Overview MLAG

MLAG is a mechanism to realize cross-device link aggregation, which gathers one device and two other paired devices to form a dual-active system.

As a weak association horizontal virtualization technology, MLAG not only has the advantages of increasing bandwidth, providing link reliability and load sharing, but also has the following advantages:

• Higher reliability: MLAG improves link reliability from board level to device level. For common cross-board link aggregation, if one board fails, the whole aggregation link can still work normally; for MLAG cross-device link aggregation, even if one device fails, the whole aggregation link can still work normally.
• Simplify Networking: MLAG, as a horizontal virtualization technology, logically virtualizes two paired devices connected by dual homing into one device. MLAG provides a two-layer topology without loop and realizes redundant backup, so the access side does not need spanning tree protocol, which greatly simplifies the networking and configuration.
• Independent upgrade: two paired devices can be upgraded separately to ensure that one device can work normally, with little impact on the running business.

The definitions of MLAG related concepts:

• MLAG (Multi-Chassis Link Aggregation Group) domain: consists of two MLAG pairing nodes, forming a dual active system.
• MLAG pairing node: in the MLAG pairing node, there is one node whose role is master and the other node whose role is slave.
• Peer link: the direct connection aggregation link between MLAG pairing nodes, which is used to interact with MLAG protocol packets and transmit part of data traffic.
• Keep alive: keep alive detection between MLAG pairing nodes, which is carried out through L3 link. The function of MLAG is to judge whether the peer link is still alive after the peer link fails.
• PTS (Peer-Link Transport Server): Reliable packet transmission service based on peer link.
• DHD (Double-Homed Device): A device (or server) that is connected to the MLAG dual-active system via dual-homing. DHD itself may be a MLAG dual-active system.
• SD (Single-Homed Device): A device (or server) that is connected to the MLAG dual-active system via single-homing.
• MLAG-Port: Indicates the aggregation group across devices between DHD and MLAG dual-active system.
• MLAG Member Port: Indicates the member port of the aggregation group across devices between DHD and MLAG dual-active system.
• Orphan-Port: It refers to the single-homed access port on the MLAG pairing node, including the single-homed access aggregation group. By saying that the orphan port belongs to MLAG VLAN, it may not belong to MLAG VLAN.
• MLAG-VLAN: The VLAN to which Peer-Link is added, all VLANs of the MLAG ports should be added to MLAG-VLAN
• non-MLAG-VLAN: The VLAN not containing Peer-Link, non-MLAG-VLAN does not contain the MLAG port
• MLAG-VLANIF: VLANIF interface corresponding to MLAG-VLAN
• non-MLAG-VLANIF: VLANIF interface corresponding to non-MLAG-VLA
• Unpaired: After the peer link port, if the keepalive is normal, it is considered that an unpaired fault has occurred, which will cause the MLAG port and MLAG-VLANIF on the slave node to be in the suspend state, making the uplink and downlink traffic be forwarded through the master node.
• Up-Delay: When the MLAG pairing node restarts or the peer link failure recovers, the MLAG port will delay for a period of time before being set to the UP state.
• Auto-Recovery: After a period of time, when the local MLAG node is judged to be completely disconnected from the peer MLAG node (the actual peer MLAG node may fail or still survive), the local MLAG node considers itself as the only node, becomes the master node, and forwards traffic.
• Dual-Master: If both peer link and keepalive fail, slave node judges to be completely disconnected with master node, slave node will be upgraded to master node, while the original master node still exists. At this time, there are two master nodes, namely dual-master. Traffic forwarding may be abnormal in dual-master state.
• Control protocol packet (also known as peering packet): the protocol packets interacted between MLAG nodes, used for node pairing, role election, etc., and sent and received through UDP.
• Data synchronization packet (also called sync packet): that is, the synchronization packet interacted between MLAG nodes, which is used to transmit control information, relay service protocol packet, synchronization service entry, etc. for each business. The sync packet is encapsulated with the PTS header information and is sent and received through TCP.
• Control VLAN: a L3 VLAN (interface VLAN) used to transmit PTS synchronous packets. Only peer link is allowed to join the VLAN, and other ports are not allowed to join the VLAN.
• Suspend: In this article, it refers to shutdown, which usually refers to the shutdown of MLAG aggregation group or MLAG-VLANIF interface.