Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 QoS Control Schemes for Two-Stage Ethernet Passive Optical Access Networks Speaker : Hsuan-Ming Chen Advisor : Ho-Ting Wu Date : 2005.10.18.

Similar presentations


Presentation on theme: "1 QoS Control Schemes for Two-Stage Ethernet Passive Optical Access Networks Speaker : Hsuan-Ming Chen Advisor : Ho-Ting Wu Date : 2005.10.18."— Presentation transcript:

1 1 QoS Control Schemes for Two-Stage Ethernet Passive Optical Access Networks Speaker : Hsuan-Ming Chen Advisor : Ho-Ting Wu Date : 2005.10.18

2 2 Outline Introduction Overview and Proposed Architecture Proposed Architecture Dynamic Bandwidth Allocation with QoS Support for IP-EPONS Multipoint Control Protocol (MPCP) Inter-ONU DBA scheme Sub-OLT Scheduling Performance Evaluation Conclusion

3 3 Introduction An EPON (Ethernet passive optical networks) a point-to-multipoint (1:N)optical access network with no active elements in the signal path The network provides two-way operation GATE REPORT

4 4 Introduction (cont.) EPON network architecture

5 5 Overview and Proposed Architecture Time-sharing protocols represent the best method of optical channel sharing in optical access networks (i.e. time division multiple access, TDMA) can either be static (simple) or dynamic Dynamic bandwidth allocation based upon a time interleaving method (i.e., the interleaved polling scheme with an adaptive cycle time (IPACT) scheme)

6 6 Proposed Architecture Present a cascaded two-stage EPON architecture Adds another intermediate level of ONU nodes to the network, termed sub-OLT Refer to this architecture as penetrated-EPON (P-EPON)

7 7 Proposed Architecture (cont.)

8 8 Illustration of sub-OLT functionalities

9 9 Proposed Architecture (cont.) Major motivations for deploying a two-stage P- EPON architecture Allow more end-users (ONU nodes) to share the uplink OLT bandwidth, without incurring extra overhead for switch-over between users Enable longer access reach/distances, since the use of intermediate sub-OLT nodes effectively adds another level of electrical regeneration Reduce OLT hardware complexity significantly

10 10 Multipoint Control Protocol (MPCP) EPON standard, the multipoint control protocol (MPCP) 0. OLT sends a GATE message to each ONU through the downstream connection 1. Gate message contains time when the ONU should start transmission and the length of its transmission window 2. Upon receiving its GATE message, the ONU performs synchronization and updates its local parameters 3. When its local clock matches the transmission start time sent by the OLT 4. The ONU starts sending information packets to the OLT

11 11 Multipoint Control Protocol (MPCP) 5. At the end of its transmission window, the ONU sends a REPORT message to the OLT to report its current buffer occupancy and request bandwidth for the next transmission cycle 6. Upon receiving REPORT message from the ONU, the OLT updates its report table and passes the message to the DBA module responsible for bandwidth allocation decision 7. At the end of the transmission cycle and upon receiving REPORT messages from all ONUs in the network, the DBA module in the OLT calculates the new window size for every ONU in the network 8. a series of GATE messages are generated and broadcast continuously to the ONUs by the OLT to initiate the next transmission cycle

12 12 Transmission Cycle Time Denote transmission cycle time by T TCT, which is the sum of all ONUs ’ transmission time If T TCT is too large the average packet queueing delay inside ONU nodes might increase randomly If T TCT is restricted too much the average packet delay might still increase unnecessarily, because more bandwidth is wasted on frequent MPCP messaging and inter-ONU guard times Static bandwidth allocation algorithms that each ONU is granted a fixed transmission timeslot per transmission cycle. Namely, the ONU timeslot can be calculated as :

13 13 Inter-ONU DBA scheme t i = (T TCT – N x t g ) x w i with Where t i is the timeslot allocated to ONU i w i is the weight of ONU I according to its service level agreement (SLA) N is the number of ONU nodes

14 14 Inter-ONU DBA scheme (cont.) However, assigning static capacity to ONU nodes will preclude idle capacity reuse and yields generally lower utilizations for bursty traffic To mitigate the above concerns, the others proposed an inter-ONU DBA scheme in which ONU nodes are portioned into two groups Underloaded requesting bandwidth below their minimum guarantee defined by the maximum transmission Overloaded

15 15 Inter-ONU DBA scheme (cont.) This algorithm can be described as follows:

16 16 Inter-ONU DBA scheme (cont.) r i is the requested transmission time of ONU node i is the total excess transmission time saved from underloaded ONU nodes is the corresponding share of the total excess transmission time allocated to overloaded ONU node K and M are the set of overloaded and underloaded ONU nodes, respectively

17 17 Inter-ONU DBA scheme (cont.) The above algorithm may still yield some wasted (unused) bandwidth capacity  The total excess transmission time might not always fully occupied by the overloaded ONU nodes

18 18 Inter-ONU DBA scheme (cont.) modify this algorithm with a more advanced dynamic cycle time upper bounded at T MAX-TCT as follows: Where denotes the total extra transmission time requested by overloaded ONU nodes

19 19 Sub-OLT Scheduling The first option for these algorithms may be to apply the above modified DBA scheme into both sections as EPON dynamic scheduling algorithm 1 (E-DSA1)

20 20 Sub-OLT Scheduling (cont.) The future incoming traffic of sub-OLT nodes is predictable The fact permits a sub-OLT to prerequest some bandwidth in its REPORT message for absent packets that will arrive before the next transmission of this sub-OLT This scheme can exempt some packets that were not reported to the OLT by the sub-OLT in E-DSA1 from waiting at least one cycle time before being transmitted (unlike E-DSA1)

21 21 Sub-OLT Scheduling (cont.) In order to implement this algorithm, each sub-OLT must estimate its future incoming traffic from the attached ONUs within some reference time based upon the information in its grant table This estimate must be added to the bandwidth request in the REPORT message of the sub-OLT

22 22 Sub-OLT Scheduling (cont.)

23 23 Sub-OLT Scheduling (cont.) If level-1 and level-2 sections run at different link capacities can be simply converted as follows: where C 1 and C 2 are the link capacities of the level-1 and level-2 sections m and n are the first and last ONU nodes that will transmit in the reference time window

24 24 Sub-OLT Scheduling (cont.) we refer to this improved algorithm that takes into consideration both imminent and predicted bandwidth requests as EPON dynamic scheduling algorithm 2 (E-DSA2)

25 25 Priority Queueing expedited forwarding (EF) class provides the highest priority for strict delay sensitive services such as voice assured forwarding (AF) class provides a lower, i.e., medium, priority level for services of nondelay sensitive nature, but still requiring guaranteed bandwidth such as video applications best effort (BE) class provides the lowest priority for delay tolerable services such as web browsing, e-mail, and file transfer

26 26 Priority Queueing (cont.) A two-stage queueing-based priority scheduler It ’ s can prevent higher priority queues from unreasonably monopolizing the granted transmission window

27 27 Priority Queueing (cont.) (a) Intra-ONU scheduling (E-DSA1)

28 28 Priority Queueing (cont.) (b) Intrasub-OLT scheduling (E-DSA1)

29 29 Priority Queueing (cont.) (c) Intrasub-OLT scheduling (E-DSA2)

30 30 Performance Evaluation Network Model P-EPON with one OLT 8 sub-OLT nodes 32 ONU nodes Traffic is distributed as follows : 20% from the local users of the sub-OLT nodes 80% from the end users of ONU nodes

31 31 Performance Evaluation (cont.)

32 32 Performance Evaluation (cont.) (a) ONU ’ s average packet delay (b) Sub-OLT average packet delay (access distance 20 km)

33 33 Conclusion Proposed a novel two-stage enhancement, penetrated-EPON (P-EPON) This scheme utilizes an intermediate state to help increase universality and boost distance and coverage within the access domains

34 34 References Shami, A.; Bai, X.; Ghani, N.; Assi, C.M.; Mouftah, H.T., “ QoS Control Schemes for Two-Stage Ethernet Passive Optical Access Networks ”, IEEE Journal on Selected Areas in Communications, Vol. 23, No. 8, pp. 1467- 1478, Aug 2005 C. Assi, Y. Ye, S. Dixit, and M. A. Ali, “ Dynamic bandwidth allocation for quality-of-service over Ethernet PONs, ” IEEE J. Sel. Areas Commun., vol. 21, no. 9, pp. 1467 – 1477, Nov. 2003

35 35 Pseudocode for E-DSA-1 DBA scheduler

36 36 Pseudocode for E-DSA-2 DBA scheduler


Download ppt "1 QoS Control Schemes for Two-Stage Ethernet Passive Optical Access Networks Speaker : Hsuan-Ming Chen Advisor : Ho-Ting Wu Date : 2005.10.18."

Similar presentations


Ads by Google