  into classes
• DiffServ dont maintain any per-flow information
  across the network

6
Differentiated services (cont.)

• DiffServ has several weaknesses
• Policing and resource reservation are migrated to
  the edge of the DiffServ domain, leaving the
  routers inside the DiffServ core to act upon the
  DSCP within a given packet
• Expedited Forwarding
• Is equivalent to a leased line provided that
  throughput stays below a certain level
• Assured Forwarding
• Classifies packets into four classes with three
  drop precedence levels within each class

7
Differentiated services (cont.)

• Relative differentiated services
• All traffic is grouped into N classes of service
• For each class i, the service provided to class i
  will be better (or at least no worse) than the
  service provided to class (i-1)
• To implement relative differentiated services,
  several approaches have been proposed
• Price differentiation
• Careful capacity provisioning

8
Differentiated services (cont.)

• Two principles were proposed in order for a
  differentiated services model to be effective for
  users and network operators
• Predictable
• Controllable
• From these two principles, the proportional
  differentiation model was proposed

9
Differentiated services (cont.)

• The premise of the proportional differential
  model
• The performance measures for packet forwarding at
  each hop are ratioed proportionally via the use
  of class differentiation parameter
• Two schemes were proposed for applying the
  proportional differential model to packet delay
• Backlog Proportional Rate (BPR)
• Waiting-Time Priority (WTR) scheduling

10
Differentiated services (cont.)

• The proportional loss dropper uses the notion of
  a loss history buffer (LHB) which captures the
  loss information for the last K packets received

11
Proposed differential schedulers

• Proportional differential model
• The goal is to appropriately space class
  performance metrics according to the
  differentiation parameters
• Consider N classes and let the performance metric
  of a class i be denoted by Qi
• Qi/QjCi/Cj (i,j1N) where C1ltC2ltCN is defined
  as the generic quality differentiation parameters

12
Proposed differential schedulers (cont.)

• (m,k) model and DBP scheduling scheme
• We adopt the (m,k) model and the Distance Based
  Priority (DBP) scheduling algorithm for
  scheduling real-time streams
• The (m,k) model provides a unique method to
  capture the loss constraints of an individual
  flow
• For every k consecutive packets in a given flow,
  at least m packets have to meet their end-to-end
  delays

13
Proposed differential schedulers (cont.)
14
Proposed differential schedulers (cont.)

• The DBP value of a flow is the number of
  transmissions required to reach a failing state
• The lower the DBP value, The higher the priority
• The DBP scheme provides a method for evaluating
  the history of a flow based on the transmission
  status of the last k packets of the flow
• The history is then used to prioritize the flow

15
Proposed differential schedulers (cont.)

• Proposed generic class scheduler (C-DBP)
• We propose a variation of the DBP scheduler,
  Class DBP (C-DBP), for use with the Proportional
  Differentiated model
• The number of queues is equal to the number of
  differentiation classes as opposed to the number
  of flows as in the DBP scheduler
• The state history is maintained at a class level,
  not a flow level as with DBP
• Given two classes Ci and Ci1, class Ci will
  receive better or at least no worse service than
  class Ci1

16
Proposed differential schedulers (cont.)

• With the C-DBP algorithm, each class in the
  scheduler can be differentiated based upon on the
  (m,k) parameters for a given class
• The C-DBP scheme consists of two main parts
• The buffer management algorithm
• The scheduling algorithm
• To distinguish the loss and delay differentiation
  behaviors, we use two sets of DBP histories
• DBPLossHistory
• DBPDelayHistory

17
Proposed differential schedulers (cont.)
18
Proposed differential schedulers (cont.)

• The proportional differentiation of the C-DBP
  algorithm arises from the (m,k) parameters of
  each of the class queues
• A set of differentiation parameter, s, is chosen
  by the network operator along with a k parameter
• The (m,k) parameter for class Ci is (k-si,k)

19
Proposed differential schedulers (cont.)

• C-DBP-delay and C-DBP-loss schedulers

20
Proposed differential schedulers (cont.)

• C-DBP-delay-loss scheduler
• The coupled delay-loss class CDL for a given flow
  with delay class CD and loss class CL is CDLCDCL

21
Simulation studies-per-hop behavior

• Uniform traffic load

22
Simulation studies-per-hop behavior (cont.)
23
Simulation studies-per-hop behavior (cont.)
24
Simulation studies-per-hop behavior (cont.)

• Non-uniform traffic load

25
Simulation studies-per-hop behavior (cont.)

• Short-term performance-uniform traffic load

26
Simulation studies-end-to-end behavior
27
Conclusions

• The C-DBP-Delay and C-DBP-Loss schedulers
  differentiate respectively for delay and loss
  only
• The C-DBP-Delay-Loss scheduler was able to
  provide predictable and controllable
  differentiation both in delay and loss
• The C-DBP-Delay-Loss scheduler is highly flexible
  in that it offers a wide range of QoS selection
  in which the delay and loss can be chosen
  independently