1. Requisitos para importar alimentos en los Estados Unidos. General. 3
2.9. Regulación aplicable a la importación de alimentos
This section describes the Split MultiLink Trunking (SMLT) feature. The following topics are included:
• "Overview" (page 35)
• "Advantages of SMLT" (page 36) • "How SMLT works" (page 38) • "Inter-Switch Trunks" (page 40)
Split MultiLink Trunking 35
• "CP-Limit and SMLT IST" (page 41)
• "Traffic flow in an SMLT environment" (page 42) • "Single port SMLT" (page 44)
• "SMLT topologies" (page 45)
• "Using MLT-based SMLT with single port SMLT" (page 49) • "SMLT network design considerations" (page 50)
• "SMLT and VRRP backup master" (page 51)
To configure SMLT using Device Manager, see"Configuring SMLT" (page 120).
Overview
Link Aggregation technologies have become popular for improving link bandwidth and to protect against link failures.
SMLT is an extension of link aggregation, which improves the level of Layer 2/Layer 3 resiliency by providing nodal protection in addition to link failure protection and flexible bandwidth scaling. SMLT achieves this by allowing edge switches using link aggregation to dual-home to two SMLT aggregation switches. SMLT is transparent to those attached devices that support link aggregation.
Because SMLT inherently avoids loops due to its superior enhanced link aggregation control protocol, when designing networks using SMLT, it is not necessary to use the IEEE 802.1d/w Spanning Tree protocols to enable loop-free triangle topologies.
With split multilink trunking, two aggregation switches can appear as a single device to edge switches, which are dual-homed to the aggregation switches. The aggregation switches are interconnected using an Inter-Switch Trunk (IST) and can exchange addressing and state information (permitting rapid fault detection and forwarding path modification). Although SMLT is primarily designed for Layer 2, it also provides benefits for Layer 3 networks.
ATTENTION
Layer 2 edge switches must support some form of link aggregation (such as MLT) to allow communications with the SMLT aggregation switches.
Advantages of SMLT
SMLT improves the reliability of Layer 2 networks that operate between edge switches and the network center aggregation switches by providing the following:
• load sharing among all links
• fast failover in case of link failures
• elimination of single point of failure
• fast recovery, in case of nodal failure
• transparent and interoperable solution
• elimination of STP convergence issues
These advantages are described in more detail in the sections that follow.
Single point of failure elimination
SMLT helps eliminate all single points of failure and create multiple paths from all edge switches to the core of the network. In case of failure, SMLT recovers as quickly as possible so that no unused capacity is created. Finally, SMLT provides a transparent and interoperable solution that requires no modification on the part of the majority of existing edge devices.
SMLT compared to Spanning Tree Protocol
Networks that are designed to have edge switches dual-homed to two aggregation switches, and that have VLANs spanning two or more edge switches, experience the following design constraints:
• spanning tree must be used to detect loops
• no load sharing exists over redundant links
• slow network convergence exists in case of failure (30–45 seconds)
Figure 8 "Resilient networks with Spanning Tree Protocol" (page 37)shows a typical aggregator switch configuration dependent upon STP for loop detection.
Split MultiLink Trunking 37
Figure 8
Resilient networks with Spanning Tree Protocol
As shown inFigure 9 "Resilient networks with SMLT" (page 38), with the introduction of SMLT, all dual-homed Layer 2 frame-switched network devices are no longer dependent upon STP for loop detection because a properly designed SMLT network inherently does not have any logical loops.
Figure 9
Resilient networks with SMLT
SMLT solves the Spanning Tree problem by combining two aggregation switches into one “logical” MLT entity, which makes it transparent to any type of edge switch. In the process, it provides quick convergence, while load sharing across all available trunks.
How SMLT works
Figure 10 "8300 switches as SMLT aggregation switches" (page
39)illustrates an SMLT configuration with a pair of 8300 switches (E and F) as aggregation switches. Also included are four separate edge switches (A, B, C, and D). Refer to the following sections for a description of the components shown in this SMLT example:
• "Inter-Switch Trunks" (page 40) • "CP-Limit and SMLT IST" (page 41)
Split MultiLink Trunking 39
Figure 10
8300 switches as SMLT aggregation switches
Other SMLT aggregation switch connections
Figure 10 "8300 switches as SMLT aggregation switches" (page 39)also includes end stations connected to each of the switches.
In this example, a, b1, b2, c1, c2, and d are clients and printers, while e and f can be servers or routers.
Edge switches B and C can use any method for determining a link of their multilink trunk connections to use for forwarding a packet, as long as the same link is used for a given Source/Destination (SA/DA) pair. This is true, regardless of whether or not the DA is known by B or C. SMLT aggregation switches always send traffic directly to an edge switch and only use the IST for traffic that they cannot forward in another more direct way.
The examples that follow explain the process in more detail:
Example 1-Traffic flow from a to b1 or b2 Assuming a and b1/b2 are communicating using Layer 2, traffic flows from A to switch E and is forwarded over the direct link to B. Traffic coming from b1 or b2 to a is sent by B on one of its MLT ports.
B sends traffic from b1 to a on the link to switch E, and traffic from b2 to a on the link to F. In the case of traffic from b1, switch E forwards the traffic directly to switch A, while traffic from b2, which arrived at F, is forwarded across the IST to E and then on to A.
Example 2-Traffic flow from b1/b2 to c1/c2 Traffic from b1/b2 to c1/c2 is always sent by switch B through the MLT to the core. No matter which switch (E or F) it arrives at, traffic is sent directly to C through the local link.
Example 3-Traffic flow from a to d Traffic from a to d (and the reverse) is forwarded across the IST because it is the shortest path. This link is treated purely as a standard link with no account taken of SMLT and the fact that it is also an IST.
Example 4-Traffic flow from f to c1/c2 Traffic from f to c1/c2 is sent directly from F. With return traffic from c1/c2, you can have one active VRRP Master for each IP subnet. The traffic is passed across the IST if switch C sends it through the link to E.
Inter-Switch Trunks
SMLT aggregation switches must be connected with an Inter-Switch Trunk (IST). For example, inFigure 10 "8300 switches as SMLT aggregation switches" (page 39), edge switches B and C are connected to the aggregation switches using multilink trunks split between the two aggregation switches. The implementation of SMLT requires only two SMLT-capable aggregation switches.
Aggregation switches use the IST to:
• Confirm that they are alive and exchange MAC address forwarding tables.
• Carry the SMLT control packets.
• Send traffic between single switches attached to the aggregation switches.
• Serve as a backup if one SMLT link fails.
Because the IST is required for the SMLT, Nortel recommends that you use multiple links on the IST to ensure reliability and high availability. Nortel recommends using Gigabit Ethernet links for IST connectivity to provide enough bandwidth for potential cross traffic.
Split MultiLink Trunking 41
ATTENTION
Nortel recommends that an IST MLT contain at least 2 physical ports.
CP-Limit and SMLT IST
Control packet rate limit (CP-Limit) controls the amount of multicast and broadcast traffic that can be sent to the CPU from a physical port. It protects the CPU from being flooded by traffic from a single, unstable port. The CP-Limit default settings are:
• default state = enabled
• default multicast packets-per-second (pps) value = 15 000
• default broadcast pps value = 10 000
ATTENTION
Nortel recommends setting the multicast packets-per-second value to 6000 pps when you configure SMLT links.
If the actual rate of packets-per-second sent from a port exceeds the defined rate, the port is administratively shut down to protect the CPU from continued bombardment. Disabling IST ports in this way can impair network traffic flow in an SMLT configuration.
To avoid this scenario, the 8300 Series switch automatically disables CP-Limit on all IST port members.
Disabling CP-Limit on IST MLT ports forces another, less-critical port to be disabled if the defined CP-Limits are exceeded. In doing so, the switch preserves network stability if a protection condition (CP-Limit) arises. Note that, although it is likely that one of the SMLT MLT ports (risers) is disabled in such a condition, traffic continues to flow uninterrupted through the remaining SMLT ports.
When you remove the IST configuration from an IST port member, the switch returns the CP-Limit for the port to the default state (enabled). Do not confuse CP-Limit with port rate limiting. Port rate limiting and CP-Limit serve different purposes. Port level rate limiting, if enabled, limits all packets with broadcast and multicast addresses to control the amount of
Traffic flow in an SMLT environment
Traffic flow in an SMLT environment follows these rules:
• If a packet is received from an interswitch trunk port, it is not forwarded to any active SMLT groups, which is key in preventing network loops.
• When a packet is received, a look-up is performed on the forwarding database. If an entry exists, and if the entry was learned locally from the split multilink trunk or through the interswitch trunk as a remote split multilink trunk, it is forwarded out the local port (the packet cannot be sent to the interswitch trunk for forwarding unless there is no local connection). Unknown and Broadcast packets are flooded out all ports that are members of this VLAN.
• For loadsharing purposes in an SMLT scenario, the Ethernet Routing Switch 8300 obeys the trunk distribution algorithm. See Nortel Ethernet
Routing Switch 8300 Planning and Engineering—Network Design Guidelines (NN46200-200) for more details about the algorithms.
Traffic flow example
In an SMLT environment, the two aggregation switches share the same forwarding database by exchanging forwarding entries using the IST. In the following figure,Figure 11 "show vlan info fdb-entry 10 sample output" (page 43), the forwarding databases are shown for a pair of IST nodes (B and C). Note that the entry for 00:E0:7B:B3:04:00 is shown on node C as being learned on MLT-1, but because SMLT REMOTE is true, this entry was actually learned from node B. On B, that same entry is shown as being directly learned through MLT-1 because SMLT REMOTE is false. Figure 12 "Network topology for traffic flow example" (page 43)shows the network topology.
When a packet arrives at node C destined for 00:E0:7B:B3:04:00, if the SMLT REMOTE status is true, the switch tries to send the packet out MLT-1 first, rather than through the interswitch trunk. Traffic rarely traverses the interswitch trunk unless there is a failure. If this same packet arrives at B, it is forwarded to MLT-1 on the local ports.
Split MultiLink Trunking 43
Figure 11
show vlan info fdb-entry 10 sample output
Figure 12
Single port SMLT
With single port SMLT, you can configure a split multilink trunk using a single port and scale the number of split multilink trunks on a switch to a maximum number of available ports. Single port SMLT behaves just like an MLT-based SMLT and can coexist with SMLTs in the same system.
Split MLT links can exist in the following combinations on the SMLT aggregation switch pair:
• MLT-based SMLT + MLT-based SMLT
• MLT-based SMLT + single port SMLT
• single port SMLT + single port SMLT
The rules for configuring single port SMLT are the following:
• The dual-homed device connecting to the aggregation switches must be capable of supporting MLT.
• Single port SMLT is supported on Ethernet ports.
• Each single port SMLT is assigned an SMLT ID from 1 to 512.
• Single port SMLT ports can be designated as Access or Trunk (that is, IEEE 802.1Q tagged or not), and changing the type does not affect their behavior.
• You cannot change a single port SMLT into an MLT-based SMLT by adding more ports. You must delete the single port SMLT, and then reconfigure the port as SMLT/MLT.
• You cannot change an MLT-based SMLT into a single port SMLT by deleting all ports but one. You must first remove the SMLT/MLT and then reconfigure the port as single port SMLT.
• A port cannot be configured as MLT-based SMLT and as single port SMLT at the same time.
Figure 13 "Single port SMLT example" (page 45)shows a configuration, in which both aggregation switches have single port SMLTs with the same IDs. With this configuration, you can have as many single port SMLTs as there are available ports on the switch.
Split MultiLink Trunking 45
Figure 13
Single port SMLT example
SMLT topologies
Four generic topologies are available, in which SMLT can be deployed. Depending on the resiliency and redundancy you require, you can choose among one of the following configurations:
• "Single port SMLT topology" (page 45) • "SMLT triangle topology" (page 46) • "SMLT square topology" (page 47) • "SMLT full mesh topology" (page 48)
Single port SMLT topology
Sometimes you need to exceed the Ethernet Routing Switch 8300 multilink trunk Group ID limit for server farm applications. In this case, you can use Single Port SMLT (seeFigure 14 "Single Port SMLT topology" (page 46)). With this topology, you can scale up to the maximum number of ports on a switch. Any Layer 2 switch capable of link aggregation can be used as the client in this case.
Figure 14
Single Port SMLT topology
SMLT triangle topology
The most often used configuration, the triangle configuration, connects multiple access switches to a pair of Ethernet Routing Switch 8300 devices. In many cases, dual-NIC servers capable of link aggregation are connected directly to the Ethernet Routing Switch 8300 devices in a similar fashion. The following figure,Figure 15 "SMLT triangle topology" (page 47), depicts Extranet Switches (ES) as the SMLT Clients. In real-world applications, any Layer 2 device capable of link aggregation can become the SMLT client.
Split MultiLink Trunking 47
Figure 15
SMLT triangle topology
SMLT square topology
Often used in an enterprise core, the square SMLT configuration provides network resiliency. The following figure,Figure 16 "SMLT square topology" (page 48), shows this topology.
Figure 16
SMLT square topology
SMLT full mesh topology
For maximum reliability and resiliency, all SMLT nodes can be fully meshed. This may not be an economical solution for many cases, but if traffic loss cannot be tolerated, this design can route traffic around any failure. The following figure,Figure 17 "SMLT full mesh topology" (page 49), shows the full mesh topology.
Split MultiLink Trunking 49
Figure 17
SMLT full mesh topology
Using MLT-based SMLT with single port SMLT
You can configure a split trunk with a single port SMLT on one side and an MLT-based SMLT on the other. Both must have the same SMLT ID. In addition to general use,Figure 18 "Changing a split trunk from MLT-based SMLT to single port SMLT" (page 50)shows how this configuration can be used for upgrading an MLT-based SMLT to a single port SMLT without taking down the split trunk.
Figure 18
Changing a split trunk from MLT-based SMLT to single port SMLT
SMLT network design considerations
Use the following base guidelines when designing an SMLT network (for more information, refer to Nortel Ethernet Routing Switch 8300 Planning
and Engineering — Network Design Guidelines (NN46200-200)).
Step Action
1 Define a separate VLAN for the IST protocol:
config mlt 1 ist create ip <value> vlan-id <value>
Split MultiLink Trunking 51
config ethernet <slot/port> perform-tagging enable
3 Enable dropping of untagged frames on IST trunk links:
config ethernet <slot/port> untagged-frames- discard enable
—End—
SMLT and VRRP backup master
When configuring routing on SMLT aggregation switches, Nortel recommends that you use VRRP for default gateway redundancy. With the standard implementation in a VRRP environment, you can have one active primary router per IP subnet, with all other network VRRP interfaces in backup mode.
A deficiency occurs when VRRP-enabled switches use SMLT. If VRRP switches are aggregated into two SMLT switches, the end host traffic is load-shared on all uplinks to the aggregation switches (based on the MLT traffic distribution algorithm).
VRRP normally has only one active routing interface enabled. All other VRRP routers are in backup (standby) mode. Therefore, all traffic that reaches the backup VRRP router is forwarded over the Inter Switch Trunk (IST) link towards the master VRRP router. In this case, the IST link does not have enough bandwidth to carry all the aggregated traffic.
You can overcome this issue by assigning the backup router as the Backup Master router. The Backup Master router is a backup router permitted to actively load-share the routing traffic with a master router.
When enabled, the VRRP Backup Master acts as an IP router for packets destined for the logical VRRP IP address. With the Backup Master router enabled, the incoming host traffic is forwarded over the SMLT links as normal. The Backup Master routes traffic received on the SMLT VLAN, thus avoiding traffic flow across the IST trunk. This eliminates the potential limitation in the available IST bandwidth and provides true load-sharing capabilities.
The Backup Master feature provides an additional benefit. Under normal VRRP operation, a hello packet is sent every second. When three hellos are not received, all switches automatically revert to master mode. This results in a 3 second outage. When you are using VRRP in an SMLT environment, and a link goes down, traffic is automatically forwarded to the remaining ports configured for SMLT VRRP Backup Master. Because both switches are processing traffic, the node immediately recognizes the VRRP state change, so there is faster failure recovery (less than 1 second).
Network design considerations for SMLT with VRRP
When you enable the VRRP BackupMaster with SMLT, refer to the following guidelines:
• The VRRP virtual IP address and the VLAN IP address cannot be the same.
• Configure the hold-down timer for VRRP to a value approximately 150 percent of the IGP (Interior Gateway Protocol, such as RIP or OSPF) convergence time to allow the IGP enough time to reconverge following a failure. That is, if OSPF takes 40 seconds to reconverge, set the holddown timer to 60 seconds.
• Stagger the hold-down timers with ARP requests. This means that the Ethernet Routing Switch 8300 does not have to run ARP at the same time, causing excess CPU load. For example, if one node has the hold-down timer set for 60 seconds, you can set the other to 65 seconds.
• Enable hold-down times on both VRRP sides (Master and BackupMaster).