JP5959011B2 - Packet scheduler, communication network, and packet scheduling method - Google Patents

Description

  The present invention belongs to a packet communication technology, and particularly relates to a packet scheduler, a communication network, and a packet scheduling method.

  In recent years, with the spread of smartphones, mobile communication packets, so-called mobile traffic, has increased rapidly, and large-scale deployment of small cells with smaller cell sizes than conventional macro cells has been considered to accommodate this rapidly increasing traffic. ing. In the deployment of small cells, it is desirable to select and deploy areas with high traffic demand. Therefore, a method for connecting small cell base station devices installed at arbitrary locations to a core network (backhaul network) is desired. . Further, such a connection method is desired to be low-cost when a large number of small cells are deployed. In addition, the traffic accommodated in the small cell is characterized by a large variation in traffic volume because the number of accommodated users is small (see Non-Patent Document 1).

  As a future RAN (Radio Access Network) architecture, a split-RAN architecture in which a base station apparatus is divided into a master station and a slave station and the functions are distributed has been proposed (see Non-Patent Document 2). In this architecture, high-performance processing functions are centrally deployed on the master station side, and the functions deployed on the slave station side (antenna site) are reduced, thereby reducing the cost of deploying base stations. Non-Patent Document 2 proposes a method of dividing the function of an eNB that is a base station apparatus in LTE (Long Term Evolution), for example, between a PHY layer and a MAC layer.

  Hereinafter, a base station apparatus having a high-function processing unit that is centrally deployed is referred to as a parent eNB, and a base station apparatus having a function processing unit that is deployed at an antenna site is referred to as a child eNB. According to Non-Patent Document 2, it is mentioned that by adopting this configuration, it is possible to connect a parent eNB and a child eNB (hereinafter, abbreviated as “between parent and child eNB”) via a packet transfer network.

  According to Non-Patent Document 2, the packet network connecting the parent and child eNBs has a delay constraint due to a random access response waiting time in LTE, a HARQ (Hybrid Automatic Repeat reQuest) processing sequence, and a scheduling grant processing sequence. (Round Trip Time) is about 8 msec and one-way delay is about 4 msec. Usually, such a delay constraint of the order of several milliseconds is severe as a network service. Although packet loss is not specifically mentioned, it is assumed that the RLC (Radio Link Control) layer is deployed on the parent eNB side, and that it is possible to deploy a function that provides a certain level of loss tolerance by its retransmission mechanism. Is done.

  In order to realize low-cost connection between the parent and child eNBs in the Split-RAN as described above, it is conceivable to connect using a public network. By using a public network, it is possible to use an access infrastructure such as an optical fiber that is widely deployed, and it is possible to connect at an arbitrary location. In addition, it is expected to reduce the cost by superimposing traffic with an access service such as FTTH (Fiber to the Home) represented by PON (Passive Optical Network). At that time, services that require a high quality of service (QoS) with a guaranteed bandwidth, such as telephone service and TV service, are provided on the public network, so it is necessary to minimize the impact on those services. .

  In order to control QoS such as delay and packet loss in a packet network, for example, a packet scheduler and a packet scheduling method as disclosed in Patent Document 1 are used in each communication node in a normal network. According to this, packets input to the communication node are classified into a plurality of QoS classes via the packet distribution unit, for example, classified into a high priority class, a priority class, and a low priority class, and high priority class packets are assigned high priority. In the class queue, the priority class packet is sorted and stored in the priority class queue and the low priority class packet is sorted and stored in the low priority class queue. The packet stored in the high priority class queue is output with priority over the packet stored in the priority class queue and the low priority class queue, thereby enabling packet transfer with reduced delay and loss.

  When superimposing the above-mentioned traffic between the parent and child eNBs and the traffic of a high quality service that has been conventionally provided in the public network, there are the following problems. This will be described with reference to the drawings.

  FIG. 1 shows the configuration of the conventional packet scheduler shown in Patent Document 1 together with an example of its operation. Here, a high priority class queue 1, a priority class queue 2, a low priority class queue 3, a packet distribution unit 4, a packet length In the packet scheduler including the notification unit 5, the WFQ (Weighted Fair Queuing) control unit 6, and the SPQ (Strict Priority Queuing) processing unit 7, in order to suppress delay, the traffic between the parent and child eNBs is treated as a high priority class queue 1 This shows the case of storing.

  By adopting this storage method, the traffic between parent and child eNBs can be transferred with low delay. However, when the traffic between parent and child eNBs greatly increases, priority class packets that require bandwidth guarantee are not output, and there is a lot of packet loss. There was a problem of becoming.

  FIG. 2 shows a case where the traffic between the parent and child eNBs is stored in the low priority class queue 3 as a low priority class handling in the conventional packet scheduler described above. In this case, although the high quality service provided in the public network can be maintained, there has been a problem that the delay of the traffic between the parent and child eNB becomes large.

  The present invention has been made in view of such a problem, and the purpose of the present invention is when parent-child eNB traffic in Split-RAN is overlapped with traffic of high-quality service in a public network, between parent-child eNBs in Split-RAN. An object of the present invention is to provide a packet scheduler, a communication network, and a packet scheduling method that suppress the traffic transfer delay as much as possible and do not affect the transfer quality of a high-quality service in the public network as much as possible.

In the present invention, in order to achieve the object,
A high-priority class queue that stores high-priority class packets that have a small delay tolerance and need bandwidth guarantee;
A priority class queue that stores priority class packets that have a moderate delay tolerance but need bandwidth guarantee;
A low-priority class queue that stores low-priority class packets that have a large delay tolerance and no bandwidth guarantee;
A low-delay class queue that stores low-delay class packets that have a small delay tolerance but do not require bandwidth guarantee;
Of the input packets, the high priority class packet is the high priority class queue, the priority class packet is the priority class queue, the low priority class packet is the low priority class queue, and the low delay class packet is the low delay priority class queue. A packet distribution unit for distributing and storing;
Means for outputting packets stored in the low-latency class queue in an arbitrary ratio with respect to packets stored in the high-priority class queue and giving priority to packets stored in the priority class queue and the low-priority class queue;
Means for discarding the output packet of the low-delay class queue when the packet storage state in the priority class queue causes a queue overflow,
Means for outputting packets stored in the low-delay class queue in an arbitrary ratio with respect to packets stored in the high-priority class queue and giving priority to packets stored in the priority class queue and the low-priority class queue; ,
An RR processing unit or a WRR processing unit that alternately reads and outputs packets stored in the high priority class queue and the low delay class queue;
A first SPQ processing unit that reads and outputs a packet stored in the priority class queue in preference to a packet stored in the low priority class queue;
A second SPQ processing unit that outputs a packet output from the RR processing unit or the WRR processing unit in preference to a packet output from the first SPQ processing unit;
When the bandwidth of the priority class packet cannot be guaranteed, the means for discarding the packet stored in the low latency class queue is:
A packet discard unit disposed between the low delay class queue and the RR processing unit or the WRR processing unit, and discarding the output packet of the low delay class queue in response to a packet discard command;
When the queue length in the priority class queue exceeds the maximum length that can be stored in the priority class queue minus the maximum packet length, the packet discard control unit outputs a packet discard command to the packet discard unit. It becomes possible to propose a packet scheduler characterized by.

In order to achieve the above purpose,
Comprising a communication node comprising the scheduler;
We propose a communication network characterized by setting parent-child eNB traffic in Split-RAN as a low-latency class packet in the scheduler.

In order to achieve the above purpose,
A high priority class that has a small delay tolerance and requires a bandwidth guarantee, and a priority class that requires a medium delay tolerance but requires a bandwidth guarantee, and a low delay that has a large delay tolerance and does not require a bandwidth guarantee. Classifying the priority class into low-latency class packets that have a small delay tolerance but do not require bandwidth guarantee, and distribute and store them in corresponding queues,
Outputting the packets stored in the low-latency class queue in an arbitrary ratio with respect to the packets stored in the high-priority class queue and giving priority to the packets stored in the priority class queue and the low-priority class queue ;
If the packet storage state in the priority class queue that produce overflow queue viewing contains a step of discarding the output packet of the low delay class queues,
Outputting the packets stored in the low-delay class queue in an arbitrary ratio with respect to the packets stored in the high-priority class queue and giving priority to the packets stored in the priority-class queue and the low-priority class queue; ,
A step in which the RR processing unit or the WRR processing unit alternately reads and outputs the packets stored in the high priority class queue and the low delay class queue;
A step in which the first SPQ processing unit reads and outputs a packet stored in the priority class queue in preference to a packet stored in the low priority class queue;
The SPQ processing unit includes a step of outputting the packet output from the RR processing unit or the WRR processing unit with priority over the packet output from the first SPQ processing unit,
When the packet storage state in the priority class queue causes a queue overflow, the step of discarding the output packet of the low latency class queue is as follows:
A step of discarding an output packet of the low delay class queue in response to a packet discard instruction, a packet discard unit disposed between the low delay class queue and the RR processing unit or the WRR processing unit;
The packet discard control unit outputs a packet discard command to the packet discard unit when the queue length in the priority class queue exceeds the maximum length that can be stored in the priority class queue minus the maximum packet length. A packet scheduling method characterized by comprising steps is proposed.

  According to the present invention, when the traffic between the parent and child eNBs in the Split-RAN is superimposed on the traffic of the high quality service in the public network, the transfer delay of the traffic between the Split and RAN parent and child eNBs is suppressed as much as possible, and the traffic in the public network is increased. Packet scheduling that does not affect the transfer quality of quality service as much as possible can be performed.

Explanatory drawing which shows the structure of the conventional packet scheduler with an example of the operation | movement. Explanatory drawing which shows the structure of the conventional packet scheduler with the other example of the operation | movement. The block diagram which shows an example of embodiment of the packet scheduler of this invention Explanatory drawing which shows the connection method between parent-child eNB using the packet scheduler of this invention

  Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings, but the present invention is not limited to these embodiments.

  FIG. 3 shows an example of an embodiment of the packet scheduler of the present invention. In the figure, 11 is a high priority class queue, 12 is a priority class queue, 13 is a low priority class queue, 14 is a low delay class queue, 15 Is a packet sorting unit, 16 is an RR (Round-Robin) processing unit, 17a and 17b are SPQ (Strict Priority Queuing) processing units, 18 is a packet discarding unit, and 19 is a packet discarding control unit. Each QoS class queue is a FIFO (First In First Out) queue.

  The high priority class queue 11, the priority class queue 12, the low priority class queue 13, and the low delay class queue 14 are packets corresponding to a plurality of QoS classes to be provided, here, high priority class packets, priority class packets, and low priority class packets. , Low latency class packets are stored respectively.

  Here, the “high priority class packet” is a packet having a small allowable delay and requiring a bandwidth guarantee, and corresponds to a telephone service packet or the like. Further, the “priority class packet” is a packet that has a medium delay tolerance but requires a bandwidth guarantee, and corresponds to a packet of TV service or the like. The “low priority class packet” is a packet that has a large delay tolerance and does not require a bandwidth guarantee, and corresponds to an IP packet or the like in general data communication. The “low-delay class packet” is a packet that has a small delay tolerance but does not require bandwidth guarantee, and corresponds to the traffic between the parent and child eNBs in the above-described Split-RAN.

  The packet distribution unit 15 classifies the input packets into a plurality of QoS classes, here, classifies the packets into a high priority class, a priority class, a low priority class, and a low delay class, and places the high priority class packets in the high priority class queue 11. The priority class packet is sorted and stored in the priority class queue 12, the low priority class packet is sorted in the low priority class queue 13, and the low delay class packet is sorted in the low delay priority class queue 14. For QoS class identification, TOS (Type of Service) value in IPv4 and IPv6 packets, COS (Class of Service) value in VLAN (Virtual Local Area Network), exp bit in MPLS (Multi-Protocol Label Switching), etc. Any identifier may be used.

  The RR processing unit 16 performs processing to read out and output the packets stored in the two queues fairly, that is, alternately. Here, the packets stored in the high priority class queue 11 and the low delay class queue 14 are alternately displayed. Read out and output.

  The SPQ processing units 17a and 17b perform processing to read and output a packet stored in a specific queue with priority over packets stored in other queues, and the SPQ processing unit 17a enters the priority class queue 12. The stored packet is read out and output in preference to the packet stored in the low priority class queue 13, and the SPQ processing unit 17b gives priority to the packet output from the RR processing unit 16 over the packet output from the SPQ processing unit 17a. And output.

  The packet discard unit 18 is arranged between the low delay class queue 14 and the RR processing unit 16, and when the packet discard command is not received, the packet stored in the low delay class queue 14 is output according to the RR process. When the packet is output as it is and a packet discard command is received, the packet is discarded when the packet stored in the low delay class queue 14 is output according to the RR process.

  The packet discard control unit 19 outputs a packet discard instruction according to the packet storage state in the queue in which the QoS class packet requiring bandwidth guarantee is stored, and the queue in which the QoS class packet not requiring bandwidth guarantee is stored. Output packet is discarded. Here, a packet discard command is output to the packet discard unit 18 according to the queue length in the priority class queue 12 in which the priority class packet is stored, and the output packet from the low delay class queue 14 is discarded. Let

  In the above configuration, the packets output from the high priority class queue 11 are alternately output by the RR processing unit 16 and the packets output from the low delay class queue 14 via the packet discard unit 18. The output from the RR processing unit 16 is output by the SPQ processing unit 17b in preference to the packets of the priority class queue 12 and the low priority class queue 13 with lower priority.

  Here, since the high-priority class packet stored in the high-priority class queue 11 and the priority class packet stored in the priority class queue 12 are packets that require bandwidth guarantee, traffic between the Split-RAN parent and child eNB is stored. The control described later is performed so as not to be affected by the packet storage state of the low delay class queue 14.

  Further, a threshold Qth is set in the priority class queue 12, and if the priority class queue length, that is, the total length of all the packets stored in the priority class queue 12 exceeds Qth, that fact is discarded. The control unit 19 is notified. Qth is, for example, a value obtained by subtracting the maximum packet length (the length of the longest packet among the packets exchanged on the network) from the maximum value of the priority class queue length, and whether or not the packet is stored in the queue. When determining, it is set so that the queue of the determination target packet itself does not overflow. Therefore, if the next packet arrives when the priority class queue length exceeds Qth, packet loss may occur due to overflow of the priority class queue.

  Hereinafter, the operation of the present embodiment will be described.

  Since the high priority class packet is RR merged with the low delay class packet via the packet discard unit 18 and is SPQ output, the bandwidth up to half of the output rate is guaranteed and output with low delay.

  If the priority class queue 12 does not exceed Qth, the packet control discard unit 19 does not output a packet discard command to the packet control unit 18. For this reason, the packet control unit 18 does not discard the packet output from the low delay class queue 14 in accordance with the RR process. At that time, if the low-delay class packet has a rate up to half the output rate, the low-delay class packet, that is, the traffic between the parent and child eNBs is output without any packet loss. If the low-delay class packet exceeds half of the output rate, it cannot be stored in the low-delay class queue 14, so that the low-delay class packet causes a packet loss but does not affect the high-priority class packet.

  When the priority class queue exceeds Qth, the packet control discard unit 19 outputs a packet discard command to the packet control unit 18. Therefore, the packet control unit 18 discards the packet output from the low delay class queue 14 according to the RR process. With such control, when there is a queue overflow due to output waiting in the priority class queue 12 due to the packet storage state of the low delay class queue 14, the output packet of the low delay class queue 14 is discarded and becomes a low delay class packet. By eliminating the waiting time, the bandwidth of the priority class packet can be guaranteed.

  In order to prevent the queue of priority class packets from overflowing due to waiting for low delay class packets, it is conceivable to stop the output of the low delay class queue 14 when the priority class queue exceeds Qth. However, in this case, the waiting delay of the low-delay class queue 14 increases, and therefore the increase in transfer delay of the low-delay class packet can be prevented by discarding the low-delay class queue output.

  The merging of the high priority class queue 11 and the low delay class queue 14 is RR control. However, control is performed so as to output at an arbitrary rate, for example, WRR (Weighted Round-Robin) control is performed according to an assumed traffic amount. May be.

  By the above operation, when the traffic between parent and child eNB in Split-RAN is superimposed on the traffic of high quality service in the public network, the transmission delay of Split-RAN parent and child eNB traffic is suppressed as much as possible, and the high quality within the public network is maintained. Packet scheduling that does not affect the transfer quality of service as much as possible can be executed.

  FIG. 4 shows a connection method between parent and child eNBs in the Split-RAN using the packet scheduler of the present invention, and is a communication network composed of the communication nodes 21 provided with the above-described scheduler, here the parent eNB 31 in the public network 20. An example is shown in which a child eNB 32 is connected, traffic between parent and child eNBs in Split-RAN is set as a low delay class packet in the scheduler, and communication between parent and child eNBs 31.32 is performed by IPv6 P2P connection. By using the IPv6 P2P connection, it is possible to provide a connection in which the transmission distance can be suppressed with a short path length in the L3 layer, that is, a transmission delay is suppressed.

  In this example, the IPv6 P2P connection is described as an example, but a connection using PPPoE or L2TP may be used. Further, any connection method such as loopback communication in the PON method or connection between bases by VLAN in the Ethernet (registered trademark) leased line service may be used.

  With the above operation, in the connection between parent and child eNBs in Split-RAN, it is possible to realize connection between any installation locations using the public network at the same time as providing the transfer control method by the packet scheduler described in the embodiment. .

  11: High priority class queue, 12: Priority class queue, 13: Low priority class queue, 14: Low delay class queue, 15: Packet distribution unit, 16: RR (Round-Robin) processing unit, 17a, 17b: SPQ ( Strict Priority Queuing) processing unit, 18: packet discarding unit, 19: packet discarding control unit, 20: public network, 21: communication node, 31: parent eNB, 32: child eNB.

Japanese Patent No. 4444721

Trends in small cell enhancements in LTE advanced, Nakamura, T. et., Al, Communications Magazine, IEEE, Volume 51, Issue 2, February 2013, Pages 98-105 Quantitative Analysis of Split Base Station Processing and Determination of Advantageous Architectures for LTE, Uwe Dotsch, et., Al, Bell Labs Technical Journal, Volume 18, Issue 1, June 2013, Pages 105-128

Claims (6)

A high-priority class queue that stores high-priority class packets that have a small delay tolerance and need bandwidth guarantee;
A priority class queue that stores priority class packets that have a moderate delay tolerance but need bandwidth guarantee;
A low-priority class queue that stores low-priority class packets that have a large delay tolerance and no bandwidth guarantee;
A low-delay class queue that stores low-delay class packets that have a small delay tolerance but do not require bandwidth guarantee;
Of the input packets, the high priority class packet is the high priority class queue, the priority class packet is the priority class queue, the low priority class packet is the low priority class queue, and the low delay class packet is the low delay priority class queue. A packet distribution unit for distributing and storing;
Means for outputting packets stored in the low-latency class queue in an arbitrary ratio with respect to packets stored in the high-priority class queue and giving priority to packets stored in the priority class queue and the low-priority class queue;
Means for discarding the output packet of the low-delay class queue when the packet storage state in the priority class queue causes a queue overflow ,
Means for outputting packets stored in the low-delay class queue in an arbitrary ratio with respect to packets stored in the high-priority class queue and giving priority to packets stored in the priority class queue and the low-priority class queue; ,
An RR processing unit or a WRR processing unit that alternately reads and outputs packets stored in the high priority class queue and the low delay class queue;
A first SPQ processing unit that reads and outputs a packet stored in the priority class queue in preference to a packet stored in the low priority class queue;
A second SPQ processing unit that outputs a packet output from the RR processing unit or the WRR processing unit in preference to a packet output from the first SPQ processing unit;
When the bandwidth of the priority class packet cannot be guaranteed, the means for discarding the packet stored in the low latency class queue is:
A packet discard unit disposed between the low delay class queue and the RR processing unit or the WRR processing unit, and discarding the output packet of the low delay class queue in response to a packet discard command;
When the queue length in the priority class queue exceeds the maximum length that can be stored in the priority class queue minus the maximum packet length, the packet discard control unit outputs a packet discard command to the packet discard unit. packet scheduler, characterized by comprising. Comprising a communication node comprising the scheduler according to claim 1 ;
A communication network characterized in that parent-child eNB traffic in Split-RAN is set as a low-latency class packet in the scheduler. The communication network according to claim 2 , wherein communication between a parent and a child eNB in Split-RAN is realized by IPv6 P2P connection. The communication network according to claim 2 , wherein communication between the parent and child eNBs in Split-RAN is realized by PPPoE connection. The communication network according to claim 2 , wherein communication between a parent and a child eNB in Split-RAN is realized by L2TP connection. A high priority class that has a small delay tolerance and requires a bandwidth guarantee, and a priority class that requires a medium delay tolerance but requires a bandwidth guarantee, and a low delay that has a large delay tolerance and does not require a bandwidth guarantee. Classifying the priority class into low-latency class packets that have a small delay tolerance but do not require bandwidth guarantee, and distribute and store them in corresponding queues,
Outputting the packets stored in the low-latency class queue in an arbitrary ratio with respect to the packets stored in the high-priority class queue and giving priority to the packets stored in the priority class queue and the low-priority class queue ;
If the packet storage state in the priority class queue that produce overflow queue viewing contains a step of discarding the output packet of the low delay class queues,
Outputting the packets stored in the low-delay class queue in an arbitrary ratio with respect to the packets stored in the high-priority class queue and giving priority to the packets stored in the priority-class queue and the low-priority class queue; ,
A step in which the RR processing unit or the WRR processing unit alternately reads and outputs the packets stored in the high priority class queue and the low delay class queue;
A step in which the first SPQ processing unit reads and outputs a packet stored in the priority class queue in preference to a packet stored in the low priority class queue;
The SPQ processing unit includes a step of outputting the packet output from the RR processing unit or the WRR processing unit with priority over the packet output from the first SPQ processing unit,
When the packet storage state in the priority class queue causes a queue overflow, the step of discarding the output packet of the low latency class queue is as follows:
A step of discarding an output packet of the low delay class queue in response to a packet discard instruction, a packet discard unit disposed between the low delay class queue and the RR processing unit or the WRR processing unit;
The packet discard control unit outputs a packet discard command to the packet discard unit when the queue length in the priority class queue exceeds the maximum length that can be stored in the priority class queue minus the maximum packet length. A packet scheduling method comprising the steps of :

Images