Overview

Traditional Layer 2 switching environments consist of Layer 2 devices (such as switches) that partition data into broadcast domains. The broadcast domains can be created through physical topologies or through virtual local area networks (VLANs).

On Juniper Networks routers you can logically configure broadcast domains within virtual switch routing instances, VPLS routing instances, or bridge domains. The individual routing instances or bridge domains are differentiated through VLAN IDs, and these instances or domains function much like traditional VLANs.

As is the case with traditional VLANs, in order to avoid loops within bridge domains, you must configure some kind of loop prevention mechanism. Spanning Tree Protocol (STP), Rapid Spanning Tree Protocol (RSTP), and Multiple Spanning Tree Protocol (MSTP) are supported on Juniper Networks routers.

These protocols prevent loops within a local area network (LAN) or VLAN by creating a tree topology in which there is only one path from each source to each destination. Before going into configuration details, let us look more closely at how the different loop prevention protocols work.

STP

STP is the simplest loop prevention protocol and is the basis for RSTP and MSTP. As is the case with other spanning-tree protocols, STP uses bridge protocol data units (BPDUs) to identify a network's tree topology. There are two types of BPDUs: configuration BPDUs and topology change notification (TCN) BPDUs. Configuration BPDUs determine the tree topology of a LAN.

An STP tree topology can be compared to an actual tree. There is a root device and there are leaf devices. All leaf devices calculate the best path to the root device and place their ports in blocking or forwarding states based on the best path to the root. This prevents loops. Unlike a tree, however, the root device in an STP LAN can change. So if the root device goes down, the leaf devices will recalculate the network and move the root to a device that is functioning properly.

The root device is determined by comparing bridge IDs of the devices. The bridge IDs consist of the bridge priority (which you can configure) and the MAC address of the bridge. The device with the lowest bridge ID becomes the root device of an STP topology.

Every nonroot device on the LAN has a root port. The root port is the “best path” to the root device. All other ports are either blocked or are the designated port for a LAN segment between two nonroot devices.

If a path is faulty, the root port does not receive configuration BPDUs and eventually the BPDUs time out. If the configuration BPDUs time out, the device sends BPDUs announcing the next best device as the root, and the process begins again.

RSTP

Although STP provides basic loop prevention functionality, it does not provide fast network convergence when there are topology changes. The reason is that STP uses a slower state transition process, meaning that a device must reinitialize every time a topology change occurs, starting in listening state and transitioning through to learning state and eventually to forwarding or blocking state. When default values are used for the maximum age (20 seconds) and forward delay (15 seconds), it takes 50 seconds for the device to converge. This causes network delay, because no data traffic can traverse this device until the topology is readjusted. RSTP greatly decreases the state transition time.

RSTP provides a faster convergence time and better network stability than STP because it allows newly elected root or forwarding ports to enter forwarding states more rapidly.

With STP, a nonroot device generates configuration BPDUs only when it receives configuration BPDUs on its root port. In contrast, you configure hello times for RSTP devices. An RSTP device generates configuration messages once every hello time interval, even if it does not receive a configuration BPDU on its root port. If an RSTP device does not receive a configuration message from its neighbor after an interval of three hello times, it assumes it has lost connection with that neighbor.

A nonroot device running RSTP has a root port, which is the “best path” to the root device; a designated port, which is the shortest path connection to the root device for a LAN segment between two nonroot devices, an alternate port, which provides an alternate root port, and a backup port, which provides an alternate designated port.

Port assignments change through proposal-agreement mechanisms. When a root port or a designated port fails on a device, the device generates a configuration message with the proposal bit set. Once its neighbor device receives this message, it verifies that this configuration message is better than the one saved for that port and then it starts a synchronizing operation to ensure that all of its ports are in sync with the new information.

Similar waves of proposal agreement handshake messages propagate toward the leaves of the network, restoring the connectivity very quickly after a topology change (in a well-designed network that uses RSTP, network convergence can take as little as 0.5 seconds). If a device does not receive an agreement to a proposal message it has sent, it returns to the original IEEE 802.D convention.

MSTP

Although RSTP provides faster convergence time than STP, it still does not solve a problem that is inherent in STP: all bridges within a LAN must share the same spanning tree. This is especially problematic if you want to configure VLANs. In STP and RSTP, redundant links in VLANs cannot be blocked, and all traffic from all VLANs must be forwarded along the same tree.

MSTP uses the better convergence functionality of RSTP and also provides a way for data from different VLANs to be forwarded along their own paths. This functionality allows for better load sharing for redundant links.

MSTP divides a network into regions. Each region can have a maximum of 64 instances of spanning tree. Figure 88 shows an example of a network running MSTP. In the figure, there are three MST regions, each containing three devices running MSTP. A Common Spanning Tree (CST) spans all MSTP regions, and the regions communicate among themselves using the CST. The Common and Internal Spanning Tree (CIST) refers to instance 0. This is called Common Spanning Tree (CST) across the regions and Internal Spanning Tree (IST) within a region. This is used for traffic within the region for VLANs that are not covered by any MSTI and is used for inter-region traffic for all VLANs.

Each device in each MST region must have the same region name, the same VLAN-to-instance mapping configuration, and the same MSTP revision level.