Download: PT-Topology-STP2.pkt

Transcription

1 IEEE Documents Describing the STP IEEE 802.1D IEEE 802.1Q IEEE 802.1w IEEE 802.1s - Media Access Control (MAC) bridges - Virtual Bridged Local Area Networks - Rapid Reconfiguration (Supp. to 802.1D) - Multiple Spanning Tree (Supp. to 802.1Q) 2003, Cisco Systems, Inc. All rights reserved , Cisco Systems, Inc. All rights reserved. BCMSN v Enhancements to STP STP PortFast BPDU Guard Root Guard UplinkFast BackboneFast Per VLAN Spanning Tree (PVST+) Rapid Spanning Tree Protocol (RSTP) Multiple Spanning Tree Protocol (MST) MST is also known as Multiple Instance Spanning Tree Protocol (MISTP) on Cisco Catalyst 6500 switches and above Download this file Download: PT-Topology-STP2.pkt pkt 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

2 Powercycle a host and watch link lights PortFast Powered On Forwarding Learning Listening Blocking gg State State I m adding any addresses on this port to my MAC Address Table. How long until switch link light turns green? Host powered on. Port moves from blocking state immediately to listening state (15 seconds). Determines where switch fits into spanning tree topology. After 15 seconds port moves to learning state (15 seconds). Switch learns MAC addresses on this port. After 15 seconds port moves to forwarding state (30 seconds total). 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v PortFast Problem DHCP PortFast Powered On Forwarding Blocking Learning Listening State Powered On Forwarding Portfast enabled State DHCP Discovery Timeout IP Address = 169.x.x.x Host sends DHCP Discovery Host never gets IP addressing information Also: Insignificant Topology Change A users PC causes the link to go up or down (normal booting or shutdown process). No significant impact but given enough hosts switches could be in a constant state of flushing MAC address tables. Causes unknown unicast floods. DHCP Discovery DHCP Offer The purpose p of PortFast is to minimize the time that access ports wait for STP to converge. When a port comes up, the port immediately moves into Forwarding state. The advantage of enabling PortFast is to prevent DHCP timeouts. Host sends DHCP Discovery Host can now can IP addressing information. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

3 Configuring Portfast UplinkFast Warning: PortFast t should only be enabled on ports that are connected to a single host. If hubs or switches are connected to the interface when PortFast is enabled, temporary bridging loops can occur. If a loop is detected on the port, it will move into Blocking state. Uplinkfast allows access layer switches that have redundant links to multiple distribution switches the ability to converge quickly when a link has failed. For Leafs (end nodes) of the spanning tree. Not for use within backbone or distribution switches (BackboneFast). 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v UplinkFast Root UplinkFast Unblock G 1/1 skips listening and learning and goes directly to forwarding X Not supported with Packet Tracer Access1(config)#spanning-tree uplinkfast UplinkFast must have direct knowledge of the link failure in order to move a blocked port into a forwarding state. Single Root Port but multiple potential root ports. If Root Port fails, next-lowest path cost is unblocked and used without delay (almost). This switchover occurs within 1 second. Uplinkfast is enabled for the entire switch and all VLANs. Not supported on a per-vlan basis. Uplinkfast keeps track of all possible paths to the Root Bridge. So, not allowed on the Root Bridge Switches BID: Raised to 49,152 to make it unlikely it will be the Root Bridge. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

4 BackboneFast X Root My link to the Root has gone down. I have no Thanks for alternate path telling me Core to it. So, I m the is the Root. I new root and will change my send out my RP to Fa 0/5. BPDUs on all ports. RP X RP Inferior BPDU Root Blocking Forwarding Listening I just This heard new from BPDU is Core inferior that they to the are one still it the had Root. stored I will: for Send this BPDU port so to I D1 will Transition ignore port it. immediately Let me send to my listening current state Root a saving After query seconds (RLQ). (Max this port Age) will now go into Forwarding state. Switch(config)#spanning-tree backbonefast Backbone fast is a Cisco proprietary feature that, once enabled on all switches can save a switch up to 20 seconds (Max Age) when it recovers from an indirect link failure. Configured in global configuration mode and should be enabled on all switches in the network. Disabled by default. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v BackboneFast is initiated when a root port or blocked port on a switch receives inferior BPDUs from a designated bridge. Inferior BPDUs are sent from a designated bridge that has lost its connection to the root bridge. Normally, a switch must wait for Max Age (20 seconds) to expire before responding to an inferior i BPDU. With Backbonefast, switch determines alternate paths to Root. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v BackboneFast Normal BPDU Protecting against unexpected BPDUs = Core = Dist1 Root Guard Inferior BPDU = Dist1 = Dist1 Same Switch FYI More Information An inferior BPDU identifies one switch as both the root bridge and the designate bridge. Distribution 1 is the Designated Bridge. Normally, sends BPDUs with Root Bridge as the Core BID. Inferior BPDU A received BPDU that identifies the root bridge and the designated bridge as the same switch. ( I was only just the Designated Bridge, but now that I can t get to the Root Bridge, so now I am also the Root Bridge. ) BPDU Guard Loop Guard Coast Guard 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

5 Problem: Unexpected BPDUs BPDU Blocking and now listening to BPDUs Portfast X Forwards BPDUs to other switches. Solution: BPDU Guard BPDU Portfast & BPDU Guard Err-Disable, Shutdown No BPDUs sent STP Reconvergence? A port configured with PortFast will go into blocking state if it receives a Bridge Protocol Data Unit (BPDU). This could lead to false STP information that enters the switched network and causes unexpected STP behavior. Newly connected switch could advertise itself as the root. BPDU Guard: Developed to protect integrity of switch ports with PortFast enabled but also keeps maintains STP integrity by disallowing unauthorized switches. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v Not supported with Packet Tracer Distribution1(config)#interface range fa 0/10-24 Distribution1(config-if-range)#spanning-tree if range)#spanning tree bpduguard enable When the BPDU guard feature is enabled on the switch, STP shuts down PortFast enabled interfaces that receive BPDUs instead of putting them into a blocking state. Errdisable: Port must be manually re-enabled or automatically ti recovered via timers. BPDU guard will also keep switches added outside the wiring closet by users from impacting and possibly violating Spanning Tree Protocol. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v Root Guard Root Guard Protect Potential Root Protect Potential Root UplinkFast must be disabled d because it cannot be used with root guard. Root Guard prevents a switch from becoming the root bridge. Typically access switches Configured on switches that connect to this switch. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

6 Root Guard Unidirectional Link Detection Protocol (ULDP) Root Guard I STP will now Inconsistent transition to listening State no sate, traffic then is learning state, passed. t then forwarding sate. Superior BPDU I no longer want to be root. I have been reconfigured to I want to be be a non-root root bridge. bridge Blocked Port Designated Port This message appears after root guard blocks a port: %SPANTREE-2-ROOTGUARDBLOCK: Port 0/3 tried to become non- designated in VLAN 1. Moved to root-inconsistent state Spanning-Tree Protocol (STP) resolves redundant physical topology into a loop-free, tree-like forwarding topology. This is done by blocking one or more ports. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v ULDP ULDP BPDU BPDU BPDU Loop BPDU BPDU BPDU No BPDU s Received Change to Forwarding State BPDU No BPDU s Received Change to Forwarding State STP uses Bridge Protocol Data Units (BPDUs). If a switch ss port in blocking port stops receiving BPDUs: STP eventually ages out the STP information for the port (up to 50 secs) Moves port to forwarding state. This creates a forwarding loop or STP loop. How is it possible for the switch to stop receiving BPDUs while the port is up? The reason is unidirectional link. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v RFC 5171: Issues arise when, due to mis-wirings or to hardware faults, the communication path behaves abnormally and generates forwarding anomalies. Link fails in the direction of SwitchC. SwitchC stops receiving traffic from SwitchB. However, e SwitchB still receives es traffic from C. UDLD is a Layer 2 (L2) protocol that works with the Layer 1 (L1) mechanisms to determine the physical status of a link. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v

7 ULDP ULDP My device/port ID & your device port ID My device/port ID & your device port ID Layer 1:Auto- negotiation configured (speed/duplex) Layer 2: UDLD configured My device/port ID & your device port ID My device/port ID & your device port ID X Unidirectional link failure UDLD-3-DISABLE: DISABLE: Unidirectional link detected on port 1/2. Port disabled Port disabled Enable both auto-negotiation and UDLD to prevent unidirectional connection. With UDLD switches share Device/Port ID information. Port shutdown by UDLD remains disabled until: Manually reenabled or errdisable timeout expires (if configured) 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v Rapid Spanning Tree Protocol Configuring ULDL Switch(config)# udld {enable aggressive} or Switch(config)# interface fa 1/2 Switch(config-if)# udld {enable aggressive} Normal mode (enable) Port is allowed to continue it s operation merely marks the port as being in undetermined state and generates a syslog message. Aggressive mode Port is place in Errdisable state and cannot be used. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

8 Loopguard Loopguard No Loopguard Configured Loop X BPDU No BPDU s Received Change to Forwarding State Loopguard Configured BPDU BPDU X Unidirectional link failure %SPANTREE-2-LOOPGUARD LOOPGUARD_BLOCK: Loop guard blocking port FastEthernet1/0 on VLAN0010 Inconsistent Blocking State Loopguard also protects t against ports erroneously transitioning i to forwarding mode. Loopguard will also protect against STP failures, designated switch not sending BPDUs due to software problems. If the switch begins to receive BPDUs again, it will transition through normal STP states. Loopguard does NOT protect against problems due to wiring issues. Highest level of protection is to enable both Loopguard and UDLD. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v Configuring Loopguard Switch(config)# spanning-tree loopguard default or Switch(config)# interface fa 1/2 Switch(config-if)# spanning-tree guard loop RSTP IEEE 802.1w (Rapid Spanning Tree Protocol) 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. 2-32

9 Rapid Spanning Tree Protocol Rapid Spanning Tree Protocol The immediate hindrance of STP is convergence. Depending on the type of failure, it takes anywhere from 30 to 50 seconds, to converge the network. RSTP helps with convergence issues that t plague legacy STP. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v STP vs RSTP RSTP RSTP is based on IEEE 802.1w standard. IEEE 802.1w took 802.1D s principle concepts and made convergence faster. STP topology change takes 30 seconds (two intervals of Forward Delay timer). RSTP is proactive and therefore negates the need for the 802.1D delay timers. RSTP (802.1w) supersedes 802.1D, while still remaining i backward compatible. RSTP BPDU format is the same as the IEEE 802.1D BPDU format, except that the Version field is set to 2 to indicate RSTP. The RSTP spanning tree algorithm (STA) elects a root bridge in exactly the same way as 802.1D elects a root. RSTP can be applied on Cisco switches as: A single instance per VLAN Rapid PVST+ (RPVST+) Multiple instances IEEE 802.1s Multiple Spanning Tree (MST) 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

10 STP Port Behavior and States 802.1D Ports Root Port Designated Port Blocking Port Not Designated Port and Not Root Port Cisco s proprietary UplinkFast has a hidden Alternative Port offering parallel paths, but in Blocking state. States Disabled (Not 802.1D state) Blocking Listening Learning Forwarding Only state that sends/receives data. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v RSTP Root Bridge: Same election process as 802.1D (lowest BID) Ports Root Port (802.1D Root Port) The one switch port on each switch that has the best root path cost to the root. Designated Port (802.1D Designated Port) The switch port on a network segment that has the best root path cost to the root. Alternate Port (802.1D Blocking Port) A port with an alternate path the root. An alternate port receives more useful BPDUs from another switch and is a port blocked. Similar to how Cisco UplinkFast works. Backup Port (802.1D Blocking Port) A port that provides a redundant (but less desirable) connection to a segment where another switch port already connects. A backup port receives more useful BPDUs from the same switch it is on and is a port blocked. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v RSTP Port States Operational Port State STP Port State RSTP Port State Disabled Disabled Discarding Enabled Blocking Discarding Enabled Listening Discarding Enabled Learning Learning Enabled Forwarding Forwarding RSTP defines port states based on what it does with incoming data frames. Discarding Incoming frames are dropped No MAC Addresses learned Combination of 802.1D (Disabled), Blocking and Listening Learning Incoming frames are dropped MAC Addresses learned Forwarding Incoming frames are forward. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v RSTP BPDUs STP Port State STP BPDUs RSTP Port State RSTP BPDUs Disabled Not Sent/Received Discarding Not Sent/Received Blocking Receive only Discarding Sent/Received Listening Sent/Received Discarding Sent/Received Learning Sent/Received Learning Sent/Received Forwarding Sent/Received Forwarding Sent/Received RSTP uses same 802.1D BPDU format for backward compatibility D and 802.1w switches can coexist. BPDUs sent out every switch port at Hello Time intervals regardless if BPDUs are sent on the port. When three BPDUs in a row (6 seconds) are missed: the neighbor switch is presumed down All MAC address information pointing to that switch (out that port) is immediately aged out (flushed) Switch can detect a neighbor down in 6 seconds instead of MaxAge of 20 seconds. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v

11 RSTP Convergence Convergence is a two step process: 1. Elect a Root Bridge 2. Examine all switch ports which by default are in Blocking state and advance to the appropriate state to prevent loops. STP requires the expiration of several timers before switch ports can be moved to Forwarding state. RSTP takes a different approach: When a switch joins the topology (powered-up) or detects a failure in the existing topology Determines its forwarding decisions based on the type of port. Edge Port Root Port Point-to-Point Port Edge Ports Edge port will never have a switch connected to it so cannot form bridging loops. Immediately transitions to forwarding state. Traditional identified with STP PortFast feature. For familiarity the command is the same: spanning-tree portfast Never generates topology changes notifications (TCNs) when the port transitions to a disabled or enabled status. If an edge port receives a BPDU, it loses its Edge Port status becomes a normal spanning-tree port. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v Non-Edge Ports Point-to-Point: The Quick Handshake Proposal Root DP RP A Agreement B Root Port The one switch port on each switch that has the best root path cost to the root. Point-to-Point Port (Link Type) Port operating in full-duplex mode. Connects to another switch and becomes a Designated Port. Uses a quick handshake with neighboring switch rather than timers to decide port state. Shared Medium Port (Link Type) Port operating in half-duplex mode. It is assumed that the port is connected to shared media where multiple switches might exist. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v Switch A is connected to Switch B through a point-to-point p link, All ports are in the Discarding (Blocking) state. Switch A has a lower BID than Switch B. Switch A sends a proposal message (Configuration BPDU) to Switch B, proposing itself as the Root Bridge and the designated switch on the segment. Switch B: Selects its new root port the port from which the proposal message was received and immediately goes into Forwarding State Forces all nonedge ports to the Discarding (Blocking) state, Sends an agreement message. Switch A: Immediately transitions its designated port to the forwarding state. No loops in the network are formed because Switch B blocked all of its nonedge ports and because there is a point-to-point link between Switches A and B. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v

12 RSTP Topology Change Notifications Proposal Root DP RP A B Agreement Proposal Root DP DP DP RP A B Agreement C Proposal Root DP RP DP RP DP RP A B C D Agreement Switch C is connected to Switch B: a similar set of handshaking messages are exchanged. Switch C selects the port connected to Switch B as its root port, and both ends immediately transition to the forwarding state. Handshaking process continues throughout topology D 802.1D 802.1D Switch detects a state change (up or down), it sends the Root Bridge a TCN BPDU. The Root Bridge sends out a Configuration BPDU (TCN bit set) to all switches to tell them about the change. (30 seconds before Forwarding) RSTP Detects a topology change only when a nonedge port transitions to the Forwarding State. RSTP uses its convergence mechanisms (Edge Ports, Point-to-Point to ports, handshaking, etc.) to prevent bridging loops. Therefore, topology changes are detected only so MAC address tables can be updated and corrected. This means that a loss of connectivity is not considered as a topology change any more, contrary to 802.1D (that is, a port that moves to blocking no longer generates a TC). 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v RSTP Topology Change Notifications RSTP When a topology change occurs: Switch flushes the MAC addresses associated with all nonedge ports. Switch sends BPDU with TCN bit set to all needs to be notified neighbors so they can update their MAC Address tables too. When a bridge receives a BPDU with the TCN bit set from a neighbor: It clears the MAC addresses learned on all its ports, except the one the port that it receives the topology change. It sends BPDUs with TCN set on all its designated ports and root port (RSTP no longer uses the specific TCN BPDU, unless a legacy bridge needs to be notified). This way, the TCN floods very quickly across the whole network - now a one step process. The initiator of the topology change floods this information throughout the network, as opposed to 802.1D where only the root did. Much faster than the 802.1D equivalent < wait for the root bridge to be notified, and then max age plus forward delays>. In just a few seconds, or a small multiple of hello-times, most of the entries in the CAM tables of the entire network (VLAN) flush. This approach results in potentially more temporary flooding, but on the other hand it clears potential stale information and allows rapid convergence. RSTP no longer uses the specific TCN BPDU, unless a legacy bridge 2003, Cisco Systems, Inc. All rights reserved. BCMSN v RSTP BPDU Flag Byte Use 2003, Cisco Systems, Inc. All rights reserved. BCMSN v

13 Rapid PVST Implementation Commands Cisco implements RSTP with Rapid PVST+ Rapid PVST Implementation Commands Cisco implements RSTP with Rapid PVST+ Switch(config)# spanning-tree mode rapid-pvst To revert back to the default PVST+ using traditional 802.1D: Switch(config)# spanning-tree mode pvst 2003, Cisco Systems, Inc. All rights reserved. BCMSN v To configure an RSTP edge port: Switch(config-if)# spanning-tree portfast RSTP automatically decides if a port is point-to-point link operating in full duplex or half-duplex. If you need to set it manually, other switch is in Half-Duplex but still point-to- point (by the way, both ends must then be Half-Duplex): Switch(config-if)# spanning-tree link-type point-to-point 2003, Cisco Systems, Inc. All rights reserved. BCMSN v Rapid PVST Implementation Commands Rapid PVST Implementation Commands 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

14 Multiple Spanning Tree Protocol 802.1s Instance 1 maps to VLANs Instance 2 maps to VLANs Implementing MSTP 2003, Cisco Systems, Inc. All rights reserved Multiple Spanning Tree (MST) extends the IEEE 802.1w RST algorithm to multiple spanning trees. The main purpose of MST is to: Reduce the total number of spanning-tree instances to match the physical topology of the network Thus reduce the CPU cycles of a switch. Allows the network administrator to configure the exact number of instances. PVST+ runs a single instance of STP for each VLAN and does not take into consideration the physical topology. May have 1,000 VLANs but only 2 different topologies (2 different Root Bridges). PVST+ will still create 1,000 instances of STP MST, on the other hand, uses a minimum number of STP instances to match the number of physical topologies present. May have 1,000 VLANs but only 2 different topologies (2 different Root Bridges). MST will let you specify only 2 instances of STP. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v MST Regions 802.1D 802.1D MST MST Region MST Region is a group of switches placed under a common administration (like an AS). In most networks a single MST region is sufficient. A single MST Region can handle 15 STP instances (topologies). Within a region, all switches must run the instance of MST as defined by: MST configuration name (32 characters) MST configuration revision number ( 0 to 65,535) MST instance-to-vlan to mapping table (4,096 entries) MST was designed to work with all forms of STP. IST (Internal Spanning Tree) instance runs to work out a loop-free topology inside the MST Region. IST presents the entire MST region as a single virtual switch (bridge) to the CST (802.1D) outside. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v Remember, the whole idea of MST is to map multiple VLANs to a smaller number of STP instances. Cisco supports a maximum of 16 MST Instances (MSTIs) in a region. The IST uses MST 0 leaving 1 through 15 available for use. The Distribution1 switch is the primary root bridge for the data VLANs 10, 30, and 100 Secondary root bridge for the voice VLANs 20, 40, and 200. The Distribution2 switch the primary root bridge for the voice VLANs 20, 40, and 200 Secondary root bridge for the data VLANs 10, 30, and 100. Distribution1 is chosen as CIST regional root. It means that Distribution1 is the root for IST , Cisco Systems, Inc. All rights reserved. BCMSN v

15 MST MST 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v MST Extended System ID in Bridge ID Field 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

16 802.1D creates a single instance of STP for all VLANs. Cisco s RSTP is Rapid PVST+ Cisco s RSTP is Rapid PVST+ PVST+ and RPVST create a single instance of STP for each VLAN. If there are 500 VLANs in the network that would be 500 instances of STP running PVST+ does allow different VLANs to have different Root Bridges which can allow for the use of redundant links. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v Spanning Tree and EtherChannel Configuring Link Aggregation with EtherChannel Etherchannel Bundle Spanning Tree only allows a single link between switches to prevent bridging loops. Cisco s s EtherChannel technology allows for the scaling of link bandwidth by aggregating or bundling parallel links. Treated as a single, logical link. Access or Trunk link Allows you to expand the link s capacity without having to purchase new hardware (modules, devices). Load Balances, View as one logical port, Redundancy. 2003, Cisco Systems, Inc. All rights reserved , Cisco Systems, Inc. All rights reserved. BCMSN v

17 EtherChannel EtherChannel EtherChannel allows for two to eight links. Fast Ethernet (FE) Fast EtherChannel Up to 1600 Mbps Gigabit Ethernet (GE) Gigabit EtherChannel Up to 16 Gbps 10-Gigabit Ethernet (10GE) 10 Gigabit EtherChannel Up to 160 Gbps This does not mean the total bandwidth of the bundle equals the sum of the links. The load is not always distributed evenly (coming). The Cisco Catalyst family of switches supports two types of link aggregation: Port Aggregation Protocol (PAgP) - Cisco proprietary Default when port channel is created (coming) Link Aggregation Control Protocol (LACP) - Industry standard 802.3adbased protocol Eh EtherChannel provides redundancy. d If one link fails traffic is automatically moved to an active link. Transparent to end user. LACP (coming) also allows for standby links (coming). 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v Distribution of Traffic and Load Balancing The key is consistency for all links in the bundle: Media Same media type and speed Same duplex VLANs All ports within the bundle must be configure with: Same VLAN (if access) Same trunking encapsulation and mode (if trunk) Mode on opposite switches do not have to be the same as long as it still forms a trunk. Same Native VLAN Pass the same set of VLANs Load is not balanced equally across links. EtherChannel uses a hashing algorithm. Single input is used (such as Source IP address), the hash will only look at the bits associated with this input. (coming) Two inputs are used (such as Source IP address and Destination IP address), the hash will perform an exclusive OR (XOR) operation on both inputs. (coming) Both of these will compute a binary number that selects a link number in the bundle to carry the frame. (coming) 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

18 Load Balancing Load Balancing 0 1 Let s take a brief look at how this works. We will focus on the 2, 4 and 8 link possibilities as this is easier to understand and the only options that provide more ideal load balancing. A 2 link EtherChannel bundle requires a 1-bit index usinganxor XOR. If the index is 0, link 0 is selected If the index is 1, link 1 is selected A 4 link EtherChannel bundle requires a 2-bit index using an XOR. 4 possible links: 00, 01, 10, 11 An 8 link EtherChannel bundle requires a 3-bit index using an XOR. 8 possible links: 000, 001, 010, 011, 100, 101, 110, , Cisco Systems, Inc. All rights reserved. BCMSN v Example: 2 Link EtherChannel. Packet sent from to The chosen hash uses Source IP and Destination IP address At most there can only be 8 links in bundle, so only the last 3 rightmost bits (least-significant) of the addresses will ever need to be indexed or examined. 3 bits will give us 8 choices (8 links max in a bundle) => => In our example we have 2 links in the EtherChannel (1 bit index): The XOR is performed only on the rightmost bit 1 XOR 0 1 XOR 0 = 1 Link 1 is used 2003, Cisco Systems, Inc. All rights reserved. BCMSN v Load Balancing 0 Load Balancing Example: 4 Link EtherChannel Packet sent from to Our hash used the Source IP and Destination IP address => => If there are 4 links in the EtherChannel (2 bit index): The XOR is performed only on 2 rightmost bits 01 XOR 10 Each bit is computed separately 01 XOR 10 = 11 1 XOR 0 = 1 0 XOR 1 = 1 Link 3 (11 2 ) is used 2003, Cisco Systems, Inc. All rights reserved. BCMSN v Example: 8 Link EtherChannel Packet sent from to Our hash used the Source IP and Destination IP address => => If there are 8 links in the EtherChannel (3 bit index): The XOR is performed only on the 3 rightmost bits 001 XOR 110 Each bit is computed separately 001 XOR 110 = XOR 0 = 1 0 XOR 1 = 1 0 XOR 1 = 1 Link 7 (111 2 ) is used 2003, Cisco Systems, Inc. All rights reserved. BCMSN v

19 Configuring EtherChannel Load Balancing Load Balancing Switch(config)# port-channel load-balance method The load balancing method is configured in global configuration mode. 2003, Cisco Systems, Inc. All rights reserved. 73 BCMSN v and 4500 switches also allow hash input to be based on: dst-port (destination port) src-dst-port (source and destination ports) Dafaults for 29xx and 35xx (this may vary so check documentation) Layer 2 switching (switched port) is src-mac (coming) Layer 3 switching (routed port) is src-dst-ip (coming) For non-ip traffic the switch will distribute frames based on MAC addresses. Multicasts and broadcasts sent over one link in the EtherChannel are not sent back over other links in the EtherChannel. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v Load Balancing Switch(config)# port-channel load-balance? dst-ip Dst IP Addr bits dst-mac Dst Mac Addr bits src-dst-ip Src XOR Dst IP Addr XOR src-dst-mac Src XOR Dst Mac Addr XOR src-ip Src IP Addr bits src-mac Src Mac Addr bits EtherChannel Protocols PAgP LACP LACP PAgP Switch(config)# port-channel load-balance src-dst-ip Normally, the default Source IP and Destination IP addresses will result in a fair statistical distribution of frames. This is because of the random nature of multiple Source and Destination IP addresses. However, if a single server s destination IP address is receiving most of the traffic this may cause one link to be overused in a two link EtherChannel. Two links in a four link EtherChannel Four links in an eight link EtherChannel. Use only Source IP address or include MAC addresses to create a more balanced load across the bundle. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v The Cisco Catalyst family of switches supports both: Port Aggregation Protocol (PAgP) - Cisco proprietary Default when port channel is created (coming) Link Aggregation Control Protocol (LACP) - Industry standard d 802.3ad-based protocol Not many differences. When a Cisco switch is connected to a non-cisco switch use LACP. Must be the same on both ends 2003, Cisco Systems, Inc. All rights reserved. BCMSN v

20 EtherChannel Protocols Fa0/1 Fa0/4 EtherChannel Protocols Fa0/1 Fa0/4 Channel Group PAgP requires identical static VLANs or trunking encapsulation with same allowed VLANs. If the VLAN, speed or duplex on a port in the bundle is changed PAgP automatically reconfigures the rest of the ports in that bundle. 2003, Cisco Systems, Inc. All rights reserved. 77 BCMSN v Channel-group number: 1 64 Does not need to be the same on both switches but its recommended that it usually is. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v No PAgP or LACP negotiation PAgP modes on EtherChannel on desirable EtherChannel desirable auto on Forces port to channel without PAgP negotiation. Both ends must be on. All ports channeling You can use channel-group # mode on when the connecting device does not support PAgP and you need to set up the channel unconditionally. An interface in desirable mode can form an EtherChannel with another interface that is in desirable or auto mode. Desirable (Active) - Actively asks to form a channel 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

21 PAgP modes LACP modes auto EtherChannel desirable active EtherChannel active passive An interface in auto mode can form an EtherChannel with another interface in desirable mode. Auto (default, passive) - Waits to be asked to form a channel. An interface in auto mode cannot form an EtherChannel with another interface that is also in auto mode because neither interface starts PAgP negotiation. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v An interface in the active mode can form an EtherChannel with another interface that is in the active or passive mode. An interface in the passive mode can form an EtherChannel with another interface that is in the active mode. An interface in the passive mode cannot form an EtherChannel with another interface that is also in the passive mode because neither interface starts LACP negotiation. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v Forming EtherChannels Configuring PAgP EtherChannel on on desirable PAgP Negotiated EtherChannel desirable e auto active LACP Negotiated EtherChannel active passive Notice: Load balancing does not have to match but usually it does. DTP on DLS2 is dynamic auto (result is trunk with DLS1) PAgP configured on both ends 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

22 Verifying Verifying DLS1#show run port-channel load-balance dst-ip interface Port-channel1 switchport trunk encapsulation dot1q switchport mode trunk interface FastEthernet0/1... interface FastEthernet0/11 switchport trunk encapsulation dot1q switchport mode trunk channel-group 1 mode desirable interface FastEthernet0/12 switchport trunk encapsulation dot1q switchport mode trunk channel-group 1 mode desirable DLS2#show run port-channel load-balance src-dst-ip interface Port-channel1 switchport trunk encapsulation dot1q interface FastEthernet0/1 t et0/... interface FastEthernet0/11 switchport trunk encapsulation dot1q channel-group 1 mode auto interface FastEthernet0/12 switchport trunk encapsulation dot1q channel-group 1 mode auto 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v Verifying Configuring g LACP 2003, Cisco Systems, Inc. All rights reserved. BCMSN v Port Priority - (Optional for LACP) LACP uses the port priority to decide which ports should be put in standby mode. Not typically used (more with hardware limitation). Ports with lower priority are active, rest are standby. (Default is 32,768) System Priority - (Optional for LACP) Valid values are 1 through Higher numbers have lower priority. (Default is 32768, switch MAC is tiebreaker) Recommended only when some ports are in standby. 2003, Cisco Systems, Inc. All rights reserved. BCMSN v

23 Configuring g LACP: DLS1 and DLS2 Verifying (only showing DLS1) DLS1#show run port-channel load-balance dst-ip interface Port-channel1 switchport trunk encapsulation dot1q switchport mode trunk interface FastEthernet0/11 switchport trunk encapsulation dot1q switchport mode trunk lacp port-priority 99 channel-group 1 mode active interface FastEthernet0/12 t0/12 switchport trunk encapsulation dot1q switchport mode trunk lacp port-prioritypriority 99 channel-group 1 mode active interface FastEthernet0/13 switchport trunk encapsulation dot1q switchport mode trunk channel-group 1 mode active interface FastEthernet0/14 switchport trunk encapsulation dot1q switchport mode trunk channel-group 1 mode active 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v Verifying Verifying 2003, Cisco Systems, Inc. All rights reserved. BCMSN v , Cisco Systems, Inc. All rights reserved. BCMSN v

24 Configuring Link Aggregation with EtherChannel (End) 2003, Cisco Systems, Inc. All rights reserved. 2-93