The two new bandwidth allocation algorithms supporting QoS [1] proposed an interpolated polling algorithm that uses a variable authorization period and is therefore not suitable for real-time services that are sensitive to delay and delay jitter. In order to solve this kind of problem, a fixed slot allocation (Fixed Slot AllocaTIon, FSA) algorithm is mentioned in [2]. However, the disadvantage of fixed slot allocation is: most optical network units (OpTIcal Network) at light load Unit, ONU) has a significant waste of bandwidth, and the upstream channel utilization rate is very low. However, under heavy load, bandwidth sharing cannot be formed between ONUs, which has a negative impact on the delay of high-priority services and queues of various service priorities. In view of this, this paper proposes two new EPON upstream channel dynamic bandwidth allocation algorithms that support multi-service quality of service (QoS): single-level authorized hybrid Transmission algorithm (Single-level Grant Merged Transmission AllocaTIon, SGMTA) and multi-level authorization separation transmission algorithm (MulTIple-level Grant Separated Transmission Allocation, MGSTA).
1.1 Frame structure Considering the compatibility of EPON with the Ethernet frame structure defined by IEEE802.3, the two algorithms designed involve three types of frames: data frame (DATA), authorization frame (GRANT) and request frame ( REQUEST). The data frame still uses the frame structure in IEEE802.3 without any changes. For authorization frames and request frames, in order to save the downlink bandwidth, it is decided to use the minimum Ethernet frame of 64 B. The specific format is shown in Figure 1 and Figure 2, respectively. From the figure, Preamble is the preamble and SFD is Delimiter, DA is the destination address, SA is the original address, and Type is the Ethernet type field. Starting from Subtype, the fields of the request frame are defined as follows: Subtype is used to distinguish between data frames, authorization frames and request frames; Timestamp is used to determine the round-trip time value between the optical line terminal (Optical Line Termination, OLT) and each ONU; Req_EF and Req_BE is used by each ONU to request the authorization of the high-priority service EF and the low-priority service BE from the OLT; the Reserved / Data field is reserved as an extension of the multi-priority service or with additional transmission data; FCS is the check field. The Sendtime_BE field of the authorization frame is the sending time of the low-priority service BE of the authorized ONU. Grant_EF and Grant_BE are the transmission windows of the high-priority service EF and low-priority service BE allocated by the OLT to the authorized ONU, respectively.
Figure 2 Authorization frame format
1.2 SGMTA dynamic bandwidth allocation algorithm
The SGMTA algorithm is a multi-priority dynamic bandwidth allocation scheme mainly based on the ONU end design. The steps are as follows:
1) The OLT sends bandwidth authorization to each ONU according to the authorization table. The authorization mainly includes two parts of information: the logo of the authorized ONU and the size of the authorization window. The authorization window here refers to the sum of the size of each priority window requested by the ONU.
2) When the user data arrives, the ONU first caches the data in the corresponding queue according to its priority according to the priority processing strategy of the ONU terminal, and further sets the subpriority for the data in the high priority queue: sub_priority = (t_wait + t_service ) / t_service, where t_wait is the time that the data packet has been waiting in the queue, and t_service is the transmission delay required by the data packet. All data will be sent in a unified manner when the authorization arrives.
3) After the authorization frame of the OLT reaches the ONU, the ONU reads the authorization window and immediately sends data according to the window size. The sending strategy is: first send high-priority EF data, and then send them in order according to their sub-priority. When the high-priority queue is empty or the sent data is equal to the window of the previous round of application, stop sending. The ONU then uses the remaining bandwidth in the authorization to continue sending low-priority services until the authorization window is used up. After sending the data, the ONU sends a request frame with its contents, including the current queue status of each priority and the corresponding Timestamp value.
4) The OLT accepts ONU data and requests while sending authorization to the ONU. The data is handed over to the upper layer for processing, and the request updates the authorization table according to its content. To further increase the chance of sending high-priority services that arrive between the previous round of requests and the current round of authorization, a supplementary window (supply_window) can be added to the authorization window of the current high-priority service as the next round of high-priority services Pre-allocated bandwidth. At the same time, in order to ensure fairness, the OLT sets the corresponding maximum authorization window Wmax for each ONU.
5) After completing this round of polling and receiving each ONU request, the OLT starts a new round of polling based on the updated authorization table. Through analysis, it is found that the SGMTA algorithm inherits the optical line termination (Optical Line Termination, OLT) algorithm to make full use of the upstream channel and avoid the advantages of synchronization and ranging technology. At the same time, it adopts "one authorization" for high-priority EF services. , Priority transmission "," pre-assign additional windows "and" fine molecular priority "and other strategies, so that high-priority services enjoy better delay characteristics in the SGMTA algorithm, and for low-priority BE services, it will also guarantee its The applied transmission bandwidth prevents its sending opportunities from being forcibly occupied by EF services, which in turn leads to the unstable growth of its buffer queue.
1.3 MGSTA dynamic bandwidth allocation algorithm
The MGSTA algorithm is a multi-priority dynamic bandwidth allocation scheme designed mainly on the OLT side. Before clarifying this algorithm, let's analyze the two shortcomings in the SGMTA algorithm: 1) The high-priority services (EF) of each ONU in SGMTA are conducted between the low-priority services (BE) of different ONUs Sent. In this way, when the BE service of each ONU is relatively busy, the entire polling cycle of SGMTA will increase, and the transmission interval of the EF service of each ONU will also increase accordingly. 2) Since the request frame is sent incidentally when the low-priority service is sent, when the OLT starts the next poll, the bandwidth request time reported by the first ONU polled in the last round has been separated by a long time. For a while. The EF service data reached within this period of time has not been applied for in this round of authorization, and its sending opportunity only depends on the additional window allocated by the OLT according to the prediction, and its accuracy is limited.
In order to solve the above problems, the MGSTA dynamic bandwidth allocation scheme is proposed based on the improvement of the SGMTA algorithm. The main changes are as follows:
1) The entire polling cycle is divided into multiple polling sub-cycles according to the priority of the business. The arrangement order of each polling sub-period is in accordance with the corresponding business priority, high-priority services are polled first, and low-priority services are polled later.
2) The authorization mechanism of OLT is further subdivided, and the overall authorization in the SGMTA algorithm is changed to hierarchical authorization for different priorities.
3) Considering the saving of downlink bandwidth, the multi-priority service authorization of the ONU by the OLT is still encapsulated in an authorization frame. After receiving the authorization, each ONU immediately sends high-priority services, and at the same time obtains the sending time of the next low-priority service.
4) An "in-band windowing" mechanism is added to the low-priority sending window to provide high-priority services with multiple sending opportunities in the same polling cycle. At the same time, the request frame is still attached after the last priority service is sent.
2 Simulation results and performance analysis In order to verify the feasibility of SGMTA algorithm and MGSTA algorithm, simulation experiments were carried out. The simulation parameters used in the experiment are set as follows (the simulation parameters of the FSA, SGMTA and MGSTA algorithms are the same): Number of ONUs: 16; Priority settings: High priority: EF, Low priority: BE; Distance from OLT to ONU : 10 ~ 20 km; uplink and downlink rate: 1 Gbps; service sources: Paerto and Poisson; frame length: 64 ~ 1518 B; protection time: 1 μs; ONU maximum allocation window: 15 000 B.
In order to better simulate the data source in the actual EPON, a burst data source with self-similarity and long correlation and a data source with Poisson distribution are used in the simulation. Among them, high-priority services account for 20% of the entire network load, and low-priority services account for 80%. The entire simulation model considers the data queuing delay Tq, transmission delay Ts and link propagation delay Tp.
Fig. 3 has measured the situation that the end-to-end delay d of the high-priority service EF under FSA, SGMTA and MGSTA three algorithms changes with the network load l. The difference is: Figure 3a is measured under the condition that the load of the high-priority service source is constant, and the EF service source in Figure 3b increases as the network load increases, but the ratio with the low-priority service always remains constant. Through comparison, it is found that under the two test conditions, the MGSTA and SGMTA allocation algorithms exhibit smaller delays and smoother delay changes than the FSA algorithm. At full load or overload, the EF delay of FSA is about 2.5 to 3 times that of MGSTA, and 1.2 to 1.5 times that of SGMTA. This is because the MGSTA and SGMTA algorithms respectively adopt mechanisms such as "multi-level authorization, separate transmission", "pre-divided bandwidth, in-band windowing" and "fine molecular priority" as possible for high-priority services. Provide more opportunities to send.
Figure 4 shows the change of the queue size b at the ONU under the FSA, SGMTA and MGSTA algorithms. Figure 4a shows the queue size b of the high-priority services of the two algorithms relative to the FSA algorithm. You can see the change trend and their respective The delay characteristics of the algorithm are basically the same. The timely delivery of high-priority services in the MGSTA algorithm will inevitably result in its smaller queue length and more stable growth and change under the same load. Figure 4b shows the required buffer size at the ONU. It is found that even if the MGSTA algorithm allows high-priority services to seize the transmission bandwidth of low-priority services, the total required buffer space is still less than the SGMTA algorithm. The reasons are: 1) The MGSTA algorithm pre-allocates an additional window for each ONU, so that high-priority services will not exhaust the transmission bandwidth of low-priority services. 2) While allowing high-priority to seize low-priority bandwidth, it also allows low-priority to send with additional windows when there is no high-priority service, making bandwidth allocation more flexible.
(a) Average queue size of high-priority service EF (b) ONU buffer size Figure 4 Queue situation at ONU under FSA, SGMTA and MGSTA algorithms
Design of EPON Uplink Dynamic Bandwidth Allocation Algorithm Supporting QoS
Hp Laptop 14-Dq 14-Fq,Hp 14-Fq Palmrest,Hp 14-Dq Palmrest,Hp 14-Fq Keyboard
S-yuan Electronic Technology Limited , https://www.syuanelectronic.com