Using LACP with OLAN
Introduction
The Tellabs OLAN product supports the LACP or Link Aggregation Control protocol for the management of LAGs or Link Aggregation Groups. This application note will define how to configure and administer systems utilizing LACP and LAGs.
Applies To
This Application Note applies to all OLAN systems.
Document Number
ENG-10675
Definitions
- OLAN – Optical LAN
- ONT – Optical Network Terminal
- OLT – Optical Line Terminal
- PON – Passive Optical Networking
What Is Link Aggregation and LACP?
Link Aggregation Groups
Link aggregation groups or LAGs allow you to bond multiple uplinks into a single Link Aggregation group. This has a number of benefits:
- Facility Protection - Allows protection from facility failures (failure of an Ethernet link)
- Equipment Protection - Allows for equipment failures (failure of an ESU or other uplink card)
- Minimize Software Upgrade outages - Spanning LAG Uplinks across both ESUs ensures that during the upgrade that the standby card can upgrade and reboot without affecting service.
- Increased Bandwidth - LAGs allow bonding of multiple links into a single logical data path with a higher aggregate bandwidth. Once a system has filled a 1G port, for example, a LAG can be introduced, enabling a 2nd link to be added and up to 2G of bandwidth to be carried between the two systems. There is no hard limit on the number of links that can be added to the LAG.
LAG Load Balancing
Link Aggregation Groups perform load balancing of the traffic across all the links in the group. To ensure that packets within a given conversation are not re-ordered (this could be bad for UDP traffic, and slow down TCP performance), packets are hashed across the links. The hash takes into account fields within the packet to decide which link a given traffic flow hashes into. The hashing function can operate on many packet fields, such as: source mac, destination mac, source ip, destination ip, destination port, source port, etc. Once the hash fields are selected, this will ensure that a given conversation (designated by source/destination addressing) will always hash to a given link. This solves the packet re-ordering problem but means that sometimes the hash is not random enough for proper load balancing. This is managed by changing the hash algorithm assigned to the LAG. If the load is all being carried by one of the links in the LAG, changing the hash algorithm can often help to re-distribute traffic across the links in a more balanced way. This ensures maximum bandwidth and efficiency of the links.
Link Aggregation Control Protocol
LACP or Link Aggregation Control protocol layers a control protocol over the top of link aggregation group that brings additional benefits. It is defined in IEEE 802.3ad and implements dynamic link aggregation.
- Allows detection of unidirectional link failures - Ethernet link signal is a unidirectional phenomenon and a link can be failed in only one direction. LACP ensures that both directions are working prior to putting an Ethernet link into the forwarding state. If the link identification on the transmit and receive are not the same, it indicates a link has been split or frogged, and the link will not move to a forwarding state.
- Allows detection of miswired links - The LACP protocol will detect if the transmit and receive both belong to the same link. It it detects that a half connection exists due to miswiring or misconnection of optical links, it will prevent the link from going into a forwarding state. It will also detect if a link is not actually a part of this LAG and reject it. The LACP protocol ensures that all links are actually a member of the lag and that the same LAG ID is shared by all members of the LAG. Each port communicates the LAG identity and the port identity. A mismatch of either will cause the LAG to stop forwarding on that port.
- Allows hitless addition or deletion of LAG Members - New links can be added to the LAG without affecting the overall bonded link and without dropping packets. Links can also be removed from the LAG in a hitless manner.
Switches exchange LACP Protocol Data Units back and forth (LACPDUs) to get the status and configuration of links and communicate control signals. LAGs on other vendor systems may be referred to as a Port Channel.
LAGs in Tellabs OLTs are called NET interfaces (or Network Interfaces) and are named NET1 through NET12. There is no special meaning to the name, it just allows assignment of a name to the LAG. Any Uplinks that are of the same speed can be joined into a LAG. The system will prevent links with two speeds being added to a LAG. Up to 12 links can be bonded into the same LAG. In redundant systems it is always recommended that LAGs span both ESU cards.
LAG Configuration
To perform LAG configuration, set the View to LAGs, then select the OLT which you want to configure a LAG on and select Create LAG if you are creating a new LAG. If the LAG already exists, click on it to edit it. Select the configuration tab which will show all the configurable LAG options.
- Interface AID - Select the LAG name you wish to utilize for this LAG; Any number from NET1 through NET12 can be used. This is just a name for the LAG and is just used to distinguish the LAGs.
- User Label - The User Label is an arbitrary name used to label the LAG and give it a more user-friendly name or description.
- Module Type - SFP or XFP allows selection of either the 1G SFP or 10G XFP ports to be placed into the LAG.
- Speed - Shows the speed of the port.
- Hashing Algorithm - The criteria used to hash packets across all ports of the LAG. The following options are supported.
- Source MAC - Traffic is only hashed on the source MAC.
- Destination MAC - Traffic is only hashed on the destination MAC in the packet. In routed networks, this is usually a bad choice as it will always be the next hop router Mac and therefore will always hash all traffic onto a single link.
- Source and Destination MAC - Traffic is hashed based on both the source and destination mac. Hashing on both often improves the randomness of the packet distribution.
- Source IP - Traffic is only hashed on the source IP in the frames.
- Destination IP - Traffic is only hashed on the destination IP in the frames.
- Source and Destination IP - Both the source and destination IP of the frames are utilized to hash the packets onto ports in the LAG. Since both the Source and Destination tend to change a lot between traffic on the ports, this is often a good choice that load balances well.
- Selected Links - Click on Available Links and use the Add button to add Available Links to the LAG or click on a selected link and press the remove button to remove a link from an existing LAG.
Press the Apply button to apply the configuration to the LAG.
LACP Configuration
The LACP is configured on the LACP Configuration Tab of the LAGs view.
The LACP Configuration has the following configurable elements:
- Configure (enable) LACP - Whether to enable LACP on the LAG. Please note it is critical that the setting be the same on both sides of a LAG. The attached switch or router must have the same settings as the OLT or traffic will not flow correctly and issues may exist during failure scenarios.
- Mode - The mode that the OLT takes with LACP. Active participants send LACP Data Units (DUs) to the partner device. Passive mode devices only respond to incoming DUs. At least one member of a LAG must be an active participant. If both are passive, the group will not come up. It allows for both sides to be active.
- Timer - The timer can be set to short or long for the LACP protocol. With long, a LACP PDU is sent every 30 seconds and the LACP timeout is 90 seconds. If the LACP PDU timer is set to short, it will send a PDU every 1 second and timeout in 3 seconds. Setting it to short will cause links to converge must faster, but send more PDUs. To minimize CPU loading and interop issues seen with some switches, it is recommended that the timer be set to long.
LACP Status
LACP Status is found on the LACP Status Tab of the LAGs view. A particular link must be selected to see the status of LACP on that link.
The following status is available from the link:
- LACP Activity - Indicates that LACP messages were sent / received on the link. Typically, LACP Activity is unchecked on one side. It is indicative of the other end not having LACP enabled, or both sides are passive.
- LACP Timeout - Indicates that a LACP Timeout has occurred.
- Aggregation - Indicates the ability to aggregate traffic across the LAG links.
- Synchronization - Indicates that the link is considered synchronized. The ID of the LAG group is consistent end to end and operational key transmited.
- Collecting - The port is active and receiving (collecting) packets from the far side.
- Distributing - The port is active on the transmit side and transmitting packets to the far side.
- Defaulted - The actors receive the machine using the default operational partner information that was configured. If the default is not checked, it has received information in a LACP PDU that it is using for its calculations. A working link is typically not defaulted.
- Expired State - Indicates the actor or partner is in an expired state and has timed out.