KR20010038486A - Structure of Buffer and Queues for Suppling Ethernet QoS and Operating Method thereof - Google Patents

Abstract

PURPOSE: An Ethernet QoS(Quality of Service) support buffer and queue structure and a method for operating thereof are provided to base an operating algorithm of buffers and queues on a framework for QoS based routing in accordance with IEEE 802.1p and IETF RFC2386, in supporting QoS at an Ethernet. CONSTITUTION: A bypass buffer array(40) is provided with buffers that process delay-sensed packets for MAC(Media Access Control) determined by an address resolution. A bypass queue link list(20) is provided for the queuing of the bypass buffer array(40). A priority weight buffer array(50) is provided with buffers corresponding to bandwidths supported according to service classes. A plurality of transmitting queue link lists(30-1-30-m) are provided for the queuing of the priority weight buffer array(50). A buffer timeout scheduler(90) sets a timeout value to manage the buffers of the priority weight buffer array(50). A buffer loss manager(100) processes delay-sensed packets according to a timeout value of the buffer timeout scheduler(90). A plurality of receiving FIFOs(60-1-60-m) store packets received from various ports. A received packet arbitrator and address detector(70) quests for the addresses of the packets received through the FIFOs(60-1-60-m) and queues them to the buffers of the bypass buffer array(40) and the priority weight buffer array(50). A transmitting packet arbitrator(80) transmits the queued packets by weights.

Description

Structure of Buffer and Queues for Suppling Ethernet QoS and Operating Method

The present invention relates to an Ethernet QoS support buffer and queue structure and a method of operating the same. In particular, the present invention relates to a buffer and a queue considering delay, loss, and throughput, which are used as performance indicators of a network service in Ethernet. The present invention relates to an Ethernet QoS support buffer and queue structure implemented and operated to support QoS, and a method of operating the same.

The conventional Ethernet packet processing scheme operates a buffer based on the traffic load and the media simply to handle the Ethernet traffic and manages it depending on the system.

Here, since it operates according to the Best Effort method, which is a characteristic of the Ethernet traffic, it has difficulties such as excessive service waiting time, inability to guarantee bandwidth, and no QoS guarantee for real-time service. In general, the QoS can be divided into a quality of expression and haptic that the user actually feels the quality of service and a quality of use and control that ensures the convenience of the user's service use and control.

In addition, the QoS is variously defined according to standardization organizations such as ITU-T and IETF, and the ITU-T E.800 recommendation defines the overall effect of service performance to determine the satisfaction level of service users. The IETF Recommendation defines the performance indicators of network services that can be adapted to the needs of users according to the type, characteristics, and level of use of the service.

Thus, conventionally adopted technologies include statically assigning or dynamically allocating buffers for each Ethernet port, and differently operating media for each media.

Here, the method of statically allocating buffers for each Ethernet port allocates the number of buffers required by the Ethernet chip to each port supported by the Ethernet chip and queues them in a FIFO method to provide services. That's the way it is.

The method of dynamically allocating buffers for each Ethernet port is to queue the number of buffers required by the Ethernet chip in a FIFO manner according to packets coming from the Ethernet port. Do not.

In addition, the method of operating differently for each media is prioritized according to other media such as 10M Ethernet and 100M Ethernet, and when the queue is queued or transmitted, 100M, which is faster than the media, is scanned in a given order before 10M. (Scan) to service. This is the way Media is used by different systems.

First, a method of operating a statically allocated buffer for each Ethernet port, that is, a static buffer Ethernet packet processing method is illustrated in FIG. 1. Each port in the Port Receive Buffer Array 12 is illustrated. It allocates a certain buffer, and adds it to the transmission link queue list 11 in sequence, thereby performing a service while rotating.

Second, a method of operating a buffer dynamically allocated to each Ethernet port, that is, a method of processing a dynamic buffer Ethernet packet is illustrated in FIG. 2, and the buffers are allocated to the port receiving buffer array 14 according to a reception order. Queuing it and sending it.

Thirdly, the Media Dependent Allocation method allocates a large number of buffers in consideration of speed in the static buffer Ethernet packet processing method.

However, as described above, it is quite difficult to support QoS in the best-in-class Ethernet appropriately for multimedia services, because excessive service latency is expected in a head of line blocking (HOL) situation. In the case of using buffers or using dynamic buffers, the required bandwidth cannot be allocated to users, and QoS cannot be guaranteed in applications requiring real-time services. In addition, there is a problem in that it is impossible to appropriately cope with services classes and privacy of IPv6, which is a next-generation Internet address.

In order to solve the above problems, the present invention implements buffers and queues in Ethernet to support QoS, which is one of the technologies of next-generation networks, that is, buffers and queues, which are the core in supporting QoS in Ethernet. The purpose of this paper is to provide a method of designing and operating an operation algorithm based on IEEE 802.1p and IETF RFC2386 (a framework for QoS based routing in the Internet).

BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a view for explaining a conventional static buffer Ethernet packet processing method.

FIG. 2 is a diagram for explaining a conventional dynamic buffer Ethernet packet processing method. FIG.

3 is a block diagram illustrating an Ethernet Quality of Service (QoS) support buffer and queues structure in accordance with an embodiment of the present invention.

4 is a flowchart illustrating a method of operating an Ethernet QoS support buffer and queue according to an embodiment of the present invention.

Explanation of symbols on the main parts of the drawings

20: By-pass Queues Linked List

30-1 ~ 30-m: Transmit Queues Linked List

40: By-pass Buffer Array

50: Priority Weighted Buffer Array

60-1 to 60-m: Receive First In Fist Out

70: Receive Packet Arbitrator and Address Detector

80: Transmit Packet Arbitrator

90: Buffer Time-out Scheduler

100: Buffer Loss Manager

An Ethernet QoS support buffer and queue structure according to an embodiment of the present invention for achieving the above object includes: a bypass buffer array having buffers for processing delayed sensed packets for a MAC determined by address resolution; A bypass queue link list for queuing buffers in the bypass buffer array; A priority weighted buffer array having buffers applying a priority and corresponding to bandwidths supported according to a service class; A plurality of transmit queue link lists for queuing buffers in the priority weighted buffer array; A buffer timeout scheduler for setting a timeout value for managing buffers in the bypass buffer array and the priority weighted buffer array; A buffer loss manager for processing the delayed detected packets according to a timeout value of the buffer timeout scheduler; A plurality of receive FIFOs for storing packets received from each port; A receive packet arbiter and an address detector for retrieving the addresses of packets received through the receive FIFO and queuing them to buffers in the bypass buffer array and the priority weighted buffer array; And a transmission packet arbiter for transmitting the packets queued by the received packet arbiter and the address detector by weight.

Meanwhile, in the method of operating the Ethernet QoS support buffer and queue structure according to the embodiment of the present invention for achieving the above object, the priority of each address is determined by determining the priority for each port, and the time for each queue. Setting the out value to a default value; Receiving a packet through each port to search for an address to determine whether the packet exists in an address table; Storing the received packet in a buffer corresponding to a service class according to the address and recording a pointer and a destination port of the buffer; And checking the case where the read pointer and the write pointer of the buffer are different, scanning the buffer according to the priority, reading the packet, and transmitting the packet to the destination port. Here, the timeout setting process may include reconfiguring the timeout value by a user request value or a currently used application; Confirming whether a database exists at an address corresponding to the address learning for address resolution; And if the database exists, performing address learning again with a priority according to the address. Alternatively, the timeout setting process may include: performing address learning according to a priority determined for each port when the database does not exist; And dynamically allocating all buffers and cues when the priority does not exist.

Alternatively, the method of operating the Ethernet QoS support buffer and queue structure according to an embodiment of the present invention may further include performing address learning when the searched address does not exist in the address table. .

Alternatively, the method of operating the Ethernet QoS support buffer and queue structure according to the embodiment of the present invention increases the readout pointer of the buffer in response to the reception of the packet and the timeout value until the packet stored in the buffer is transmitted. It characterized in that the process further comprises the management according to.

Alternatively, the method of operating the Ethernet QoS support buffer and queue structure according to an embodiment of the present invention includes the steps of forcibly freeing the buffer after a predetermined time delay in the case of a HOL or a buffer that cannot be freed; Counting the free buffer and notifying when the counted value reaches a threshold for a predetermined time; The method may further include performing traffic management on a port or a host corresponding to the buffer.

The present invention operates a buffer for delay sensitive packet processing by operating a bypass buffer and a queue, and uses an eight-level buffer for loss sensitive packet processing by operating a fryer weighted buffer and a queue. Operating. The buffer timeout scheduler and loss manager manage the buffer for maximum buffer utilization, and the receive packet arbiter and address detector allow priority weighted basic address resolution and packet storage.

In other words, the present invention comprehensively proposes a buffer and queue operation method that considers delay, loss, and throughput used as performance indicators of network services according to the definition of the IETF recommendation. It simultaneously applies dynamic operation methods, and handles processing delays and queuing delays in packet processing to suit high-priority packets, such as delay-detected voice communications and the quality of noninteractive real-time services such as real-time interactive services and video distribution. By designing a bypass buffer array that minimizes and designing a priority weighted buffer array to support throughput based on the amount of bandwidth that can be allocated to a particular application service, it has the proper structure for loss detection packets.

In addition, the present invention includes a buffer timeout scheduler and a buffer loss manager to manage or support the simultaneous application of static and dynamic operation methods, the design of the bypass buffer array, and the design of the priority weighted buffer array. There is a receiving packet arbiter and an address detector that searches for an address according to a source address and a destination address and adds it to the given array of buffers and queues, and sends packets to the corresponding ports in the given array of transmit queues. There is a transport packet arbiter. Hereinafter, an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

As shown in FIG. 3, the Ethernet QoS support buffer and queue structure according to an embodiment of the present invention includes a bypass queue link list 20, a plurality of transmission queue link lists 30-1 to 30-m, Bypass buffer array 40, priority weighted buffer array 50, a plurality of receive FIFOs 60-1 to 60-m, receive packet arbiter and address detector 70, and transmit packet arbiter 80, a buffer timeout scheduler 90, and a buffer loss manager 100. As shown in FIG.

In other words, the Ethernet QoS support buffer and queue structure according to the embodiment of the present invention has a bypass buffer array 40 having buffers for processing delayed detected packets for Media Access Control (MAC) determined by address resolution. And a bypass queue link list 20 for queuing the corresponding bypass buffer array 40, and also having buffers corresponding to bandwidths to which the priority is applied and supported according to a service class. And a plurality of transmit queue link lists 30-1 to 30-m for queuing the priority weighted buffer array 50, and the corresponding bypass buffer array. A buffer timeout scheduler 90 that sets a timeout value for managing buffers in the 40 and priority weighted buffer array 50, and a corresponding buffer timeout schedule. A plurality of receive FIFOs (60-1 to 60-m) having a buffer loss manager (100) for processing delayed packets according to the timeout value of the scheduler (90), and also storing packets received from each port. Search for the addresses of the packets received through the respective reception FIFOs 60-1 to 60-m, and search the addresses in the bypass buffer array 40 and the priority weighted buffer array 50 according to the searched addresses. A receiving packet arbiter and an address detector 70 for queuing to a buffer, and a transmitting packet arbiter 80 for transmitting packets queued by the corresponding receiving packet arbiter and address detector 70 by weight. It is made.

Here, the bypass buffer array 40 is designed to minimize processing delays and queuing delays of packet processing to be suitable for quality of delayed sensed voice communication and non-interactive real-time services such as real-time interactive service and video distribution. Lose. In addition, the bypass buffer array 40 is associated with the bypass queue link list 20 to apply and queue the packet to which the delay minimization of the IETF RFC 1349 is set. In order to manage the buffering of packets, timeout values of the buffer timeout scheduler 90 are set, and corresponding packets are processed according to timeout values given by the buffer loss manager 100 according to the corresponding timeout values.

The priority weighted buffer array 50 is designed to support the throughput according to the amount of bandwidth that can be allocated to a specific application service. In addition, the priority weighted buffer array 50 includes buffers managed by dividing it into eight classes, applying the privacy described in IEEE 802.1p and IPv6, and queuing the buffer corresponding to the bandwidth supported according to the service class. Or the receiving packet arbiter and the address detector 70 to determine this, and the buffer timeout scheduler 90 and the buffer loss manager 100 to manage the buffers. ).

The receiving FIFOs 60-1 to 60-m receive the packets through each port and output the packets to the receiving packet arbiter and the address detector 70.

The received packet arbiter and address detector 70 searches for an address according to a source address and a destination address in a given traffic and adds the address to an array of a given buffer and queue. Queuing the packets in the buffer according to the address, to maintain the priority level in the address table.

The transmit packet arbiter 80 transmits packets from a given array of transmit queues to the corresponding ports, sequentially scanning the packets queued in the transmit queue from priority bypass queues to priority weighted transmit queues. Tate and send.

The buffer timeout scheduler 90 and the buffer loss manager 100 manage or support the buffers in the priority weighted buffer array 50. The buffer timeout scheduler 90 and the buffer loss manager 100 use the buffer timeout value for each buffer. This setting is applied according to the delay, and the buffer that is not free under HOL or heavy traffic is freed to optimize the operation of the buffer.

An operation method of the Ethernet QoS support buffer and queue structure according to an embodiment of the present invention will be described with reference to the flowchart of eh 4 as follows. Here, although two packets having different priorities will be described by way of example, it should be understood that the present invention is not limited thereto.

First, basically, the priority is determined for each port initially, and the priority of each address is determined based on the determined priority value. In addition, the timeout value is set as default for each cue (step S1).

At this time, the manager may reconfigure the set timeout value by values requested by a user or an application currently used, and the database may be configured to a corresponding address during address learning for address resolution. If is present, the address learning can be performed again with the priority according to the value. In addition, in the case of Connection Oriented Traffic such as Transmission Control Protocol (TCP), it may be set so as not to be aged out.

If the database does not exist, address learning may be performed according to a given priority for each port. If the corresponding priority does not exist, all buffers and queues are dynamically allocated and used.

Then, if one packet is called a delayed detection packet and the other packet is called a priority weighted loss detection packet, the receiving FIFO (60-1 to 60-m) of each port identifies the corresponding delay detection packet and loss detection packet. It receives and outputs the received packet arbiter and the address detector 70.

Then, the received packet arbiter and the address detector 70 search for the source address and the destination address in the given traffic (step S2). In this case, if the address does not exist in the address table, the address is transmitted to the CPU and the address is transmitted. The learning can be performed, and of course, the address learning can be performed by default without the CPU learning. In addition, the received packet arbiter and the address detector 70 transmit the delay detection packet to the bypass buffer array 40 according to the learned address, and transmit the buffer pointer and the destination port to the bypass queue link list ( 20). In addition, the loss detection packet is stored in a corresponding buffer in the priority weighted buffer array 50 by a priority determined by address resolution, and a pointer and a destination port of the buffer are stored in the corresponding transmission queue link list 30-. 1 to 30-m) (step S3).

At this time, the read and write queue pointers of the buffer are the same at first, but as the packets are received, the write pointer of the buffer increases so that the read pointer and the write pointer are different from each other, and the bypass buffer array 40 and the priority weighting are different. Packets stored in the buffer array 50 are managed until transmitted according to the timeout value set by the buffer timeout scheduler 90.

Accordingly, the transmission packet arbiter 80 checks the case where the read buffer pointer and the write buffer pointer of the queues are different from each other, so that the bypass queue link list 20 and the transmission queue link list 30-1 to 30-m are checked. The packet buffer pointer is read and the packets stored in the bypass buffer array 40 and the priority weighted buffer array 50 are transmitted to the destination port of the packet until the end of packet (EOP) of the packet (step S4). At this time, the order of transmission and the order of scanning are based on the priority, and the priority of all the scanning processes the bypass buffers in the bypass buffer array 40 by giving priority to the delay detection packet, and then according to the priority weight. Process the priority weighted transmission buffers in the priority weighted buffer array 50. Here, the eighth priority packet is scanned by eight more times than the first priority packet, and the packets stored in the bypass buffer are always scanned before the packet is transmitted by scanning the priority weighted transmission buffer. First scan and send.

On the other hand, since Ethernet traffic is a best-in-class service, there are quite a few cases in which a buffer is not available in the service or a buffer does not exist in a congestion situation due to heavy traffic. 90) and the buffer loss manager 100 after the delay allowed by the buffer timeout scheduler 90 in the case of a HOL or a packet buffer that is not freed by the CPU has elapsed. Forcibly frees the packet buffer, and notifies the manager when the free packet buffer count reaches a given threshold for a certain amount of time for traffic management.

Accordingly, the manager maintains the value and when the limit is reached, traffic management for the corresponding port or host can be performed.

Accordingly, the loss detection packet and the delay detection packet can support QoS required by using different buffers, and manage a buffer that can be appropriately adapted to various network conditions.

As described above, the present invention is designed to manage the buffers and queues, which are key to supporting the next-generation Ethernet QoS, so that the present invention can be efficiently used to support reception orient QoS, which has been a problem in the existing Ethernet network. There is no application for this yet, but IETF is considering RSVP (Resource Reservation Protocol), and in terms of QoS, IETF is standardizing integrated service and differentiated service.

Accordingly, the present invention can be suitably used for differential services and can be applied to integrated services, and the IPv6 priority, which is the next generation Internet address resolution, can be differentially applied to the 8th grade, and also the IEEE 802.1p 8 It can also be used for precedence.

In addition, the present invention provides a real-time interactive service and enlargement requiring access guarantee and adaptability of service control to adapt to a latency time indicating address and name recognition in use and control quality. It can provide quality assurance of video distribution, a type of service, facilitate traffic engineering, policy, and resource management, which indicates resource utilization efficiency, and also hops on presentation and haptic quality. Minimization of the hop level may be supported at the end point.

In addition, the present invention can be applied to the end switch or router because it can implement the traffic management at the end.

Claims (7)

A bypass buffer array having buffers for processing delay sensed packets for a MAC determined by address resolution; A bypass queue link list for queuing buffers in the bypass buffer array; A priority weighted buffer array having buffers applying a priority and corresponding to bandwidths supported according to a service class; A plurality of transmit queue link lists for queuing buffers in the priority weighted buffer array; A buffer timeout scheduler for setting a timeout value for managing buffers in the bypass buffer array and the priority weighted buffer array; A buffer loss manager for processing the delayed detected packets according to a timeout value of the buffer timeout scheduler; A plurality of receive FIFOs for storing packets received from each port; A receive packet arbiter and an address detector for retrieving the addresses of packets received through the receive FIFO and queuing them to buffers in the bypass buffer array and the priority weighted buffer array; And a transmitting packet arbiter for transmitting packets queued by the receiving packet arbiter and an address detector by weight. Determining a priority for each port to determine the priority of each address, and setting a timeout value by default for each queue; Receiving a packet through each port to search for an address to determine whether the packet exists in an address table; Storing the received packet in a buffer corresponding to a service class according to the address and recording a pointer and a destination port of the buffer; Checking if the read pointer and the write pointer of the buffer are different, scanning the buffer according to the priority, reading the packet, and transmitting the packet to the destination port; How to operate. The method of claim 2, The timeout setting process may include reconfiguring the timeout value by a user request value or a currently used application; Confirming whether a database exists at an address corresponding to the address learning for address resolution; And performing address learning again with a priority according to the address when the database exists. The method of claim 3, The timeout setting process may include: performing address learning according to a priority determined for each port when the database does not exist; And dynamically allocating all buffers and queues when the priority does not exist. The method of claim 2, And performing address learning when the searched address does not exist in the address table. The method of claim 2, And increasing the read pointer of the buffer according to the reception of the packet and managing the packet according to the timeout value before transmitting the packet stored in the buffer. How to operate. The method of claim 2, Forcibly freeing the buffer after a predetermined time delay in case of HOL or when there is a buffer that cannot be freed; Counting the free buffer and notifying when the counted value reaches a threshold for a predetermined time; And performing traffic management on the port or host corresponding to the buffer.