Huawei CloudEngine Series Switches-What is M-LAG

[Music] in a live network horizontal virtualization technology and virtualized to physical devices the backup mode into one logical device as one node on the network this technology simplifies network management and configuration and approved that work reliability CSS is most frequently used on live networks to realize horizontal virtualization however since CSS is configured with strict synchronization mechanism all entries must be completely synchronized between master and backup switches the master switch controls and manages the entire CSS and serves as the only control panel when the CSS splits services are interrupted moreover CSS upgrades must be performed simultaneously while master and backup switches resulting in lengthy service interruption similar problems significantly reduce CSS reliability to solve these problems unlike another horizontal virtualization technology emerges compared with CSS and lag enables two devices to present the same status only during link aggregation negotiation furthermore different from CSS where all information needs to be synchronized between master and backup switches and lag only requires synchronization of interface any information when an EM lock breaks the two devices can operate independently which dramatically improves control panel reliability upgrades for master and backup devices in an EM lag do not need to be synchronized ensuring service traffic forwarding during an upgrade in an EM lag a server can connects to a network through link aggregation of switch a and switch be when the aggregation succeeds the two switches will negotiate the master and backup relationship when an M lag is running normally the two devices is synchronized information with each other through a peer link in real time and lag fault detection depends on the heartbeat packets sent periodically over the interconnected link between two switches on the network side [Music] now we know the technical background of em later let's move to working principle of Emily when an EM lag is running normally her destination load balancing enables em lag master and backup switches to forward known unicast traffic sent from an EM lag member interface to a network side device simultaneously similarly per destination load balancing also enables M lag master and backup switches to forward known unicast flows sent from a network side device to an M lag member interface simultaneously now which forwarding mode is used for multicast broadcast and unknown unicast traffic inside of an M Lac in the scenario where multicast broadcast or unknown unicast flows sent from my memory interfaces to network side devices flood between devices M lag unidirectional isolation is configured to avoid potential loops and prevent traffic forwarding to em like member interfaces through the PA link similarly in the scenario where the multicast broadcast or unknown unicast traffic sent from the network side devices to em like member interfaces flood between devices um like unidirectional isolation is also configured to avoid potential loops and prevent traffic forwarding to em like member interfaces to the peer link how does an EM lag work in full scenarios let's see the working conditions of an M lag in the scenarios of an M lag member interface fault hey link fault device fault and uplink fault when an M lag member interface fails the traffic sent by a server to a network side device is load balanced and forwarded by a normal link the network side devices do not detect the fault and send traffic to the dual homing devices due to the EM lag member interface fault the traffic is single home to one switch an M lag unidirectional isolation is disabled when the switz with the full TM lag member interface receives traffic from a network to be sent to a server it sends the traffic to the normal M like switch through the peer link and the latter device forwards the traffic to the server when the faulty M log member interface recovers mark entry synchronization of the M lag that is triggered single homing mode returns to dual home mode and load balancing of traffic forwarding is restored once a dual homing switch detects that the peer link is down it immediately initiates dual master detection on the people live link if the switch does not receive keepalive packets from the remote end in a specific period it considers that the remote switch fails if the switch receives keepalive packets from the remote end it considers that the peer link fails when the peer link fails the two switches cannot forward traffic simultaneously otherwise problems such as broadcast storms and MAC address flapping may arise therefore only one switch can forward traffic my backups which shuts down all physical interfaces except the peer link interface stack interface and management interface at this point all traffic is forwarded only by the EM lag master switch when the clear link recovers and the peer link interface goes up the em like initiates a new negotiation after the negotiation the recovery delay is configured for em like member interfaces to make em like interface isolation take effect after protocols on the network side er converged interfaces accept the em lag remember interfaces recover immediately whereas the EM lag member interfaces in the arrow down states on a backup switch will have a recovery delayed normally 2 minutes 5 minutes if the M lag is dual home to an SPL system [Music] when the master device fails the M lag backup device becomes the master and continues forwarding traffic heads if crank is still up the eath trunk on the original master device becomes down and the dual homing topology changes into a single homing topology revert of switchover can be disabled for interfaces on M lag master and backup devices in this case the new master device remains their master state after the original master device recovers an original master device becomes the backup device [Music] when an uplink fails traffic forwarding inside an M lag is not affected but the traffic to the Cee will not be forwarded to switch a to to the interface failure if the 40 interface is a keepalive interface or if the two switches were considered that the remote end fails when performing dual active detection then a dual active condition occurs and traffic sent from the Cee to switch a will be discarded since there is no up and outbound interface we have two solutions to solve this problem one configure the management interface as a keepalive interface to enable monitor link to associate M log member interfaces with the uplink interface when the uplink fails the Associated M log member interfaces will be shut down to prevent traffic loss [Music] now we know that an M lag provides load balancing and backup let's move to M lag application scenarios mm lag can be used to dual home servers or switches to a common Ethernet network or trill the X LAN or IP network when an M lag is dual homed to an Ethernet network and the STP network configuration cannot be easily adjusted the STP network must be interconnected with the M lack the to M lag nodes are virtualized into one logical STP node to connect to the STP Network two methods are available to virtualize the two dual homing devices into one STP node one configure the two devices as the root bridge of the STP network to configure vstp to synchronize STP status between the two devices so that they present the same status perform STP negotiation with other nodes [Music] when an M lag is dual homes to a trill network the mblaq is assigned a virtual nickname manually or through peer link negotiation as the two devices are virtualized into one illogical trill device network side devices consider that they have established trill neighbor relationship with one device and all routes to servers will be directed to the virtual nickname when an M lag is dual homes to a be excellent network the two devices are virtualized into a VT EP then both switch a and switch B use the IP address of the VT EP to establish vehicle and tunnels with other nodes no matter whether the tunnels are manually configured or dynamically established using MP BGP [Music] when an M Lac is dual homes to an IP network the dual homing devices act as the boundary between layer 2 and layer 3 networks and function as a gateway the two devices must present the same gateway IP and MAC addresses to clients therefore vrrp needs to run on switch a and switch B to allow them to have a virtual IP and MAC addresses we shield vrrp packets to enable the two devices to work in dual master mode so that they can provide active active gateways the layer 3 traffic load balancing [Music] on a large-scale network you can deploy multiple M legs at the core layer aggregation layer an access layer to enable load balancing among links on the entire network and ensure link reliability multi-level M lags can be considered the same as a single M lakh as all M lags are connected through the same eath trunk however in the case of multi-level M lakhs you cannot manually configure an STP root bridge if you configure the dual homing devices of an M like as the root bridge other devices cannot work therefore the STP must be deployed for interconnection to synchronize STP status between M like dual homing devices [Music] for more information about em like please visit us on evaluate calm [Music]

Channel: Huawei Documentation Insights

Views: 2,853

Keywords: M-LAG, SDN, ce, huawei, cloudengine, cloud engine, huawei cloudengine 12800, huawei cloudengine 6800, ce12800, ce6800

Id: KTUTc_C-1Ls

Length: 13min 8sec (788 seconds)

Published: Tue Jun 12 2018