Supporting DiffServ with Per-Class Traffic Engineering in MPLS

1
Supporting DiffServ with Per-Class Traffic
Engineering in MPLS
2
Outlines

• The paper is proposing a new MPLS traffic
  engineering scheme to meet the demand for Quality
  of Service (QoS).
• This scheme is a modified version of a previously
  proposed QoS routing algorithm.
• The proposed scheme enhances E-LSP with per-class
  TE and load balancing.
• Compared with the original E-LSP and E-LSP with
  load balancing for the support of EF, AF, and BE.
• Simulation results are shown.

3
Background

• MPLS and DiffServ together provide a potential
  model that supports QoS over IP network.
• DiffServ model divides traffic into a small
  number of classes and allocates resources on a
  per-class basis.
• Because DiffServ has a few different classes, a
  packets class can be marked directly in the
  packet.
• The mark in the packet is called a Differentiated
  Service Cod Point, or DSCP, that has 6-bit long
  within the IP header.

4
Background (2)

• DSCP identifies a per-hop behavior or PHB.
• The standard PHBs include
• Expedited Forwarding (EF) Packets are forwarded
  with minimal delay and low loss.
• Assured Forwarding (AF) Packets have different
  classes and different drop preferences.
• Best Effort (BE) No special treatment.
• DSCPs is carried in the IP header, but MPLS LSRs
  dont examine that. ? we need a way to determine
  the PHB from the label header.
• Two ways to solve that E-LSP, and L-LSP.

5

Background (3)

• E-LSP The EXP field of the MPLS Shim Header
  conveys to the label-switch router (LSR) the PHB
  to be applied to the packet.
• L-LSP - That the packets scheduling treatment is
  inferred by the LSR exclusively from the packets
  label value while the packets drop precedence is
  conveyed in the EXP field of the MPLS Shim Header
  or in the encapsulating link-layer-specific
  selective drop mechanism.

6
Features of the proposed Scheme

• Labeling
• The EXP field encodes the scheduling treatment
  and drop precedence just as in E-LSP.
• Load balancing by Service Classes
• Traffic flows of different service classes is
  distributed over different LSPs.
• Routing
• Uses a simple constraint-based routing algorithm
  that is modified from an Optimal QoS Routing
  Algorithm previously developed 10.

7
Optimal QoS Routing Algorithm

• It solves the QoS multicast routing problem that
  requires two constraints delay and bandwidth.
• It is called also Maximum-Bandwidth with Delay
  Constraint algorithm, MBDC.

MBDC Algorithm
8
The Proposed Algorithm

• i. Prunes the topology database of all links that
  dont have sufficient residual/reservable
  bandwidth or that is administratively forbidden
  (including excess delay) for the LSP
• ii. Find the minimum-cost path towards the LSPs
  egress router (use propagation delay as the cost
  for each link)
• Iii. If several equal-cost paths remain, select
  the one with the fewest number of hops
• Iv. If several equal-cost paths remain, apply the
  load- balancing rule choose the path that has
  the maximum residual bandwidth
• V. Steps i to iv are repeated for each LSP
  computation, beginning from the highest service
  class

9
Performance Evaluation

• Three schemes are simulated
• Original E-LSP
• E-LSP with multiple path load balancing
• E-LSP with per-class traffic engineering, the
  proposed scheme

10
(No Transcript)
11
Simulation Results
Flow 1 (EF) Case 3a, 3b,and 3c have the smallest
delay and delay jitter. Throughput is the same
for all schemes.
12
Simulation Results (2)
Flow 2 (AF) All schemes have similar delay and
throughput. The proposed scheme yields better
delay jitter.
13
Simulation Results (3)
Flow 3 (BE) Delay and delay jitter are reduced
dramatically with the proposed scheme. Throughput
also is improved.
14
Simulation Results (4)
Flow 4 (BE) The proposed scheme in case 3a has
dramatically reduced Delay and delay
jitter. While facing more AF flows, in case 3b,
Throughput, delay, and delay jitter are close to
scheme 1.
15
Simulation Results (5)
Flow 5 (BE/AF/EF) When flow 5 is BE ? flow
3. When flow 5 is AF ? the proposed scheme in
case 3b has much better performance. When flow 5
is EF ? LSP is shifted to 4-9-8, and flow 2,3,4
will have more BW.
16
(No Transcript)
17
Conclusion

• The proposed a per-class TE scheme that enhances
  E-LSP demonstrates better utilization of the
  network resources.
• It is able to accommodate more QoS flows while
  offering better performance to existing flows.