PERFORMANCE EVALUATION OF DIFFSERV DRIVEN HFC SYSTEM

1
PERFORMANCE EVALUATION OF DIFFSERV DRIVEN HFC
SYSTEM

• Authors
• G.I.Pikrammenos, H.-C. Leligou

2
Presenter
Helen-Catherine Leligou National Technical
University of Athens Electrical and Computer
Engineering Department
3
Architecture of tree-shaped HFC System
4
Implications of the HFC Architecture

• In a tree-shaped asymmetric HFC the upstream
  channel is the bottleneck link
• The MAC controller is the only arbiter of
  upstream bandwidth within the HFC network
• Distributed multiplexing governed by the MAC
• Traffic is buffered locally at the modem queues
• Global knowledge of load conditions only
  indirictely from the MAC protocol information

5
The Importance of Service Differentiation in HFC
systems

• Provide different QoS classes
• Mixing delay-sensitive with best-effort services
  results to low quality and unacceptable
  connection service
• Efficient use of the available bandwidth
• Traffic fluctuations can be taken into account to
  form a dynamic service policy
• More flexible prising policy
• Important for the penetration of an access
  network to the residential environment

6
Features employed to provide QoS

• Priority levels distinction
• Dynamic bandwidth allocation
• Simultaneous request announcement for all
  priorities
• Polling instead of contention

7
MAC protocol scheme

• 1st For CBR-EF pre-arbitrated,unsolicited access
• 2nd For real-time VBR-High AF classes access
  based on requests and policing by token credit
  checking
• 3rd For BBE - AF with min. guaranteed rate
  priority effective while credits confirm
  compliancethe rest (OUT of profile) relegated to
  4th priority
• 4th For BE traffic with loss based congestion
  control Round Robin servicing for equal
  bandwidth
• Note ANTs subscribe to either 3rd or 4th

8
AROMA priorities
DiffServ PHBs
Service mechanism
9
Credit fetch logic
10
Simulation Set-up

• Simulation environment HFC model developed in
  the Ptolemy simulation platform
• System configuration

NOTE 1st priority exhibits deterministic behavior
11
Low delay for high priorities
1st scenario 85 total load
12
2nd priority unaffected low priorities
experience cell loss
System behaviour under heavy load
conditions (110 total load)
13
Buffer size evolution with time
14
Prioritization scheme benefits
The lab demonstrator tests have also verified the
above discussed results.
15
Fairness tests results
Equal performance for modems within the same
priority
16
Conclusions

• Tree-shaped, asymmetric multiple access networks
  impose new requirements on the traffic handling
  based on the MAC protocol employed
• The employed MAC algorithm distinguishes 4
  different priorities to provide QoS
• Simulation results prove that the employed MAC
  scheme offers different QoS levels in terms of
  delay bounds
• QoS guarantees are provided to demanding
  services.

17
Differentiated prioritized services