Route Switching Link Aggregation Technology (Concept)

With the continuous expansion of the network scale of router link aggregation technologies (Eth-Trunk, Ip-Trunk)


　　, operators put forward higher and higher requirements on the bandwidth and reliability of backbone links. In the traditional technology, the bandwidth is often increased by replacing a high-speed interface board or replacing a device supporting the high-speed interface board, but this solution requires a high cost and is not flexible enough. The link aggregation technology can increase the link bandwidth by bundling multiple physical interfaces into one logical interface without hardware upgrade. While achieving the purpose of increasing bandwidth, link aggregation adopts the mechanism of backup links, which can effectively improve the reliability of links between devices.


　　1. The concept of link aggregation technology (Eth-Trunk, Ip-Trunk)


　　Link aggregation is a method of bundling a group of physical interfaces together as a logical interface to increase bandwidth and reliability. For example, two routers are directly connected through three 100M Ethernet interfaces, and these three Ethernet interfaces are bundled to form an Eth-Trunk logical interface with a bandwidth of 300M. This is the link aggregation technology.


　　1.


　　The logical link formed by the link aggregation group bundling several physical links together is called a link aggregation group (LAG) or Trunk. If these bundled links are all Ethernet links, the aggregation group It is called an Ethernet link aggregation group, abbreviated as Eth-Trunk, and the interface of the aggregation group is called an Eth-Trunk interface; if these bundled links are all POS links, the aggregation group is called a POS link aggregation Group, abbreviated as IP-Trunk. The interface of the aggregation group is called IP-Trunk interface.


　　2. Member interface The interfaces


　　that make up the trunk are called member interfaces. For example, an Eth-Trunk interface can be used as an ordinary Ethernet interface. The only difference between it and an ordinary Ethernet interface is that when forwarding, the Eth-Trunk needs to select one or more interfaces from many member interfaces for forwarding. Therefore, except for some features that must be configured under the physical interface, the Eth-Trunk logical interface can be configured like a normal Ethernet interface. Existing Trunk member interfaces cannot be bundled into members of other Trunks.


　　3. Active interface and inactive


　　interface There are two types of link aggregation: active interface and inactive interface. An interface that forwards data is called an active interface, and an interface that does not forward data is called an inactive interface. A link corresponding to an active interface is called an active link, and a link corresponding to an inactive interface is called an inactive link.


　　In link aggregation, a backup link mechanism is introduced to improve link reliability. The interfaces corresponding to these backup links usually play the role of inactive interfaces. Only when the current active interface fails, the backup interface can be converted from an inactive interface to an active interface.


　　4. The upper threshold for the number


　　of active interfaces is configured in an Eth-Trunk. If the upper threshold for the number of active interfaces is configured, when the number of active interfaces reaches this value, adding member interfaces to the Eth-Trunk will not increase the number of active interfaces on the Eth-Trunk. number.


　　5. Lower threshold of the number of


　　active interfaces The main purpose of setting the lower threshold of the number of active interfaces is to ensure the bandwidth of the Eth-Trunk link. To prevent excessive load on these links due to too few active interfaces, packet loss of transmission data occurs.


　　In an Eth-Trunk, if the lower threshold for the number of active interfaces is configured, when the number of active interfaces is lower than this value, the Eth-Trunk interface status will change to Down, and all member interfaces in the Eth-Trunk will no longer forward data.


　　6. Classification of


　　link aggregation Link aggregation is divided into two types according to whether the link aggregation control protocol is enabled:


　　1) Manual load sharing mode Link aggregation


　　Manual load sharing mode is a most basic link aggregation mode. In this mode, the establishment of an Eth-Trunk, the addition of member interfaces, and which interfaces are active interfaces are completely manually configured without the participation of the Link Aggregation Control Protocol. In this mode, all active interfaces participate in data forwarding and share load traffic, so it is called load sharing mode.


　　In manual load balancing mode, all member interfaces can share data traffic evenly. You can also configure the weight of member interfaces so that some interfaces can share more traffic to achieve uneven traffic sharing.


　　If a faulty link occurs among the active links, the link aggregation group automatically shares the data traffic evenly or by weight among the remaining active links.


　　2) LACP protocol link aggregation


　　LACP (Link Aggregation Control Protocol) protocol link aggregation includes two types:


　　static LACP mode link aggregation


　　In static LACP mode, the establishment of Eth-Trunk and the addition of member interfaces are all manually configured Completed. However, different from link aggregation in manual load balancing mode, LACP protocol packets participate in the selection of active interfaces in this mode. That is to say, after a group of interfaces is added to an Eth-Trunk, which interfaces of these member interfaces are active interfaces and which interfaces are inactive interfaces need to be negotiated through LACP protocol packets.


　　Link Aggregation


　　in Dynamic LACP Mode In dynamic LACP mode, the establishment of Eth-Trunk, the addition of member interfaces, and the selection of active interfaces are completely completed by the LACP protocol through negotiation. This means that on the two directly connected devices with the dynamic LACP protocol enabled, there is no need to create an Eth-Trunk, nor to specify which interfaces are the member interfaces of the aggregation group. The two devices will automatically complete the link aggregation operation through LACP negotiation. .
　　8. Constraints of Trunk Interfaces


　　To treat multiple physical links logically as one logical link, and to be transparent to upper-layer data transmission, certain rules must be followed, mainly the following ones.


　　1) The physical parameters


　　of the physical interface must be the same The number of physical interfaces connected at both ends of the


　　Eth-Trunk link must be the same The rate of the physical interfaces connected at both ends of the Eth-Trunk link must be the same


　　The duplex mode of the physical interfaces connected at both ends of the


　　Eth-Trunk link must be the same as the flow control mode of the physical interfaces connected at both ends of the Eth-Trunk link.


　　2) The order of


　　data must be guaranteed. address, destination MAC address, source network layer address, destination network layer address, and a group of data packets with the same Layer 4 source and destination port numbers. For example, a Telnet or FTP connection between two devices is a data flow.


　　If it is required that the Layer 2 data frames belonging to the same data flow must arrive in sequence, this can be guaranteed when the Eth-Trunk interface is not used, because there is only one physical connection between the two devices. But if multiple physical links are aggregated without taking certain measures, this requirement cannot be guaranteed. Because there are multiple physical links between the two devices, if the first data frame is propagated on the first link and the second data frame is propagated on the second link, it is possible that the second data frame arrives at the peer device before the first data frame.


　　In order to avoid this kind of out-of-order data packets, a data packet forwarding mechanism is introduced when implementing Eth-Trunk to ensure that data frames belonging to the same data stream arrive at the destination in the order in which they are sent. This mechanism distinguishes data flows based on MAC addresses or IP addresses, and sends data frames belonging to the same data flow to the destination through the same physical link.