WO2015149460A1

WO2015149460A1 - Fiber channel over ethernet flow control method, device and system

Info

Publication number: WO2015149460A1
Application number: PCT/CN2014/083645
Authority: WO
Inventors: 牛克强
Original assignee: 中兴通讯股份有限公司
Priority date: 2014-04-04
Filing date: 2014-08-04
Publication date: 2015-10-08
Also published as: CN104980359A

Abstract

Disclosed are a fiber channel over Ethernet (FCoE) flow control method, device and system, the method comprising: determining whether the flow at an ingress node port of a magnetic array device exceeds a predetermined threshold; if yes, then transmitting a configuration message to a source node device connected to the ingress node port, the configuration message being used for instructing the source node device to adjust the data transmission rate of the applications thereof. The present invention solves the problem in the prior art that the upper layer service in the FCoE network cannot sense link congestion, and improves quality of service of the magnetic array system.

Description

TECHNICAL FIELD The present invention relates to the field of communications, and in particular to a method, device, and system for controlling traffic of an Ethernet Fibre Channel. BACKGROUND Fibre Channel (FC) is the most widely used protocol in a storage area network (SAN), that is, a data transmission based on an FC network in a SAN. However, the number of switches, network interface cards, and cables required to implement the FC network is large, and the cost of these devices is high, which makes the FC network equipment costly, difficult to maintain, and poorly scalable. To solve the above problem, the prior art uses the Fibre Channel over Ethernet (FCoE) protocol to carry the FC protocol on the basis of Ethernet to integrate the SAN and the local area network (LAN). . When most storage arrays in the data center are FC targets, the FCoE Initiator is connected to Lossless Ethernet, and the FC data is transferred to the FC SAN through the FCoE switch. Access the FC target. However, with the development of the FCoE standard, the FCoE Target (FCoE Target) has begun to appear, and the FCoE Enabler can access the FCoE Target directly through the lossless Ethernet. Since the magnetic array device is a shared resource, its front-end port is much smaller than the host. In order to facilitate more host access, a typical S AN network must have an S AN switch, so the typical network of the magnetic array. It is a port in which multiple hosts are connected to one magnetic array or one magnetic array. With the rapid development of network technology, the quality of data transmission of network services is also increasing, such as network voice services and network video services. Both have high requirements for data stream bursts, jitter and other indicators. The SAN service data flow needs to be stored and forwarded through all levels of nodes in the network. During the forwarding process, the traffic model of the network service data flow will change, and indicators such as bursts and jitter will decrease. For example, the video service based on the IP network, the data stream sent from the video source end is stored and forwarded by the nodes along the path, and the burst and jitter indicators are reduced, which exceeds the buffer processing capability of the receiving end, resulting in packet loss, mosaic, etc. Quality issues. 1 is a schematic diagram of a structure of a magnetic array network according to the related art, and FIG. 2 is a schematic diagram of fluctuations of a traffic bandwidth of a magnetic array network according to the related art. As shown in FIG. 1, when two hosts H1 and H2 pass through the same port of the switch. When P3 accesses a port T1 of the array, the rate of the host HI to the switch PI is 10 gigabits per second (Gbps, Referred to as G), the rate of the host H2 to the switch P2 is 10G, and the rate of the switch P3 to the magnetic array T1 is only 10G. Therefore, the networking module must have a large amount of congestion before the switch and the magnetic array, resulting in the application of the host HI and H2. The performance jitter of the program. The traffic bandwidth of the host HI and H2 applications will fluctuate as shown in Figure 2. The related art solution is to implement loss-based Ethernet flow control, based on 802.1Qbb and 802.3bd, to implement priority-based flow control on full-duplex point-to-point links. Current Ethernet can also achieve no packet loss by using PAUSE frame pause, but it will block all traffic on one link, essentially suspending the entire link. Priority-based Flow Control (PFC) allows eight virtual channels to be created on an Ethernet link and an IEEE 802.1P priority level assigned to each virtual channel, allowing separate pauses and restarts Any one of the virtual channels allows traffic of other virtual channels to pass without interruption. This approach enables the network to create a service with no packet loss class for a single virtual link, enabling it to coexist with other traffic types on the same interface. In the above solution of the related art, the buffer space is allocated in the eight queues of the switch port, and eight virtualized channels in the network are formed, and the data stream carries its own channel label (identified by 802.1P). The buffer size allows each queue to have different data caching capabilities. In the event of transient congestion (that is, a device's queue cache consumes faster and exceeds a certain threshold), the device sends back pressure information in the direction in which the data enters. When the upstream device receives the back pressure information, it stops sending according to the back pressure information. Delay sending data, and store the data in the local port buffer; if the buffer consumption of the local port exceeds the threshold, continue to backflush upstream, so the first level back pressure, until the network terminal equipment, thereby eliminating the loss of the network node due to congestion package. It can be seen that, in the foregoing solution of the related art, when the local end detects the congestion, the PAUSE frame is used to notify the opposite end to temporarily stop the transmission request, and the PAUSE frame carries the time length information of the stop request transmission and the IEEE 802. IP priority. The vector information is used to notify the peer end to temporarily stop sending the request carrying the above priority within the above time period.

The advantage of PFC is that it responds quickly to congestion and can flow control in time; the disadvantage is that it tends to cause congestion to spread. In addition, if the PAUSE frame of the PFC is sent to the upstream more frequently, it will have a negative impact on the bandwidth utilization of the link. 3 is a schematic diagram of congestion spreading caused by a PFC according to the related art. As shown in FIG. 3, when congestion occurs on the P7 port of the switch S3, a PAUSE frame is sent along the dotted line in the figure to the upstream device, including the switch S1 and the switcher. S2, which may cause congestion of upstream devices of S1 and S2. It can be seen from the above description that the drawback of the flow control mechanism in the related art is that the PAUSE frame is a flow control belonging to the L2 layer, because in the related art FCoE protocol, the L2 (link link) layer only guarantees the link control, the L2 layer. FCoE stack that cannot be passed to the L3 (network network) layer when a PAUSE frame occurs. As shown in Figure 1, networking, in the switch When PAUSE occurs in the L2 layer of the P3 to the magnetic array T1, the application of the host HI and the host H2 cannot be perceived, and the service is still delivered at a rate of 10G. 4 is a schematic diagram of network congestion in a SAN network according to the related art, and a typical description of congestion phenomenon of the network is shown in FIG. 4. When the network load is small, the throughput basically increases with the load, linearly, and the response time grows slowly; when the load reaches the network capacity, the throughput shows a slow increase, and the response time increases sharply; if the load continues Increase, the router starts to drop packets; when the load exceeds a certain amount, the throughput begins to drop sharply. The congestion control mechanism in the related art includes two strategies of congestion avoidance and congestion control. The former aims to avoid congestion when the network capacity is reached; the latter is the processing after reaching the network capacity. The former is a "preventive" measure to maintain the network's high-throughput, low-latency state to avoid congestion; the latter is a "recovery" measure that allows the network to recover from congestion to enter normal operating conditions. Congestion increases the delay and jitter of the request transmission, which may cause retransmission of the request, resulting in more congestion. At the same time, the effective throughput of the network is reduced, resulting in a decrease in the utilization of network resources. The architecture of the SAN must ensure that the underlying network devices are designed to be properly loaded. Overloaded links may be allowed by some applications, but some applications cannot accept such overloads and timeouts, so FCoE must be considered from top to bottom. Flow control. And if the capacity provided by the magnetic array device fluctuates too much, it will cause some user business interruption. The key indicator for the test SPC-1 test results in the storage industry is the delay, in which the proportion of requests for different load levels falling in each time period is counted. If the jitter is too large, the last curve will cause fluctuations and the test will be invalid. The upper layer services in the FCoE networking of the related technology cannot be aware of the problem of link congestion. Currently, no effective solution has been proposed. SUMMARY OF THE INVENTION The embodiments of the present invention provide a method, a device, and a system for controlling traffic of an Ethernet Fibre Channel, so as to solve at least the problem that the upper layer service in the related art FCoE networking cannot detect the link congestion. According to an embodiment of the present invention, a method for controlling a traffic of an Ethernet Fibre Channel includes: determining whether a traffic of an ingress port of a magnetic array device exceeds a predetermined threshold; and transmitting a configuration message if the determination result is yes. And the source node device connected to the ingress port, where the configuration message is used to instruct the source node device to adjust a data transmission rate of the application. In a case that the number of the source node devices connected to the ingress node is multiple, before determining whether the traffic exceeds a predetermined threshold, the method further includes: determining the ingress node and the multiple source node devices A flow control policy for each connection between the two; according to the flow control policy, the traffic of each connection is processed. Determining a flow control policy for each connection between the ingress node and the plurality of source node devices includes: obtaining the flow control policy according to a preset configuration; and/or according to the priority of each connection and/or Or adjusting the traffic control policy by the load condition of the service on each connection. Adjusting the flow control policy according to the priority of each connection and/or the traffic condition on each connection includes: in the case that the traffic on the connection with the high priority is increased, the priority is The allocated bandwidth of the low-level connection is provided to the high-priority connection to meet the bandwidth requirement of the traffic on the high-priority connection; and/or the traffic condition of each connection is detected within a predetermined time. And adding the bandwidth allocated by the connection in which the traffic is not detected within the predetermined time to the common resource area for use by other connections. Detecting the traffic condition of each connection in the predetermined time includes: saving a receiving time of the latest read/write request of the source node device to the ingress node on each connection; according to the predetermined time The time interval detects the receiving time of each connection; if the interval between the receiving time and the current time is greater than or equal to the predetermined time, determining that no traffic is detected on the corresponding connection within the predetermined time. In the case that the traffic control policy includes the bandwidth weight, the processing, according to the traffic control policy, the traffic of each connection includes: determining, according to the bandwidth weight of each connection, the allocated bandwidth of each connection The traffic of each connection is processed by the bandwidth allocated by each of the connections. Processing the traffic of each connection by using the bandwidth allocated by each connection includes: receiving a read/write request of the source node device; determining a traffic of the connection between the source node device and the ingress node port Whether the allocated bandwidth is reached; if the judgment result is YES, the read/write request is discarded or suspended. In the case that the traffic control policy includes the priority of the traffic type, according to the traffic control policy, processing the traffic of each connection includes: in the case that the bandwidth required by the traffic type with high priority is increased, The bandwidth allocated by the traffic type with the lower priority is allocated to the traffic type with the higher priority; the traffic type of the lower priority is sent to the connection of the traffic type with the lower priority in each connection. Read and write requests. According to another embodiment of the present invention, a flow control device for an Ethernet Fibre Channel is provided, including: a determining module, configured to determine whether a traffic of an ingress port of the magnetic array device exceeds a predetermined threshold; and a sending module, configured to When the determination result is yes, sending a configuration message to the source section connected to the ingress port The point device, where the configuration message is used to instruct the source node device to adjust a data transmission rate of its application. According to another embodiment of the present invention, a flow control system for an Ethernet Fibre Channel is provided, including a source node device and a magnetic array device, and further includes: a magnetic array control module, located in the magnetic array device, configured to Monitoring the traffic of the ingress port of the magnetic array device, and sending a configuration message to the source node control module if the traffic exceeds a predetermined threshold; the source node control module, the Ethernet located at the source node device The Fibre Channel FCoE driver protocol layer is configured to receive the configuration message, and adjust, according to the configuration message, a rate at which an application in the source node device sends data to the magnetic array device. According to the embodiment of the present invention, it is determined whether the traffic of the ingress port of the magnetic array device exceeds a predetermined threshold; if the determination result is yes, the configuration message is sent to the source node device connected to the ingress port, where the configuration message is used. The method of instructing the source node device to adjust the data transmission rate of the application program solves the problem that the upper layer service in the related technology cannot detect the link congestion in the FCoE network, and improves the service quality of the magnetic array system. BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are incorporated in FIG. In the drawings: FIG. 1 is a schematic diagram of a magnetic array networking structure according to the related art; FIG. 2 is a schematic diagram of a flow bandwidth fluctuation of a magnetic array networking according to the related art; FIG. 3 is a schematic diagram of congestion diffusion caused by a PFC according to the related art. 4 is a schematic diagram of network congestion in a SAN network according to the related art; FIG. 5 is a schematic flowchart of a flow control method of an Ethernet Fibre Channel according to an embodiment of the present invention; FIG. 6 is an Ethernet Fibre Channel according to an embodiment of the present invention; FIG. 7 is a schematic structural diagram of a flow control system for an Ethernet Fibre Channel according to an embodiment of the present invention; FIG. 8 is a schematic structural diagram of an FCoE frame structure according to a preferred embodiment of the present invention; FIG. 10 is a system structural diagram of a flow control system according to a preferred embodiment of the present invention; FIG. 11 is a flow chart showing a flow control method of a magnetic array device according to a preferred embodiment of the present invention; FIG. 12 is a flow chart showing IOPS control in a flow control method for a magnetic array device according to a preferred embodiment of the present invention; FIG. 14 is a schematic diagram of a process flow of a magnetic array device receiving host 10 request according to a preferred embodiment of the present invention; FIG. 15a is a schematic diagram of a magnetic array port receiving host request according to a preferred embodiment of the present invention; Figure 15b is a schematic diagram of a magnetic array device sorting all traffic types according to a priority growth direction according to a preferred embodiment of the present invention; Figure 15c is a magnetic array device according to a preferred embodiment of the present invention, sorting according to the growth direction of the magnetic array scheduling priority FIG. 16 is a flow chart showing the flow monitoring process requested by the magnetic array device receiving host 10 according to a preferred embodiment of the present invention; FIG. 17 is a flow chart for adjusting the current weight and default weight of the magnetic array device according to a preferred embodiment of the present invention. schematic diagram. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS It should be noted that the embodiments in the present application and the features in the embodiments may be combined with each other without conflict. The invention will be described in detail below with reference to the drawings in conjunction with the embodiments. Also, the steps illustrated in the flowchart of the figures may be executed in a computer system such as a set of computer executable instructions, although the logical order is illustrated in the flowchart, in some cases, may be different from this The steps shown are performed in the order shown or described. This embodiment provides a flow control method for an Ethernet Fibre Channel. FIG. 5 is a schematic flowchart of a method for controlling a traffic of an Ethernet Fibre Channel according to an embodiment of the present invention. As shown in FIG. 5, the process includes the following steps: S502, determining whether the traffic of the ingress port of the magnetic array device exceeds a predetermined threshold; Step S504, if the determination result is yes, sending a configuration message to the source node device connected to the ingress port, where the configuration message is used to indicate The source node device adjusts the data transmission rate of its application. Through the above steps, when the traffic of the ingress port of the magnetic array device exceeds a predetermined threshold, a configuration message is sent to the source node device that is also in the L3 layer and connected to the ingress port, to indicate that the source node device adjusts its application. The data transmission rate solves the problem that the upper layer service in the related art FCoE networking cannot sense the link congestion, and improves the service quality of the magnetic array system. Preferably, in the case that the number of the source node devices connected to the ingress node is multiple, the process may further include: before step S502, step S501: determining each port between the ingress node and the plurality of source node devices a connected flow control policy; Step S500, processing the traffic of each connection according to the flow control policy. In this manner, a flow control policy is set for each connection, so that the magnetic array device can perform flow control processing on each connected traffic according to the set flow control policy, thereby further improving the effect of avoiding traffic congestion. Preferably, the flow control policy for each connection is obtained in various manners. For example, the flow control policy for determining each connection between the ingress node and the plurality of source node devices in step S501 may be: Set the configuration to obtain a flow control policy; and/or adjust the flow control policy based on the priority of each connection and/or the load of the traffic on each connection. Preferably, adjusting the flow control policy according to the priority of each connection and/or the traffic condition on each connection includes: the bandwidth allocated by the connection with the lower priority in the case of the increase of the traffic on the connection with the higher priority Providing a connection with a higher priority to meet the bandwidth requirement of the traffic on the connection with a higher priority; and/or detecting the traffic condition of each connection within a predetermined time, and allocating the connection for which the traffic is not detected within the predetermined time The bandwidth is added to the common resource area for use by other connections. In this way, a bandwidth-based allocation mechanism based on the priority of the connection is provided, and the bandwidth of the low-priority connection is allocated to the high-priority connection, thereby ensuring the transmission of the service on the high-priority connection; The bandwidth resources of the connection are shared for use by other working connections, thereby improving the utilization efficiency of the bandwidth resources. Preferably, in the above manner, detecting the traffic condition of each connection within a predetermined time may be as follows: saving the receiving time of the latest read/write request of the source node device to the ingress node of each connection; according to the predetermined time The time interval detects the receiving time of each connection; when the interval between the receiving time and the current time is greater than or equal to the predetermined time, it is determined that the traffic is not detected on the corresponding connection within the predetermined time. In the above manner, the recording time of the latest read/write request is recorded, and the traffic condition of each connection is traversed by using a certain interval time to detect the receiving time, thereby avoiding the magnetic array device frequently detecting the traffic on each connection. Additional system resources are consumed, reducing the overhead of the magnetic array device. Preferably, in the case that the traffic control policy includes the bandwidth weight, the step S500 may include: determining, according to the bandwidth weight of each connection, the allocated bandwidth of each connection; processing each connection by using the allocated bandwidth of each connection flow. Preferably, in the above manner, processing the traffic of each connection by using the allocated bandwidth of each connection includes: receiving a read/write request of the source node device; determining whether the traffic of the connection between the source node device and the ingress node reaches the The allocated bandwidth; in the case of a yes result, the read or write request is discarded or suspended. Preferably, in the case that the traffic control policy includes the priority of the traffic type, the step S500 may include: allocating the allocated bandwidth of the traffic type with the lower priority in the case that the required bandwidth of the traffic type with the higher priority is increased. Give the traffic type with high priority; randomly discard the read and write requests corresponding to the traffic type with lower priority on the connection of the traffic type with lower priority in each connection. In this way, the corresponding priority is set for each traffic type, and the traffic types with different priorities are processed separately, thereby ensuring the transmission of the traffic of the high priority. The embodiment further provides a flow control device for the Fibre Channel of the Ethernet, and the device is configured to implement the flow control method of the Ethernet Fibre Channel. The functions of the modules and units involved in the device may be combined with the description and description of the functions corresponding to the file blocking method described above, and will not be further described in this embodiment. FIG. 6 is a schematic structural diagram of a flow control apparatus for an Ethernet Fibre Channel according to an embodiment of the present invention. As shown in FIG. 6, the apparatus includes: a judging module 62 and a sending module 64, wherein the judging module 62 is configured to determine a magnetic array. Whether the traffic of the ingress port of the device exceeds a predetermined threshold; the sending module 64 is coupled to the determining module 62, and configured to send a configuration message to the source node device connected to the ingress port, where the determination result is yes, where the configuration message is Used to instruct the source node device to adjust the data sending rate of its application. Preferably, the apparatus further comprises a determination module 60 configured to determine the traffic of each connection between the ingress node port and the plurality of source node devices if the number of source node devices connected to the ingress port is plural The control module; the processing module 61, coupled to the determining module 60 and the determining module 62, is configured to process the traffic of each connection according to the flow control policy. Preferably, the determining module 60 is further configured to obtain a flow control policy according to a preset configuration; and/or adjust the flow control policy according to a priority of each connection and/or a load condition of the service on each connection. Preferably, the determining module 60 is further configured to provide the bandwidth allocated by the connection with the lower priority to the connection with the higher priority in the case of the increase of the traffic on the connection with the higher priority, so as to satisfy the connection with the higher priority. The bandwidth requirement of the traffic; and/or detecting the traffic condition of each connection within a predetermined time, and adding the allocated bandwidth of the connection that does not detect the traffic within the predetermined time to the common resource area for use by other connections. Preferably, the determining module 60 is further configured to save a receiving time of the latest read/write request of the source node device to the ingress port port on each connection; detecting the receiving time of each connection according to the time interval of the predetermined time; When the interval of the current time is greater than or equal to the predetermined time, it is determined that the traffic is not detected on the corresponding connection within the predetermined time. Preferably, in the case that the traffic control policy includes bandwidth weight, according to the traffic control policy, the processing module

61 can be set to determine the bandwidth allocated for each connection based on the bandwidth weight of each connection; handle the traffic of each connection through the bandwidth allocated by each connection. Preferably, the processing module 61 is further configured to receive a read/write request of the source node device; determine whether the traffic of the connection between the source node device and the ingress port reaches the allocated bandwidth; and if the judgment result is yes, discard Or suspend the read and write request. Preferably, the processing module 61 is further configured to allocate the allocated bandwidth of the low priority traffic type in the case that the traffic control policy includes the priority of the traffic type, and the bandwidth required by the high priority traffic type increases. Give the traffic type with high priority; randomly discard the read and write requests corresponding to the traffic type with lower priority on the connection of the traffic type with lower priority in each connection. The embodiment also provides a flow control system for an Ethernet Fibre Channel. FIG. 7 is a schematic structural diagram of a flow control system for an Ethernet Fibre Channel according to an embodiment of the present invention. As shown in FIG. 7, the system includes a source node device. 72 and magnetic array device 74, further comprising: a magnetic array control module 742 and a source node control module 722. The magnetic array control module 742 is located in the magnetic array device 74, and is configured to monitor the traffic of the ingress port of the magnetic array device 74, and send a configuration message to the source node control module 722 if the traffic exceeds a predetermined threshold; Module 722 is coupled to magnetic array control module 742, located in the FCoE drive protocol layer of source node device 72, configured to receive configuration messages, and to adjust the rate at which applications in the source node device transmit data to the magnetic array device based on the configuration messages. Description and description are made below in conjunction with the preferred embodiments. The preferred embodiment provides a method and system for Ethernet Fibre Channel flow control, which is a precautionary measure before network capacity is achieved, so that the FCoE protocol stack of the L3 (network) layer is effectively controlled according to the processing capability of the target node. The bandwidth and rate of the source node. The preferred embodiment adopts the following technical solution: FIG. 8 is a schematic structural diagram of an FCoE frame structure according to a preferred embodiment of the present invention. As shown in FIG. 8, a bandwidth-related member is added to an FCoE frame structure, for example, in FCoE protocol data. Unit (Protocol Data Unit, The reserved field of the PDU format (FCoE PDU format) is added to the member, where the weight indicates the proportion of bandwidth allocated by the destination node to the source node; the priority is for applications that require higher latency. The priority is adjusted to ensure the execution delay; the queue depth is the queue depth currently being processed by the destination node FCoE protocol stack, and the source node service adjusts the application sending rhythm and flow control according to the actual load condition of the destination node. Among them, the FCoE PDU key fields and their meanings are:

Version: The version number of the FCoE frame;

Encapsulated FC Frame: Encapsulated FC frame; In the Reserved field, the added members in the field include: Default weight: The user configures the weight for a connection configuration on the ingress port; Current weight: The current weight of the connection; Priority: The user configures the priority for a certain connection configuration on the ingress port; Queue depth: The number of requests is being processed by the ingress port; In the preferred embodiment, in order to ensure smooth processing of traffic, it is determined according to the priority. The processing sequence inside the magnetic array first satisfies the high-priority resources, including the central processing unit (CPU) scheduling order, input/output (I/O) processing priority, and adjustment cache (CACHE). The resource then processes the low-priority request and determines that when the bandwidth exceeds the pre-allocated amount, the request is discarded according to the corresponding policy, and the weight of the discard is calculated. As shown in Figure 1, if the hosts HI and H2 are configured as follows, the default weight = 50, the current weight = x, the priority = 0, and the queue depth = x. In order not to waste bandwidth and traffic, the weight is basically increased by a factor. If it is 1.2, the magnetic array configured in Figure 1 provides 6G bandwidth to the hosts HI and H2, respectively. FIG. 9 is a schematic diagram showing fluctuations in traffic bandwidth of a magnetic array network according to an embodiment of the present invention. The traffic bandwidth of the applications of the hosts HI and H2 fluctuates as shown in FIG. 9. As can be seen from Figure 9, the bandwidth and traffic fluctuations are balanced, and the flow control balance can be achieved among applications, switches, and magnetic arrays. In the above manner, when a plurality of hosts are connected to one magnetic array or one magnetic array, the user traffic jitter caused by the LCo (FC) FCoE lacking effective flow control can be solved; Magnetic array devices cannot be used when accepting such overloads, jitters, and timeouts. The adjustment of the current weight and the default weight: Since the resources of the magnetic array are valuable, there will be many hosts connected to one magnetic array at the same time. When the user configures the weight of a certain host, but the host does not have the service to be delivered for a certain period of time, it needs Consider adjusting the current weight of the host to free up resources for the running magnetic array. As shown in Figure 1, if the hosts HI and H2 are configured as follows, the default weight is 50, the current weight is x, the priority is 0, and the queue depth is x. If the HI host does not deliver the service for a certain period of time, you need to Dynamically adjusting the current weight of the HI host to free up bandwidth to the H2 host; adding a current weight calculation device to the magnetic array port T1, setting a flag for each connection of a port, and when the connection has a service, the request is Timestamp write flag, set a timer for a port, traverse all connected timestamps every 30 seconds, when the timestamp value is found to be different from the current time by 30 seconds or more, mark the connection for the first time without service, and so on. When the statistics are counted to the third time, the current weight value of the connection is lowered, and the current weight value of other connections is increased; likewise, when the connection of the current weight reduction is requested, the current weight is restored to the default weight. In order to make the objects, technical solutions and advantages of the present invention more clear, the above preferred embodiments are further described and illustrated below. 10 is a system configuration diagram of a flow control system according to a preferred embodiment of the present invention, as shown in FIG. 10, through a host 92 (corresponding to the source node device 72 described above) and a storage device 94 (corresponding to the above-described magnetic array device 74). The communication between the two implements access to the data; wherein the method of the preferred embodiment is implemented in the host 92 is a host control module 922 (corresponding to the source node control module 722 described above), the host control module 922 is an FCoE on the host relative to the related art. A functional module added by the driver protocol layer, by adding this module enables the host 92 to perform the method of the preferred embodiment. The user can execute the flow control method of the preferred embodiment by the host control module 922 manually or according to a preset policy according to the networking and bandwidth requirements. Performing the method of the preferred embodiment in the magnetic array device is a magnetic array control module 942 (corresponding to the above-mentioned magnetic array control module 742), and the magnetic array control module 942 is an improved module of the original flow control module on the magnetic array device side. . 11 is a schematic flowchart of a flow control method of a magnetic array device according to a preferred embodiment of the present invention. As shown in FIG. 11, the flow includes the following steps: Step S1001: The magnetic array control module is initialized; Step S1002, the magnetic array acquires a preset a data object, and obtaining a flow control policy of the data object according to the preset data object, where the flow control policy may be configured by the user on the magnetic array side or using a default configuration of the magnetic array before the host accesses the storage device; Step S1003: If the user does not pre-read the flow control policy data object, the default data object is used, which is initially configured by the magnetic array, and all the connections on one port of the magnetic array are equally weighted, and the initial priorities are equal; S1004, if the user does not pre-read the flow control policy data object, the data object is a default weight of a connection, that is, the user configures a weight for a connection configuration on the ingress port port; the current weight of a connection is the current connection of the connection. The weight is equal to the default weight in the case; the priority is the priority configured by the user on the ingress port for the configuration of the connection; the queue depth is zero for the initial number of requests processed by the ingress port. If the application requires the magnetic array to provide the required number of read/write I/O operations per second (IOPS) as required, then different priorities can be set, and each application can have different priorities. For high priority applications. The access can be prioritized on the backup side of the storage array; the IOPS corresponding to a connection is monitored as the product of the current weight and the maximum IOPS of the link speed. When the connected IOPS has not yet reached the preset target, the magnetic array device It will continue to adjust the system resources allocated for the access corresponding to the connection, including increasing the CPU usage ratio, adjusting the 10 priority, and adjusting the queue depth and number; until the preset IOPS is reached. 12 is a schematic flow chart of IOPS control in a magnetic array device flow control method according to a preferred embodiment of the present invention. As shown in FIG. 12, the flow includes the following steps: Step S1101, a magnetic array receives a host request; Step S1102, a magnetic array acquisition The preset data object; Step S1103, the magnetic array determines whether the IOPS reaches the upper limit. If the upper limit is reached, the monitoring module performs corresponding processing, and the monitoring module discards or suspends waiting for processing; Step S1104, the magnetic array determines whether the IOPS reaches the upper limit. , does not reach the upper limit, increase the proportion of its CPU, adjust the 10 priority, adjust the queue depth and quantity, so that the IOPS on the connection reaches the preset IOPS as soon as possible. The magnetic array device in the embodiment of the present invention can perform the magnetic array device flow control method according to any embodiment of the present invention. FIG. 13 is a schematic structural diagram of a magnetic array device according to a preferred embodiment of the present invention. As shown in FIG. 13, the magnetic array device may include: a policy configuration module 1201, a service control module 1202, and a policy analysis module 1203. The policy configuration module 1201 is configured to extract a corresponding data object from a preset data object or a default data object, including: a default weight of a connection, that is, a weight configured on the ingress node user configuration for a connection; a connection The current weight is the current weight of the connection, which is equal to the default weight in the initial case; the priority is the priority configured by the user on the ingress node for a connection; the queue depth is the ingress port is being The number of requests is zero in the initial case; the service control module 1202 is configured to perform quality of service control on the connection. The policy analysis module 1203 is configured to analyze and manage the policies on each connection. Preferably, the magnetic array device further includes a policy monitoring module 1204 configured to monitor bandwidth and IOPS. Among them, the monitoring of bandwidth and IOPS is not real-time and synchronous. In order to reduce the performance impact of the monitoring module on the magnetic array device and ensure the availability of the magnetic array device, the magnetic array does not immediately change the weight in real time in monitoring the change of the data object. The notification host and the magnetic array internal processing system to avoid increasing the burden of the magnetic array device; regular monitoring methods can be used. The policy adjustment module 1205 may be configured to adjust the weight according to the policy. The policy sending module 1206 is configured to send the current flow control policy to the host. The magnetic array device of the preferred embodiment can control and dynamically adjust the bandwidth and IOPS of a connection by setting the policy configuration module 1201 and the service control module 1202, etc., and solve the problem that when multiple hosts are connected to one magnetic array, Or a port of a magnetic array, due to L3 (network) FCoE lack of user traffic jitter caused by effective flow control; avoid some applications can not accept this overload, jitter, timeout and can not use magnetic array equipment, saving The system resources of the magnetic array system improve the service quality of the magnetic array system and more accurately meet the user's needs. FIG. 14 is a schematic diagram of a process flow of a magnetic array device receiving host 10 according to a preferred embodiment of the present invention. As shown in FIG. 14, the process includes the following steps: Step S1301: A magnetic array receives a host request; Step S1302, a policy analysis module 1203 Obtaining the relevant configuration; Step S1303, determining whether the flow control upper limit is reached, if yes, entering the service control module 1202; Step S1304, determining whether the flow control upper limit is reached, if not, proceeding to the next step; Step S1305, determining whether it is a high priority, If yes, the service control module 1202 is entered, and the priority is processed according to the high priority; in step S1306, it is determined whether it is a high priority, and if not, it is put into the normal queue and other non-priority queues for fair processing; Step S1307, threshold and priority After the judgment ends, the policy adjustment module 1205 and the policy monitoring module are entered.

1204 processing; Step S1308, all the above processes are completed, and the processing of the host 10 request is started. Priority is used to identify the priority of request transmission. It can be divided into two categories: request carrying priority and magnetic array scheduling priority. The request carrying priority mainly means that the 802.1p priority is handled by the L2 layer. The preferred embodiment provides a method of processing a magnetic array scheduling priority. The priority of the magnetic array scheduling refers to the priority used when requesting the CPU, the cache CACHE resource, and the queue depth in the magnetic array, and is only valid for the current magnetic array device itself. The main meanings of the magnetic array scheduling priority include the following. The magnetic array is a local-priority priority assigned to the request. Each local priority corresponds to a queue. The request with the larger local priority value has priority. The higher the level, the better the scheduling can be achieved. The parameter referenced when requesting discarding, the smaller the priority value, the more preferentially discarded. For a request that the user does not specify a priority, the magnetic array device automatically acquires the priority of the requested L2 layer for the incoming traffic or processes according to the default priority of the magnetic array. 15a to 15c are schematic diagrams showing a priority processing flow of a magnetic array device receiving host 10 request according to a preferred embodiment of the present invention, the flow comprising the following steps: Step S1401, FIG. 15a is a magnetic array port receiving according to a preferred embodiment of the present invention. Schematic diagram of the host request, as shown in FIG. 15a, the magnetic array port receives the host request; step S1402, if the user specifies the priority, sorting according to the traffic type, and sorting the request into three types according to the type, due to a certain traffic type There is only one priority, within the determined traffic type, the priority of the specific packet is not distinguished; Step S1403, FIG. 15b is a schematic diagram of the magnetic array device sorting all traffic types according to the priority growth direction according to a preferred embodiment of the present invention. As shown in FIG. 15b, all traffic types are organized according to the growth direction of the priority; Step S1404, FIG. 15c is a schematic diagram of the magnetic array device sorting all traffic types according to the growth direction of the magnetic array scheduling priority according to a preferred embodiment of the present invention, As shown in Figure 15c, all traffic types are organized according to the growth direction of the magnetic array scheduling priority. Step S1405: If the user-specified magnetic array scheduling priority, the priority is determined within a certain traffic type, the magnetic array resource allocation is performed according to the scheduling in the magnetic array execution request, and the scheduling priority is reordered within the type. Traffic monitoring is to monitor the specification of a certain traffic entering the network, limit it to a reasonable range, or "punish" the excess traffic to process the request according to the preset policy; if a connection is found If the traffic exceeds the standard, traffic monitoring can choose to discard the request or reset the priority of the request. When the magnetic array discards the request, it needs to cooperate with the flow control action of the source to adjust the traffic of the network to a reasonable load state. An efficient combination of packet loss policy and source-side flow control mechanism maximizes network throughput and utilization efficiency and minimizes request drop and latency. Traffic monitoring should avoid the following situation: when the queue discards multiple connections at the same time, multiple connections will enter the congestion avoidance state at the same time. The host will lower the transmission rate and adjust the traffic, and then will appear at a certain time. Traffic peaks. Repeatedly, the network traffic is suddenly large and small, and the network keeps oscillating. Therefore, the detection method used is mainly as follows. The random drop request causes other connections to have a higher transmission speed when a connection request is discarded and deceleration is started. In this way, whenever there is always a connection for faster transmission, the utilization of the line bandwidth is improved. The upper and lower limits are set for each queue according to the preset policy. The request in the queue is processed as follows. When the length of the queue is less than the lower limit, the request is not discarded. When the length of the queue exceeds the upper limit, random discarding begins. request. The longer the queue, the higher the drop probability, but there is a maximum drop probability. Or, the index, the upper limit, the lower limit, and the drop probability when calculating the average queue length are set for the request of different priorities, thereby providing different discarding characteristics for requests of different priorities. A balance needs to be struck between discarding requests and introducing additional delays. On the one hand, flow control can be performed according to a preset policy, and on the other hand, additional delays are introduced to host requests. FIG. 16 is a schematic diagram of a traffic monitoring process requested by a magnetic array device receiving host 10 according to a preferred embodiment of the present invention. As shown in FIG. 16, the process includes the following steps: Step S1501, a magnetic array port receives a host request; Step S1502, according to traffic The type is sorted and monitored according to the time point; Step S1503, in the existing processing mode, the priority of the traffic 3 is the highest, and at the time T2, since the idle bandwidth can be occupied, the traffic 3 occupies the remaining idle traffic, and the bandwidth is increased to 4 Gbps _; In step S1504, the traffic 1 has the lowest priority. At the time T3, since the host of the traffic 3 increases the transmission bandwidth, the traffic 3 will preempt the bandwidth of the traffic 1, the traffic 3 bandwidth is increased to 5 Gb, and the bandwidth of the traffic 1 is reduced to 2 Gb. That is, the bandwidth of the traffic 1 is reduced, but the host of the traffic 1 is not aware. When the transmission bandwidth of the host corresponding to the traffic 1 exceeds 2 Gb, the magnetic array port sends a PAUSE frame to notify the opposite end to temporarily stop sending the request, and the host application cannot perceive the request. Will continue to send at 3Gb rate, which will cause traffic 1 to generate a large number of timeouts or even business interruptions for the host application; Step S1505, if each traffic type corresponds to two connections on the magnetic array side, as shown in Figure 16 for the connection 1 and the connection 2 queue, the flow control policy is preset, for example, the traffic 3 needs to occupy 5 Gb bandwidth at the time T3, the policy monitoring module The bandwidth of traffic 1 and traffic 3 is adjusted at time T3. Step S1506, after the flow control policy is set, when a traffic type exceeds the standard, the multiple connection requests cannot be discarded at the same time to avoid the network from oscillating; And the random discarding request on the connection 2, and the request discarding is performed according to the length of the queue on the connection 1 and the connection 2 and the preset discarding policy restriction, and feedback to the host side, the host side actively reduces the transmission rate, and gradually reduces the discarding amount and Probability to achieve smoothness of the bandwidth curve. Add a policy monitoring module to the magnetic array device to set a flag for each connection of a port. When the connection has a service, write the timestamp of the request to the tag, set a timer for a port, and traverse every 30 seconds. Time stamp of all connections, when the time stamp value is found to be different from the current time by 30 seconds or more, the connection is marked as having no service for the first time, and so on. When the statistics are counted to the third time, the current weight value of the connection is lowered, and the value is increased. The current weight value of other connections; Similarly, when the connection that is lowering the current weight has a request, its current weight is restored to the default weight. 17 is a schematic diagram of a process of adjusting a current weight and a default weight of a magnetic array device according to a preferred embodiment of the present invention. As shown in FIG. 17, the weight value of each connection is counted, for example, the weight of a connection is 30, and 30 are Credit, when each host request is received, the credit value is reduced. When the credit value is restored after the processing is completed, the traffic monitoring module detects every 30 seconds. If the credit on a connection is equal to the default weight, If there is no change, it is considered that there is no host business requirement for the connection, and the credit can be contributed out and put into a common resource area. The other credits use the public credit according to the load. If there is enough credit on the connection, it can be used to process the host request. In summary, the embodiments of the present invention and the preferred embodiment can effectively solve the problem that the upper layer service in the FCoE network cannot be aware of the link congestion. The existing implementation method is that the maintenance personnel manually configure the switch bandwidth to ensure that the service jitter is small. The bandwidth caused by the static configuration of the switch is wasted, and when the networking is complicated, the manual configuration efficiency is low and the load adjustment between all nodes is complicated. Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general-purpose computing device, which can be concentrated on a single computing device or distributed over a network composed of multiple computing devices. Alternatively, they may be implemented by program code executable by the computing device, such that they may be stored in the storage device by the computing device, or they may be separately fabricated into individual integrated circuit modules, or they may be Multiple modules or steps are made into a single integrated circuit module. Thus, the invention is not limited to any specific combination of hardware and software. The above is only the preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes can be made to the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and scope of the present invention are intended to be included within the scope of the present invention. Industrial Applicability As described above, the method, device, and system for controlling the flow rate of the Ethernet Fibre Channel provided by the embodiments of the present invention have the following beneficial effects: The service quality of the magnetic array system is improved.

Claims

Claim

A method for controlling traffic of an Ethernet Fibre Channel, comprising: determining whether a traffic of an ingress port of a magnetic array device exceeds a predetermined threshold;

If the result of the determination is yes, the configuration message is sent to the source node device connected to the ingress port, where the configuration message is used to instruct the source node device to adjust the data transmission rate of the application.

2. The method according to claim 1, wherein, in a case where the number of source node devices connected to the ingress port is plural, before determining whether the traffic exceeds a predetermined threshold, the method further includes :

Determining a flow control policy for each connection between the ingress node and the plurality of source node devices; processing the traffic of each connection according to the flow control policy.

3. The method according to claim 2, wherein the flow control policy for determining each connection between the ingress node and the plurality of source node devices comprises:

Acquiring the flow control policy according to a preset configuration; and/or adjusting the flow control policy according to the priority of each connection and/or the load condition of the service on each connection.

4. The method according to claim 3, wherein adjusting the flow control policy according to the priority of each connection and/or the traffic condition on each connection comprises:

When the traffic on the connection with the high priority is increased, the bandwidth allocated by the connection with the lower priority is provided to the connection with the higher priority to satisfy the traffic on the connection with the higher priority. Bandwidth requirements; and/or

The traffic condition of each connection is detected within a predetermined time, and the bandwidth allocated by the connection in which the traffic is not detected within the predetermined time is added to the common resource area for use by other connections.

5. The method according to claim 4, wherein detecting the traffic condition of each connection within the predetermined time comprises:

And storing, by the source node device, a receiving time of the latest read/write request of the ingress node; Detecting the receiving time of each connection according to the time interval of the predetermined time;

In a case where the interval between the reception time and the current time is greater than or equal to the predetermined time, it is determined that no traffic is detected on the corresponding connection within the predetermined time.

The method according to claim 2, wherein, in the case that the traffic control policy includes a bandwidth weight, according to the traffic control policy, processing the traffic of each connection includes: according to each connection Bandwidth weights, determining the bandwidth allocated by each of the connections; processing the traffic of each connection by the bandwidth allocated by each of the connections.

The method according to claim 6, wherein, by using the allocated bandwidth of each connection, processing the traffic of each connection comprises: receiving a read/write request of the source node device;

Determining whether the traffic of the connection between the source node device and the ingress node port reaches the allocated bandwidth;

In the case where the judgment result is YES, the read/write request is discarded or suspended.

The method according to any one of claims 2 to 7, wherein, in the case that the traffic control policy includes a priority of a traffic type, processing the traffic of each connection according to the traffic control policy The method includes: assigning, according to an increase in bandwidth required by a traffic type with a high priority, a bandwidth allocated by the traffic type with a lower priority to the traffic type with the higher priority;

The read/write request corresponding to the low priority traffic type on the connection of the low priority traffic type is randomly discarded in each connection.

9. A Fibre Channel traffic control device, comprising: a judging module configured to determine whether a traffic of an ingress port of a magnetic array device exceeds a predetermined threshold; a sending module, configured to send in a case where the determination result is yes The configuration message is sent to the source node device connected to the ingress port, where the configuration message is used to instruct the source node device to adjust the data transmission rate of its application.

10. A Fibre Channel over Ethernet traffic control system, including a source node device and a magnetic array device, further comprising: a magnetic array control module, located in the magnetic array device, configured to monitor traffic of an ingress port of the magnetic array device, and send a configuration message to a source node control module if the flow exceeds a predetermined threshold;

The source node control module is located in an Ethernet Fibre Channel FCoE driver protocol layer of the source node device, configured to receive the configuration message, and adjust an application program in the source node device according to the configuration message. The rate at which the magnetic array device transmits data.