KR20050002048A - Dynamic Bandwidth Allocation Scheme for Ethernet Passive Optical Networks - Google Patents

Abstract

PURPOSE: A method of assigning a dynamic band for an EPON(Ethernet Passive Optical Network) is provided to use a traffic descriptor of real-time traffic to suppress an increase of a relative delay of traffic having the highest priority, thereby guaranteeing delay performance for the real-time traffic. CONSTITUTION: Each ONU loads an amount of traffic stored in input buffers of each class of the ONUs into request messages, and transmit the messages to an OLT. The OLT maintains information on how many packets are waiting to be transmitted, and how much RTT(Round Trip Time) takes in each ONU, through the transmitted request messages. The OLT determines maximum transmission bands and transmission times of each ONU, by using a traffic descriptor for real-time traffic of each ONU. The OLT loads information on the maximum transmission bands into a control message, and sends the message to the ONUs to permit data transmissions.

Description

Dynamic Bandwidth Allocation Scheme for Ethernet Passive Optical Networks

The present invention relates to a dynamic band allocation method for efficiently using an uplink transmission band in an Ethernet-based passive optical communication network that must support multiple application services and guaranteeing a delay performance of real-time traffic.

The dynamic band allocation method can be applied in various ways depending on the implementation company. This depends on the performance of the system, which can have a significant impact on efficiency depending on which algorithm is applied. In particular, depending on how efficient the dynamic band allocation scheme is in an Ethernet-based passive optical network having a variable length frame structure, it may or may not effectively cope with line services such as leased lines. In addition, according to this algorithm, only part of the total capacity can be applied as line service, or the whole can be used for line service.

There are several conventional techniques associated with the present invention. The following is a description of six conventional ways in which the OLT makes this determination.

First is the exhaustive service method. There is a disadvantage in that the waiting time of another ONU is increased by serving until the traffic of a specific ONU is depleted.

Second, there is a fixed service method. Although the implementation is simple by dividing the band evenly according to the number of ONUs, there is a disadvantage in that the bandwidth efficiency is bad.

Third is the gated service method. The OLT is responsible for allocating the requested bandwidth in the request message, which provides excellent average packet delay and queue length performance, but increases the cycle time.

Fourth, it is a limited service method. Although it was the most widely applied method to guarantee the cycle time by defining the maximum band to be allocated to a specific ONU, gated service is known to be advantageous in terms of actual performance.

Of the various methods for calculating the band as described above, the present invention proposes a simple and efficient algorithm that compensates for the shortcomings of the limited method and the gated method to guarantee the delay performance of real-time traffic.

The present invention devised to solve the above problems, each traffic to provide an algorithm that guarantees the delay performance of the real-time traffic in an Ethernet-based passive optical network environment that must support multiple application services and efficiently use the limited transmission band upwards In this paper, we present a dynamic band allocation scheme that considers the quality of service. In addition, the present invention is a method for solving the relative delay increase of the real-time traffic, which is the highest priority traffic, by using the traffic descriptor of the real-time traffic when determining the band to be allocated to each ONU.

1 is a conceptual diagram of an EPON in which each ONU has M + 1 priority classes in accordance with the present invention;

* Explanation of symbols for the main parts of the drawings

1.OLT (Optical Line Terminal): Optical line termination device.

2. Splitter / coupler: The optical signal transmitted along the passive optical network is separated by the splitter into multiple optical lines from the bottom, and multiple optical lines from the upward by a single cable To be combined.

3. Single mode optical cable: Optical cable using one transmission frequency in one direction

4.ONU i : Optical network unit ( I ) in the polling order

5. Dynamic Cycle-based Polling: Adaptively determines the band allocated to each ONU and efficiently collects packets in order to use the limited transmission band efficiently.

6. B0: Input buffer for class 0 traffic with the highest delay priority, such as voice traffic or T1 emulation traffic.

7. B1: Input buffer for class 1 traffic with priority after class 0.

8. BM: Input buffer for class M traffic with the lowest priority, such as best-effort traffic.

9. User: Information source to generate various traffic such as voice, video, data

10.TC: Traffic classifier is a device that classifies traffic into ONU from the user into three traffic classes.

All packets input for transmission from the user are classified into M + 1 classes by the traffic classifier of each ONU and stored in each input buffer. The B0 queue in Figure 1 is a buffer for class 0, the highest priority class that stores real-time traffic, B1 stores the next priority class, Class 1 traffic, and finally BM is the class with the lowest priority. Used to store M traffic. When each ONU transmits uplink data, the request message transmits information on the amount of traffic stored in the input buffer for each class of the ONU to the OLT.

Through request messages sent from each ONU, the OLT polls for information on how many packets are waiting in the input buffer for each ONU and how much round trip time (RTT) is available for each ONU. Keep at the table. The OLT uses this polling table and traffic descriptors for the real-time traffic of each ONU to determine the maximum transmission band and transmission time for each ONU. The OLT sends information about the determined maximum transmission band per ONU in a control message and sends the data to the ONU at a given time. This control message is called a grant. Since the OLT broadcasts data to all ONUs in a downward direction, the grant must include the ID (identification) of the destination ONU as well as the size of the band allowed to transmit. ONU upon receiving grant from OLT Transmits its own data up to the transmission allowed band. At the same time, ONU continues to receive new data from users. At the end of the transmit allowed band, the ONU generates its own control message request. ONU The request sent to the OLT tells OLT how much data was in the ONU's class-specific input buffer at the moment it was created.

In the method of determining the maximum transmission band used in the present invention, if the OLT authorizes each ONU to send the entire contents of the buffer in one transmission, ONUs having a high data volume can monopolize the entire band. To avoid this situation, OLT allowed two scheduling schedules during one cycle time. That is, the ONUs are divided into two groups so that scheduling is performed once for each group during one cycle time, so that each group can share half of the entire band corresponding to one cycle time. It is assumed that the length of the queue requested by the i- th ONU for the m- th priority traffic through the request message is Q (i, m) and the value is in 2 byte units. In performing dynamic band allocation, it is a rule to first allocate a band for the highest priority class 0, that is, real-time identity bit rate traffic, and then allocate a band for traffic having a next priority. For real-time identity bitrate traffic, assign the sum of the band Q (i, 0) and the band expected to arrive in one cycle time to the i th ONU. If there is a band remaining for real-time constant bit rate traffic, then a band for class 1 traffic, which is a next priority class, is allocated. In this case, an additional band of Q (i, 1) is allocated to the i th ONU. In this way, as long as the band remains, all bands are allocated in order of priority. In the above algorithm, the variable bit rate traffic can be allocated in various ways when the requested band is smaller than the remaining band. In this algorithm, a method of sequentially allocating the bands requested by the ONU until all the remaining bands are exhausted is applied. In this case, there may be a problem of equity of the ONU which is not allocated a band. However, if you use an algorithm that fairly divides the remaining bands, the bandwidth provided to each ONU is not used efficiently because the current request message in IEEE 802.3ah does not have information about the length of each pending Ethernet frame. Failure may occur.

As described above, the present invention relates to a method of reducing delay time of a real-time service such as a voice service and efficiently using a limited band, and compensates for the disadvantages of the conventionally proposed gated and limited methods. In addition, the present invention ensures the delay performance for the real-time traffic by eliminating the relative delay increase of the highest priority traffic by using the traffic descriptor of the real-time traffic.

Claims (2)

Dynamic band allocation method, which divides ONUs into two groups and performs scheduling once for each group during one cycle time, thereby eliminating bandwidth loss due to scheduling In the case of dynamic bandwidth allocation, a method of guaranteeing real-time traffic delay performance by using an additional traffic descriptor of real-time traffic to allocate an additional band expected to arrive in one cycle time.