---

Fault Tolerant Operation of the IEEE 802.1D Spanning Tree Algorithm

---

• Fault Tolerance: Recovery from Bridge Failure
  • Recall the purpose of having loops in the network:

    Loops are the necessary evil that comes with providing multiple paths between a source and a destination: In other words: we need to have loops because we wanted fault tolerance

  ---

  ---

  ---

  ---

  ---

  ---

• Fault tolerant operation

  • Recall the physical network contains a loop:

    ---

    ---

  • We constructed a logical network that is loop-free:

    ---

    ---

    ---

  • Now, suppose port #3 of bridge Br 1 fails:

    Result:

    One of the LAN segment is disconnected from the rest of the network...

    We can repair the failure as follows:

    We can reconnect the LAN segment by unblocking port #2 of Bridge Br 3:

    ---

  • $64,000 Question:

    How can we unblock port #2 of Bridge Br 3

    ---

    ---

  • Note:

    The solution must not use human intervention !!!

     

     

     

    Answer:

    Use a time out mechanism !!!

  ---

  ---

  ---

  ---

  ---

  ---

• Timeout Mechanism in the 802.1D algorithm
  • The time out mechanism:

    Each state variables in a bridge: RootID State of each port has a time out counter associated with it When the time out expires: The variable is reset to its default value

    ---

    ---

  • Algorithm for the time out setting of default values:

    if ( RootID timer expires ) { /* =============================== Reset to the initial state =============================== */ RootID = myID; Distance = 0; for ( each port P ) { status[P] = D; } } if ( state[P] time expires ) { state[P] = D; }

    ---

    ---

    ---

  • Note:

    Every time out mechanism needs a timer maintainance procedures: The time out timer must be reset (from time to time) to prevent the time out from occuring !!!

  ---

  ---

  ---

  ---

  ---

  ---

• Time out maintainance procedure
  • The time out maintenance procedure:

    Recall: A root bridge will periodically (= before the time out expires) transmits a configuration message !!! Graphically:

    ---

    ---

  • The (Distributed) Spanning Tree Algorithm in pseudo code:

    Let: msg = control message received p = port on which the message msg was received /* =========================================== Check for better root =========================================== */ if ( msg.rootID < bridge.rootID ) { /* ===================================== Update state variables ===================================== */ bridge.rootID = msg.rootID; bridge.distance = msg.distance + 1; status[p] = R; // This is the root port for ( all ports q ≠ incoming port p ) { status[q] = D; // Other ports are now designated } /* ===================================== Forward new state to neighbors ===================================== */ for ( all ports q ≠ incoming port p ) { send (bridge.ID, bridge.rootID, bridge.distance) on port q; } Reset all time out timers; } /* =========================================== Check for shorter path to root bridge =========================================== */ else if ( msg.rootID == bridge.rootID && msg.distance + 1 < bridge.distance ) { /* ===================================== Update state variables ===================================== */ bridge.rootID = msg.rootID; bridge.distance = msg.distance + 1; status[p] = R; // This is the root port for ( all ports q ≠ incoming port p ) { status[q] = D; // Other ports are now designated } /* ===================================== Forward new state to neighbors ===================================== */ for ( all ports q ≠ incoming port p ) { send (bridge.ID, bridge.rootID, bridge.distance) on port q; } Reset all time out timers; } /* =========================================== Check for farthest node in even cycle =========================================== */ else if ( msg.rootID == bridge.rootID && msg.distance + 1 == bridge.distance ) { if ( status[p] != R ) { status[p] = B; // Block the incoming port Reset time out for status[p]; // Refresh blocked port } /* ================================================== Time out counter maintenance Node is one of the last nodes in odd cycle: Send control message to your neighbor to keep port blocked =================================================== */ if ( status[p] == R ) { /* ========================================== My old rootID and distance value received AGAIN on my Root port !!!! **** must be maintenance time **** ========================================== */ for ( all ports q ≠ incoming port p ) { send (bridge.ID, bridge.rootID, bridge.distance) on port q; } } } /* =========================================== Check for farthest node in odd cycle =========================================== */ else if ( msg.rootID == bridge.rootID && msg.distance == bridge.distance ) { if ( msg.ID < bridge.ID ) { status[p] = B; // Block the incoming port Reset time out for status[p]; // Refresh block port status } } else { // Do nothing, ignore a worse configuration }

    ---

    ---

    Effect:

    The processing the configuration control message will Recomputes the same value for each state variable The time out conuter has been reset

  ---

  ---

  ---

  ---

• Recovery example
  • Scenario: suppose port #3 of bridge Br 1 fails:

    ---

    ---

  • The disconnected LAN can be reconnected as follows:

    The root bridge transmits the (periodic) configuration message: Bridge 3 will renew the all timers except the timer for the block port (The block state timer must be reset by a message from bridge 1 !!!) Notice: Bridge 1 can not transmit a configuration message to bridge 3 !!! Result: The timer for the state of port #2 of bridge 3 will time out !!! The block state will become designated

    ---

    ---

  • Result: the new tree is as follows:

    the LAN is re-connected to the network !!!