Link aggregation

from Wikipedia, the free encyclopedia

Link Aggregation ( short: LA) designated as part of the network technology , various methods for combining multiple physical LAN - interfaces to a logical channel for the purpose of increasing the data throughput and reliability over a simple network interface. The Link Aggregation Control Protocol (LACP) is a widespread and manufacturer-independent implementation, which is standardized within the framework of the IEEE as IEEE 802.3ad, since 2008 referred to as IEEE 802.1AX.

History and names

The first implementations of this technology come from Kalpana (now Cisco ) and originally served only to increase the data throughput between two Ethernet - Switches , which were invented also by Kalpana. Today's implementations can also connect servers and other systems such as network attached storage (NAS) to a network structure using link aggregation with higher throughput.

Depending on the manufacturer or context, various expressions are used as synonyms for link aggregation for the bundling of Ethernet interfaces :

  • Link Aggregation (LA) and Link Aggregation Control Protocol (LACP) as a standardized term at IEEE
  • Bundling as a German term for Ethernet channel bundling
  • Bonding in the Linux environment
  • Etherchannel at Cisco
  • Port aggregation at Hewlett-Packard
  • Trunking at Brocade, Sun Microsystems and a few other manufacturers
  • Teaming at Novell Netware and Microsoft Windows

In addition to LACP, there are also proprietary solutions for dynamic bundling, for example the PAgP from Cisco and MESH from HP . There are also static link aggregation methods, all of which are more or less proprietary and therefore cannot be used across manufacturers.

General

A PC is connected to a central switch by means of two Ethernet connections and the link aggregation

In general, several physical Ethernet interfaces, all of which must be designed as full duplex connections, are interconnected to form a logical endpoint connection between two devices, for example a PC with two Ethernet interfaces, which are connected in parallel to two ports on a network switch. LA is usually used at data rates of 1 GBit / s and above with Gigabit Ethernet .

Due to the parallel connections, the total throughput can be higher depending on the number of combined connections. Another advantage is the increased availability of this type of connection. With some methods, one or more physical interfaces can fail without the logical channel being interrupted as long as at least one physical connection still exists. Only the data throughput is reduced according to the missing connections.

Bundling process

  • Round robin: Here all available lines are used alternately one after the other.
  • DA trunking: The elementary interface is selected here based on the modulo of the destination MAC address.
  • SA trunking: The elementary interface is selected here based on the modulo of the source MAC address.
  • SA-DA trunking: The elementary interface is selected here based on the modulo of the source MAC address and the destination MAC address.
  • Adaptive trunking: Here, another interface is only switched on when the first elementary interface is fully utilized.
  • Dynamic trunking is possible in the Standard Link Aggregation Control Protocol (LACP) and with proprietary methods such as PAgP .

Basic implementation

The basic implementation of channel bundling between two Ethernet switches requires surprisingly few changes (compared to a normal switch). Of course, the management interface has to be expanded to define the trunk. Then the way a switch learns its SAT ( Source Address Table - table with the MAC addresses of the sender) is affected, and finally broadcasts and packets that are directed to unknown destination MAC addresses are affected . treated separately from the drink.

  • Management Interface: Menu structures for managing the trunks must be created.
  • Address learning phase: If a switch receives an unknown sender address on a trunk port, it is not automatically assigned to the address stack (SAT) of this port; rather, the switch tries to assign the same network load to all members of the trunk and therefore distributes new sender addresses equally to all distribute ports belonging to the trunk. For example, the port that currently has the fewest entries in its SAT receives the new address. Depending on the implementation, other distribution strategies are also used, see section Bundling procedures .
  • Broadcasts and unknown addresses: In the case of a trunk, these are not sent over all ports, but only over one line per trunk; in general, the port with the lowest port number is selected. Depending on the implementation, other strategies are used to select the port, see section Bundling .

Interestingly, no further measures are necessary to realize a simple implementation. The implementation of the transport mechanisms ( cut through , store and forward etc.) does not require any changes. However, this basic implementation is far from ideal and is actually only suitable for coupling two networks. A disadvantage is that only one port of the trunk is used per MAC address - for a single station it means that it has no advantages - only the network as a whole benefits. Furthermore, it can happen that the most active stations are all randomly assigned to the same port, so that consequently the load is not evenly distributed within the trunk. There are also bundling processes that can connect powerful servers. All of these facts have led to many manufacturers offering different, but also more advanced implementations. As a rule, these are not compatible with each other. These incompatibilities have led the IEEE to address the problem and specify the Link Aggregation Control Protocol (LACP, IEEE 802.3ad).

Web links

Individual evidence

  1. ↑ IEEE P802.3ad Link Aggregation Task Force. Retrieved September 19, 2017 .
  2. ↑ IEEE Std 802.1AX-2008 - IEEE Standard for Local and metropolitan area networks - Link Aggregation. IEEE, 2008, accessed September 19, 2017 .