3. INTERCONNECTING NETWORKS WITH SWITCHES. THE SPANNING TREE PROTOCOL (STP)

Transcription

1 3. INTERCONNECTING NETWORKS WITH SWITCHES. THE SPANNING TREE PROTOCOL (STP) 3.1. STP Operation In an extended Ethernet network (a large network, including many switches) multipath propagation may exist between two or more systems, which decreases the network efficiency. The existence of two or more parallel paths closes a propagation loop for the MAC frames. Issues related to the unwanted blocking of "unicast" flooding frames, and the avalanche growth of broadcast frames (a phenomenon called "broadcast storm") are determined by the existence of propagation loops. The function of the STP (Spanning Tree Protocol IEEE 802.1D) is precisely the logic level (data link layer) interruption of the propagation loops, without eliminating the physical layer redundancy. Moreover, physical redundancy is required in most telecommunications networks, but the Ethernet propagation should be only one path at a time. The solution adopted in IEEE 802.1D is to find a so called "spanning tree" using a distributed algorithm that works only in active nodes and in the data link layer (switches). Each switch has a unique STP identifier formed by a most significant proportion that can be set up by the network administrator and a least significant part, i.e., the MAC address of the switch, to ensure uniqueness of the identifier. The switch with the lowest ID is elected as the root node of the tree (Root Bridge). ST algorithm is then used (in the data link layer) to obtain the tree with the optimal paths from each network node to root node. The selection of the optimal route is based on a criterion of cost. Thus, each switch chooses a single optimum port on the way to the root (Root Port), which has the minimum cost among all its active ports (in state Designated Port). The exception to this rule is the root switch, which does not need a Root Port and has all the ports in the state Designated Port. As a rule, a loop is always interrupted by two secondary switches (not Root Bridge), where each has at least two paths to the root and which are connected to each other through a direct link. First, one of these two switches is chosen as Designated Bridge that will be placed on the ST hierarchy closer to the root than the other switch. Thus, the selection of the optimal path to the root and the actual interruption of the loop by closing the ports will be made by the switch that was not elected Designated Bridge. The loop interruption will be achieved by interrupting the link between the two secondary switches by blocking the port that belongs to the switch that was not elected as Designated Bridge. Therefore, the switch port belonging to Designated Bridge remains open ( in the state Designated Port or " forwarding") and the port belonging to the other secondary switch is locked (state Non- Designated Port or Alternate Port). The distributed STP mode is implemented by sending from each switch at regular intervals, Bridge Protocol Data Unit (BPDU) packets to neighboring nodes, which contain the following information:

2 What is the root node, specified by its identifier, Root ID; Optimal path cost to the current root switch (minimum cost), Root Path Cost (RPC); Switch ID of the switch that has issued the BPDU, Bridge ID; The identifier of the port issuing the BPDU, Port ID; Values of timers used to control convergence. BPDU packets are transmitted inside the MAC frames, using a single multicast destination address. In the first step of the ST algorithm, each switch will transmit a BPDU specifying Root Bridge ID with the same value as its own Bridge ID. Therefore, the initialization state of each switch considers it as the root of the tree. Therefore, the cost it announces for the optimal connection to the root is zero. Upon receiving such a packet, a switch compares the information packet received values with the values already in its memory to determine the role it will have in the STP topology. Upon receiving a BPDU from a neighboring switch, each switch first checks the relationship between its Root Bridge ID and the value announced by the neighbor. If the root identifier is better than the previously saved value, it starts the calculation of the cost of the new path to the root, by adding to the cost reported by the neighbor, the cost of the local port that the BPDU was received by. If the root path cost already known by the switch is greater than the sum between the cost from the received BPDU and the cost of the receiving port, then the switch adjusts its minimum cost to root properly and assumes that the shortest path to the root is by the port where the BPDU packet has just been received. As discussed above, it is considered that the STP tree is stable (at convergence) when a root switch (specified by Root ID) was elected and each switch has chosen a Root Port (RP), Designated Bridge and one or more multiple Designated Ports (DP). Usually, the decision of a port blocking is taken when the port cost of the path to the root is greater than the cost of any different available path. However, there are situations when you can not only take the decision by means of the cost. Thus, it is possible from a particular switch, in terms of the cost, to have two paths of equal costs to the root. In such situations, the path which includes a non-root switch root having the minimum Bridge ID is always chosen. Sometimes, a switch is connected through multiple ports on the same LAN. If there are situations where the costs are identical, the port with the lower identifier (port ID, PID) is chosen, because the Bridge ID of the first switch to the root has the same value. For this reason, the BPDU packets contain also the emitter switch identifier and the emitting port identifier. Therefore, the decision is taken using the three criteria in order until one of them is satisfied: cost path, Bridge ID, and transmitter Port ID. Port Identifier is useful where a loop is formed between two neighboring switches (the smallest loop possible). When a port is blocked it does not redirect frames of information for users, but still sends and receives BPDU frames. Thus, although the port is not active for data traffic, the switch monitors the STP parameters for the paths determined on this port. Therefore, if the optimal path to the root is interrupted (the one that led to blocking this port), it is required to reactivate the port, as it offers a new optimum path. When the topology changes due to the change status of a port or a link, the root switch sends over the entire tree a message announcing the change of the topology. This procedure allows the

3 adjustment of the information about the best paths already stored in the other swithches in the network. Thus, it recalculates the costs and checks the STP parameters only for the newly announced routes. If the switches allow the formation of virtual LANs (VLAN), then it can work separately one instance of STP for each VLAN protocol. Thus, for each VLAN a root will be selected and carried out in parallel and independently all STP operations. If using one protocol instance for each VLAN, the protocol is called PVST (Per VLAN STP) STP Parameters Since Bridge ID value determines the position switch STP hierarchy identifier must meet two conditions: - Uniqueness - there are two switches that can not occupy the same position in the tree; - Scalability - sometimes the shaft position can be set manually by the administrator and therefore, it must be able to change the value of Bridge ID. At first it seems that the two conditions are mutually exclusive, since uniqueness implies often a static allocation of address values. However, if the STP problem is solved by dividing into two to cuvânului representation Bridge ID value. The most significant bits of the bridge ID ( first 2 bytes ) is the priority field that can be modified by the administrator. In the default (no administrator intervention ) the priority field is set to the value ( 8000H ), which is the decimal value in the mid- range of possible to represent two bytes. The least significant bits ( last 6 bytes ) are represented by the switch MAC address ( unique for each VLAN ). The unique Bridge ID is provided by the MAC address of the least significant because MAC addresses are unique (IEEE ), and scalability is ensured by the head of priority. Without administrator intervention all switches have equal priority and root decision will be based only on the MAC address of the bridge ID (lowest MAC address will cause the Root Bridge). The administrator can set a priority mechanism in the formation of STP topology by providing values for Bridge ID as the default. States one can find a port to the data link layer, a switch that is on the court STP are: - Block ( Blocking, non - Designated Port ) - A port in this state can not redirect the teaching of information for users; however, the port receives BPDU packets seeks to identify the root switch ; - Listening ( Listening ) - a listening port to an alarm only after STP algorithm to determine the optimal tree from the received BPDU packets. At this point, the port can send and receive BPDU packets. If the port forward BPDU packets means that it is preparing to participate in the STP active topology switches and inform neighbors about this. The default time that a port is in listening state (forward delay) is 15 s; - Learning (Learning ) - port in this state prepares to become active traffic information ( to pass the state redirects) by learning MAC addresses. The default time that a port is in learning state (forward delay) is all of 15 s; - Redirection ( Forwarding ) - the port part of the active STP topology, redirects frames of information and transmits / receives BPDU packets ;

4 - Sleep ( Disabled ) - not part of the active STP topology, thus no redirects frames of information and does not transmit / receive BPDU packets. Meters are used to update information for STP tree. The time counters are used to determine how long a port should switch from one state to another. Also, STP uses timers to determine the availability of the STP tree formation neighboring switches and the maximum storage address MAC address tables ( dynamically learned ). Meanings associated values for these counters are: - Hello time - the interval between the emission of two successive BPDU packets. The default is 2 seconds; - Maximum Age Time - the interval of time that must be a switch (receiver of this BPDU ) blocking state before moving into the state of obedience. The default is 20 seconds; - Forward delay - the interval of time that must be a switch (receiver of this BPDU ) listening state before moving to the state of learning and learning state before forwarding to an alarm. The default is 15 seconds The default values of the these counters are suitable for a wide range of configurations and operating scenarios. However, these values can be modified to determine the optimal operation in a particular case. On the other hand, improper setting of the values of these time values may lead to the network instability. The cost of use of the ST is determined in each switch, calculated independently for the two ends of a connection and depend only on the maximum flow rate of the port. Since the decision is based on the minimum cost of a port is in a relationship and linear reverse flow to port (in Mb / s). Table 3.1 presents the correlation between flow port and STP cost for two protocol variants. D(Mb/s) Cost (IEEE, new version) Cost (IEEE, old version) Tab The relationship between the datarate of a port and its STP cost Example of a STP topology Further, it will be considered a network configuration LAN - Ethernet extended. This wide area network is depicted in Figure 2 and consists of four network LAN star topology with the following components of the interconnection: LAN 1 - switch SW BASE T ports (Fast Ethernet); LAN 2 - switch SW 2 to port 1000 BASE T (Gigabit Ethernet); LAN 3 - switch SW 3 10 BASE T ports (Ethernet);

5 LAN 4 - multiport repeater ports HUB 1000 BASE T (Gigabit Ethernet). These LANs are composed of work stations, servers and other types of terminals as shown in Fig In this configuration, we used the notation "i / j" to identify the switch port j of i. Fig Example of an extended LAN Ethernet, with a loop. Expanded LAN network in Fig. 3.1 there is a closed loop direct links of the 3 switches and HUB. Thus, if a certain PC station A, the LAN 1 emits a frame whose destination station LAN4 PC -C, there are two possible ways of transmission : the first way consists of the sequence of nodes SW1 - SW3 - HUB and the second way consists of SW1 - SW2 - HUB. In view of the fact that the hub does not take part in the formation of STP tree, because it works at the data link layer, when made according to the above notation for the next pair of ports parallel paths resulting from the PC - A to PC - C : 1 / 4-3 / 1-3 / 5 1/5-2/1-2/2. Suppose that switches identifiers ( Bridge ID ) have values in the same order relation as the numbers of switches, BIDSW1 < BIDSW2 < BIDSW3. Costs links between switches are listed in red in Fig. 3.1, the values associated with the maximum flow on each port according to Table 3.1. To simplify the analysis, we assume that the costs involved, and the administrator has not modified any port priorities. Therefore, each switch port

6 identified with a small has a higher priority. We also assume that all three switches have already enabled STP protocol instance. At startup, each switch learns station addresses active on its ports. It will be assumed that all terminals connected to nodes are active. In the example in Fig. 3.1 SW1 root node is selected because it has the minimum ID Bridge. All ports are active switch SW1 in the state Designated Port (Forwarding). The other two switches to root lodging costs will be calculated SW1 and one of them will disable a port to interrupt the loop. Fig The tree (STP) for the network in Fig Along with these calculations is appointed from each of the following switches : Root Port (one for switches SW2 and SW3 ), Designated Port (one for each common bond ) and Nondesignated ( Blocking ) Port (one for each common link here celătalt port on the Connectors 2 /2-3/5 ). Thus, SW2 port 2/1 is the cost Root Port 4 to port 1/5, and the SW3 port 3/1 Root Port is the port cost 100 1/4. In segment 2 /2-3/5 switch SW2 choose Designated Bridge because it has a lower identifier than switch SW3, at the other end of the link. The port 2/2 transitions to state Designated Port, and the decision will lock the switch SW3. Because port 2/2 remains open, the path cost to the root determined port 3/5 is 104. Way switch SW3 SW1 finds two paths to the root, one to port 3/5 via SW2 cost 104 and the other from port 3/1 directly to the root, cost 100. as they choose the path of least cost, port 3/5 is blocked (becomes Nondesignated Port ) and thus end STP tree shown in Fig Further, suppose that the administrator changes the port cost of 3/1 to SW3 104 value. Thus the lock switch SW3 decision may not be based on cost paths to the root because they have equal costs. According to the ST algorithm, where the costs are equal, the decision will be based on the minimum value corresponding ID switches emitting Bridge BPDU. In the case of the network of Fig. 8.1 SW3 seen two ways to root the same cost 104, and the best way to root SW1 is direct from port 3/1 because the path from port 3/5 is achieved via SW2 which has a Bridge ID higher than that of SW1. Thus, as in

7 the previous case, port 3/5 is locked (Port is Nondesignated) and the shaft end STP is the same (see Fig. 3.2), with the observation made on costs. To illustrate another possibility of Ports closure decision we will consider the following changes are made to the network in Fig. 3.1: cost of port 3/1 is changed to the value 105 and port 1/6 of SW1 connects directly to port 2/6 of SW2, both being used in Fig So close a new loop is formed by links 1/5-2/1 to 2/6-1/6. Configuration at after these two changes is illustrated in Fig Fig Example of an extended LAN with two loops. Loops formed by parallel connection of equal cost between two switches can be interrupted only by port identifiers as and Bridge ID transmitter is the same. In the case of the example in Fig. 3.3, the switch SW2 can see two ways to root SW1, both cost 4 with the same bridge ID for the switch emitter (here even root SW1) and therefore the only way to choose a Root Port SW2 is based at Port minimum ID SW1 emitting ports: ports fifth and sixth. Considering that Port ID identifiers have the same priority, then the port will be closed is 2/6, because the other end has a port ID (1/6) higher than the other link. In parallel, large loop is interrupted by closing port 3/1 due to the high cost (value 105). In this case, the final STP tree shown in Fig. 3.4.

8 Fig The tree (STP) for the network in Fig. 3.3.