What is Link Aggregation Group?
A Link Aggregation Group (LAG) is a basic networking technology that combines several physical network connections (Ethernet ports) into a single logical link. It is also referred to by proprietary designations such port trunking or EtherChannel (Cisco terminology). IEEE 802.3ad, the original standard, or the subsequent IEEE 802.1AX describe the standardization for Link Aggregation.
LAG is a crucial component of contemporary network equipment, such as Core Layer switches and Layer 3 switches (which offer “Link aggregation”).
Advantages and Features of LAG
There are various advantages of configuring links into a LAG:
Enhanced Bandwidth and Resilience: The LAG generates a logical link with a greater aggregate throughput by combining several physical links (referred to as member links). For example, a logical link with a total potential capacity of 4 Gbps is created by joining four 1 Gbps links. This is a low-cost method of gradually increasing backbone speed without going straight to the next, frequently more expensive generation of Ethernet hardware.
STP Integration: LAG generates an aggregation that is recognized as a single logical connection by the Spanning Tree Protocol (STP). STP uses available bandwidth more efficiently and requires fewer convergences since it regards the bundle as a single interface, meaning that if at least one link is still active, convergence is not required.
Configuration Consistency: To ensure uniformity across the bundled links, the majority of configuration operations can be carried out on the resulting virtual interface, known as a Port Channel, rather than on each separate physical port.
Also Read About What is Split MAC Architecture, Functions, and Benefits
LAG Modes and Protocols
It is possible to create link aggregation statically or dynamically.
Unconditional Static Configuration: This technique employs the on mode, which compels the interface to channel unconditionally and does not require the usage of a negotiation protocol (LACP or PAgP). The EtherChannel will not function if the on option is used with dynamic negotiation modes (such as auto, desirable, active, or passive) on the adjacent switch.
Dynamic Negotiation Protocols:
- Cisco’s proprietary Port Aggregation Protocol (PAgP) facilitates the automatic establishment of EtherChannel links. Sending packets every 30 seconds, PAgP controls link additions and failures and verifies that the setup is consistent. Two modes are supported by PAgP: desirable and auto. The desired mode must be used by at least one side in order for a PAgP channel to develop; auto/auto will not do so.
- The IEEE 802.3ad specification includes the non-proprietary Link Aggregation Control Protocol (LACP), which enables it to function across multi-vendor networks. The role of LACP is comparable to that of PAgP. Up to 16 links can be supported by LACP in a channel; up to 8 of those links can be active at once, with the remaining links being put on standby. Both active and passive modes are supported by LACP. At least one side must be in the active state in order for a LACP channel to develop; passive or passive will not do so.
Requirements for LAG
A LAG needs to fulfil specific conditions in order to function properly:
- The devices on each end of the link must have the same LAG/LACP configuration.
- Physical Restrictions: Every member link needs to be of the Ethernet type. They also need to support full duplex and have the same speed.
- Member Count: Generally speaking, a bond needs one to eight members.
- Single Switch Limitation: All physical ports must be located on the same logical switch in the majority of standard implementations (balance-rr, balance-xor, broadcast, and 802.3ad modes). This creates a single point of failure in the event that the physical switch goes offline. (To remedy this, vendors frequently offer proprietary extensions like as Multi-chassis Link Aggregation Group (MC-LAG) or Split Multi-Link Trunking (SMLT), which make several switches appear as a single logical unit.
In computer networking, a Link Aggregation Group works similarly to constructing a multi-lane freeway from multiple single roads. The lanes are the distinct physical linkages, and the highway itself serves as the single logical interface (the LAG), increasing overall capacity. Traffic automatically switches to the other lanes when one lane shuts (redundancy), but all cars connected to a single journey (a data flow) usually have to remain in the same lane (flow consistency).
Also Read About Identity Services Engine Cisco Core Functions and Benefits
LAG Implementation with Wireless LAN Controllers (WLCs)
In order to provide redundancy and load balancing, LAG is frequently used to bundle the distribution system ports on a Wireless LAN Controller (WLC) to the switching infrastructure.
WLC LAG implementations, however, are rather restrictive:
- No Support for Negotiation Protocols: Cisco WLCs are not capable of supporting PAgP or LACP.
- Required Static Configuration: The switch ports linked to the WLC must be set up as an unconditional or always-on EtherChannel using the on mode (e.g., channel-group # mode on) since dynamic protocols are not supported.
- Port Requirement: The LAG has to include every Distributed System Port on the WLC.
- Port Maximum: Eight ports is the most that can be combined in a LAG between a switch and a WLC.
- Reboot Requirement: A system reboot is required in order to enable LAG on the WLC.
Load Distribution (Load Balancing)
A polymorphic algorithm in the hardware forwarding ASIC distributes (or balances) the load among the connections in an EtherChannel/LAG. This approach determines which particular link in the bundle a frame will traverse by looking at fields in the packet headers and applying a mathematical procedure (a hash function). This eliminates message reordering by guaranteeing that every message in a single application flow uses the same link.
In particular, hash-based load balancing is employed for LAG between a WLC and a switch, usually using Layer 4 source and destination ports.
Also Read About What is Dynamic Trunking Protocol DTP Spoofing Overview