RRPP Technology White Paper

Transcription

1 RR Technology White aper Keywords: RR, RR domain, RR ring, control VLAN, protected VLAN, master node, transit node, edge node, assistant-edge node, ring group. Abstract: The Rapid Ring rotection rotocol (RR) is a link layer protocol dedicated to Ethernet rings. This document introduces H3C s RR implementation, characteristics, and typical networking schemes. Acronyms: Acronym Full spelling RR SRT ST VLAN Rapid Ring rotection rotocol Sub Ring acket Tunnel in Major Ring Spanning Tree rotocol Virtual Local Area Network Hangzhou H3C Technologies Co., Limited 1/30

2 Table of Contents 1 Overview Background Benefits RR Implementation Basic Concepts in RR RR Domain RR Ring RR Control VLAN RR rotected VLAN Node Transit Node Edge Node and Assistant-Edge Node rimary ort and Secondary ort Common ort and Edge ort RRDUs RRDU Types RRDU Format Single Ring Fundamentals Single-Domain Single Ring Multi-Domain Single Ring Intersecting Ring Fundamentals Single-Domain Intersecting Rings Multi-Domain Intersecting Rings SRT State Detection Mechanism Application Scenarios Single-Domain Single Ring Multi-Domain Single Ring Tangent Rings Single-Domain Intersecting RR Rings Multi-Domain Intersecting RR Rings RR in Combination with ST Hangzhou H3C Technologies Co., Limited 2/30

3 1 Overview 1.1 Background Nowadays, most networks use the ring structure to improve reliability. With ring networking technologies, network devices are connected to form rings. To avoid broadcast storms, which are common in a ring network, a loop protection mechanism must be used. The IEEE spanning tree protocols have been widely used for loop protection. However, as ST topology convergence gets slower while network size is growing, transmission performance can be degraded in a large network. To remove the negative impact of network size on topology convergence and shorten topology convergence time, H3C developed RR. 1.2 Benefits RR is a link layer protocol dedicated to Ethernet rings. It can prevent broadcast storms caused by data loops when an Ethernet ring is healthy, and rapidly restore the communication paths between the nodes after a link is disconnected on the ring by bringing up the backup link. Compared with ST, RR has the following advantages: Fast topology convergence within 50 ms. Convergence time independent of Ethernet ring size. On intersecting rings, the topology change of an RR ring does not cause topology changes in the other rings, and therefore, data transmission is more stable. In addition, RR supports load sharing in Ethernet rings, which improves physical link bandwidth utilization. Hangzhou H3C Technologies Co., Ltd. 3/30

4 2 RR Implementation 2.1 Basic Concepts in RR RR Domain An RR domain identified by an integral ID defines a topology range calculated and controlled by the RR protocol. It consists of some interconnected devices with the same domain ID, control VLANs, and protected VLANs. A device can belong to multiple RR domains. An RR domain consists of the following elements: RR rings RR control VLANs RR protected VLANs nodes Transit nodes Edge nodes Assistant-edge nodes In Figure 1, Domain 1 is an RR domain that contains Ring 1 and Ring 2 formed by devices S1 through S6. Ring 1 is the primary ring, where S1 is the master node and S2, S3, and S4 are transit nodes. Ring 2 is the subring, where S6 is the master node, and S5 is a transit node. S3 and S2 are the edge node and the assistant-edge node respectively. The primary control VLAN and secondary control VLAN of Domain 1 are VLAN 3 and VLAN 4 respectively. Hangzhou H3C Technologies Co., Ltd. 4/30

5 RR Domain Outline Domain 1 S S3 3 4 Edge 4 4 S6 B S 4 Ring 1 Ring B 4 4 S3 3 S Assistant 4 4 S2 S5 - rimary ort S - Secondary ort B - Blocked ort Major Ctrl VLAN:3 Sub Ctrl VLAN:4 Figure 1 A sample RR domain RR Ring Each RR ring corresponds to a ring-shaped Ethernet topology and is identified by an integral ID. As described in the last section, an RR domain consists of a single RR ring or multiple connected RR rings. The topology calculation is actually based on RR rings. Typically, ring topologies fall into these three types: single ring, tangent rings, and intersecting rings. For each topology type, the RR domain configuration is different: All devices on the single ring are configured to be in the same RR domain. All devices on intersecting rings are also configured to be in the same RR domain. For two tangent rings, the devices on each ring are configured to be in the same RR domain. That is, two tangent rings need two different RR domains, one for each ring. In an RR domain with intersecting rings, to achieve independent topology calculation on each ring without affecting other rings and prevent loops, you need to configure one ring as the primary ring and the others as subrings. The primary ring as a whole serves as a logical node on the subrings, and protocol packets from the Hangzhou H3C Technologies Co., Ltd. 5/30

6 subrings are transparently transmitted through the primary ring. In this way, topology calculation is performed on the intersecting rings as a whole. rotocol packets of the primary ring are confined within the primary ring. The level of the primary ring is 0 and that of subrings is 1. For example, to configure Ring 1 in Figure 1 as the primary ring, you should set its ring level to 0 and set the level of all the other rings in the domain, Ring 2 in this example, to 1. By doing this, you can prevent loops in the intersecting ring topology and ensure the connectivity between nodes RR Control VLAN As described earlier, RR separates data traffic from RRDUs (RR packets) by transmitting RRDUs in dedicated VLANs called control VLANs. An RR domain is configured with one primary control VLAN and one secondary control VLAN. After you specify a VLAN as the primary control VLAN, the VLAN whose ID is one plus the primary control VLAN ID is configured as the secondary control VLAN automatically. The primary control VLAN transmits the RRDUs of the primary ring and the EDGE- HELLO messages of the subrings. The secondary control VLAN transmits the RRDUs of the subrings except the EDGE-HELLO messages. All the ports connecting devices to RR rings are assigned to control VLANs, and only such ports can be assigned to control VLANs. As shown in Figure 1, 3 and/or 4 near a port indicate the VLAN(s) the port is assigned to. RR ports on the primary ring must be assigned to both the primary control VLAN and the secondary control VLAN; RR ports on the subrings can be assigned to only the secondary control VLAN RR rotected VLAN A protected VLAN is a VLAN that transmits data packets. It can contain both RR ports and non-rr ports. A protected VLAN s forwarding status is controlled by its RR domain. Different RR domains on the same RR ring are configured with different protected VLANs, and each RR domain controls the forwarding status of ports in it independently Node Each device on an RR ring is called an RR node. On an RR ring, you must Hangzhou H3C Technologies Co., Ltd. 6/30

7 configure only one as the master node. The master node initiates ring status detection with the polling mechanism and makes operation decisions upon ring topology changes. In Figure 1, S1 is the master node on the primary ring, and S6 is the master node on the subring. A master node can be in one of the following states: Complete state The master node is in the complete state if it can receive at its secondary port the Hello packets sent out its primary port. In this case, the master node blocks the secondary port to prevent traffic loops. Failed state When a link in the ring fails, the master node is in the failed state. To avoid traffic interruption in the ring, the master node unblocks the secondary port to forward data traffic. Note: The state of the master node represents the state of the whole RR ring. That is, when the master node is in the complete (failed) state, the RR ring is also in the complete (failed) state Transit Node All the nodes except the master node on a ring are transit nodes. For example, in Figure 1, S2, S3, and S4 are transit nodes on the primary ring, and S5 is a transit node on the subring. A transit node transparently transmits Hello packets of the master node, monitors the state of its directly connected RR links, and reports link state changes (if any) to the master node to decide the actions to be taken. A transit node can be in one of the following states depending on the states of its primary and secondary ports: Link-up state When both the primary port and secondary port are up, the transit node is in the linkup state. Hangzhou H3C Technologies Co., Ltd. 7/30

8 Link-down state When either the primary port or the secondary port is down, the transit node is in the link-down state. re-forwarding state When either the primary port or the secondary port is blocked, the transit node is in the pre-forwarding state Edge Node and Assistant-Edge Node In an RR domain, of the two nodes at which the primary ring and a subring intersect, one is the edge node and the other is the assistant-edge node. You can configure either of them as the edge or the assistant-edge but must ensure that the roles of the two nodes are different. In Figure 1, S3 is the edge node, and S2 is the assistant-edge node. These two roles are significant only on subrings. Edge nodes and assistant-edge nodes are special transit nodes. An edge or edgeassistant node can be in one of the following three states depending on the state of its edge port: Link-up state When the edge port is up, the node is in the link-up state. Link-down state When the edge port is down, the node is in the link-down state. re-forwarding state When the edge port is blocked, the node is in the pre-forwarding state. The state transition of an edge or edge-assistant node is the same as that of a transit node but it is triggered by the link state change of the edge port only rimary ort and Secondary ort Of the ports that connect a node to an RR ring, one is the primary port and the other is the secondary port. You can configure them as needed. The primary and secondary ports of master nodes are different in functions. A master node sends HELLO messages out its primary port. If it can receive these HELLO Hangzhou H3C Technologies Co., Ltd. 8/30

9 messages on its secondary port, the master node considers the RR ring as complete and thus blocks the secondary port to avoid loops. If the master node fails to receive these HELLO messages within the specified period, it considers the ring as having failed and unblocks the secondary port to ensure service continuity. The primary and secondary ports of a transit node are the same in functions. In an RR domain, the primary ring is a logical node of each subring and it transmits subring RRDUs (except the EDGE-HELLO messages) transparently as data traffic. Therefore, no data packet or subring RRDU (except the EDGE-HELLO messages) can pass through a blocked port on the primary ring Common ort and Edge ort On an edge or assistant-edge node, the port connecting to the subring is called the edge port while the two ports connecting to the primary ring are called common ports. The link between the common port on the edge node and that on the assistant-edge node is called the common link. As a primary ring is considered as a logical node on its subrings, the common link is considered as an internal link of the primary ring node. Thus, the common link state changes are reported only to the master node of the primary ring. In Figure 1, on the edge node S3, the port connecting to S6 is an edge port, while the ports connecting to S4 and S2 are common ports. The link directly connecting the edge node S3 to the assistant-edge node S2 is the common link. 2.2 RRDUs RRDU Types HELLO LINK-DOWN RRDU type Description Sent regularly by a master node to check ring completeness. If the sent HELLO messages can finally reach the secondary port of the master node within the predefined period, the ring is considered complete; if not, the ring is considered open, in which case a link may have failed on the ring. Sent by a transit, edge, or edge-assistant node to report link failure to the master node. Hangzhou H3C Technologies Co., Ltd. 9/30

10 RRDU type COMMON-FLUSH-FDB COMLETE-FLUSH-FDB Description Sent by a master node to notify the transit nodes to update their MAC address tables and AR/ND tables when it transitions to the failed state. The nodes on the primary ring must update their MAC address table and AR/ND tables after receiving COMMON-FLUSH-FDB messages, even if they are from the master node on a subring. Sent by a master node to notify the transit nodes to update their MAC address tables and AR/ND tables when it transitions to the complete state. The transit nodes thus transition to the link-up state, unblocking the temporarily blocked ports. For the nodes on the primary ring, if the sending master node is on a subring, they will update the MAC address tables and AR/ND tables, but will not unblock the blocked ports. Sent by the edge node of a subring and received by the assistant-edge node of the same subring to check whether the SRTs of the subring are in good condition. EDGE-HELLO MAJOR-FAULT The edge node periodically sends EDGE-HELLO messages out the two common ports to the assistant-edge node across the primary ring. If the assistant-edge node receives the messages, the SRTs are considered as in good condition; if the assistant-edge node does not receive the messages within a specified period of time, the SRTs are considered as faulty. Sent by an assistant-edge node to report SRT failure to the edge node. Upon receiving a MAJOR- FAULT message, the edge node blocks its edge port. Hangzhou H3C Technologies Co., Ltd. 10/30

11 2.2.2 RRDU Format Destination MAC Address (6 bytes) Source MAC Address (6 bytes) EtherType RI VLAN ID Frame Length DSA/SSA CONTROL OUI = 0x00e02b 0x00bb 0x99 0x0b RR Length RR_VER RR Type Domain ID Ring ID 0x0000 SYSTEM_MAC_ADDR (6 bytes) HELLO_TIMER FAIL_TIMER 0x00 LEVEL 0x0000 0x0000 RESERVED(0x ) RESERVED(0x ) RESERVED(0x ) RESERVED(0x ) RESERVED(0x ) RESERVED(0x ) Figure 2 RR RRDU format The following table describes each field in an RRDU: Field Length (in bits) Description Destination Address MAC 48 Destination MAC address, in the range of 0x000FE to 0x000FE Source Mac Address 48 EtherType 8 RI 4 VLAN ID 12 Frame Length 16 DSA/SSA 16 Source MAC address, fixed to 0x000fe203fd75. Encapsulation type, fixed to 0x8100 indicating tagged encapsulation. Class of service (CoS) priority, fixed to 0xe0. ID of the VLAN to which the packet belongs. Ethernet frame length, fixed to 0x48. Destination service access point/source service access point, fixed to 0xaaaa. CONTROL 8 Fixed to 0x03. Hangzhou H3C Technologies Co., Ltd. 11/30

12 Field Length (in bits) Description OUI 24 Fixed to 0x00e02b. RR Length 16 RR protocol data unit length, fixed to 0x40. RR_VER 16 RR version, 0x0001 currently. RR Type 8 Domain ID 16 Ring ID 16 SYSTEM_MAC_ADDR 48 HELLO_TIMER 16 FAIL_TIMER 16 LEVEL 8 RRDU type: 5: HELLO 6: COMLETE-FLUSH-FDB 7: COMMON-FLUSH-FDB 8: LINK-DOWN 10: EDGE-HELLO 11: MAJOR-FAULT ID of the RR ring to which the packet belongs. ID of the RR ring to which the packet belongs. Bridge MAC address of the node sending the packet. Hello timer setting (in seconds) of the sending node. Fail timer setting (in seconds) of the sending node. Level of the RR ring to which the packet belongs. 2.3 Single Ring Fundamentals Single-Domain Single Ring This section describes how RR works and how ring topology converges by analyzing the Ethernet ring status change from complete to failed, and then back to complete. 1. Ring status detection and related operations RR uses a polling mechanism to check ring completeness: the master node sends Hello messages to the ring regularly. These Hello messages pass through each Hangzhou H3C Technologies Co., Ltd. 12/30

13 transit node on the ring in turn. If they can finally reach the secondary port of the master node, the ring is considered complete. In this case, to avoid broadcast loops on the ring, the master node keeps its secondary port blocked. Figure 3 shows a closed RR ring. Single Ring Complete state Complete state S port blocked S B HELLO - rimary ort S - Secondary ort B - Blocked ort Data acket Control acket Figure 3 A complete RR ring 2. Fault detection and related operations Ring faults can be detected by using one of the following mechanisms: olling mechanism Link down alarm mechanism 1) olling mechanism RR uses a polling mechanism to check ring faults. With this mechanism, a master node sends Hello messages to the ring regularly. These Hello messages pass through each transit node on the ring in turn. If they fail to reach the secondary port of the master node within a specified period of time, the ring is considered open, in which case at least one link may have failed on the ring. In this case, the master node transitions to the failed state, unblocks the secondary port, and sends COMLETE-FLUSH-FDB messages out its primary and secondary ports to instruct all the transit nodes to update their MAC address table entries and Hangzhou H3C Technologies Co., Ltd. 13/30

14 AR/ND entries. 2) Link down alarm mechanism A node is always monitoring its own port link status, and takes immediate actions upon detecting a failed port. When a master node s primary port goes down, the master node is able to sense the link fault by itself. It immediately unblocks its secondary port, and sends a COMLETE-FLUSH-FDB message out its secondary port to instruct all the transit nodes to update their MAC address table entries and AR/ND entries. When one port on a transit node goes down, the transit node sends a LINK- DOWN message out the other RR port that is still up to notify the master node, as shown in Figure 4. Upon receiving the message, the master node unblocks its secondary port and transitions to the failed state. To avoid packet direction errors due to the topology change, the master node updates its own MAC address table and AR/ND entries and sends a COMMON-FLUSH-FDB message out its primary and secondary ports to instruct all the transit nodes to update their MAC address table entries and AR/ND entries. See Figure 5 for the process of a master node transitioning from the complete state to the failed state. Link failure Transit Send LINK-DOWN to LINK-DOWN S B - rimary ort S - Secondary ort B - Blocked ort Data acket Control acket Figure 4 A transit node sends a LINK-DOWN message to the master node Hangzhou H3C Technologies Co., Ltd. 14/30

15 Transfer to Failed State Transfer to Failed State Unblock S port S COMMON-FLUSH-FDB - rimary ort S - Secondary ort B - Blocked ort Data acket Control acket Figure 5 The master node transitions to the failed state The link-down alarm mechanism provides faster fault protection than the polling mechanism. However, if LINK-DOWN messages are lost on the way to the master node, the link faults will be detected later by the master node using the polling mechanism. If the master node fails to receive on its secondary port the Hello messages it sends out within a period of time specified by the Fail timer, it considers the ring topology as faulty, and then takes related actions as described earlier. 3. Fault recovery detection and related operations When a failed port on a transit node recovers, the master node cannot be notified of the recovery immediately. Hence, the master port s secondary port is still up. If the transit node transitions to the Link-Up state at once, a temporary loop will be created on the ring. Therefore, when a Link-Down transit node s primary and secondary ports are both recovered, the transit node blocks them immediately and transitions to the pre-forwarding state, as shown in Figure 6. At this point, the ring is not recovered yet. The master node initiates the ring recovery process. Hangzhou H3C Technologies Co., Ltd. 15/30

16 Transit block restored port temporarily B B Transfer to preforwarding state & block restored port at Failed State S - rimary ort S - Secondary ort B - Blocked ort Data acket Control acket Figure 6 A transit node blocks its recovered ports and transitions to the pre-forwarding state When all the links on the ring recover and the master node is able to receive its own Hello packets again, it blocks the secondary port and transitions back to the complete state. Because of the ring topology change, the master node needs to update its MAC address table entries and AR/ND entries, and sends COMLETE-FLUSH-FDB messages to instruct all the transit nodes to update their MAC address table entries and AR/ND entries. When receiving the COMLETE-FLUSH-FDB message from the master node, the transit nodes in the pre-forwarding state transition to the Link-Up state. Thus the ring is recovered, as shown in Figure 7. Hangzhou H3C Technologies Co., Ltd. 16/30

17 transfer to Complete state Transfer to Link-Up State & Unblock port Transfer to Complete state & block S port S B COMLETE-FLUSH-FDB - rimary ort S - Secondary ort B - Blocked ort Data acket Control acket Figure 7 How a ring recovers In case the COMLETE-FLUSH-FDB messages fail to reach a transit node, the transit node in the pre-forwarding state unblocks the blocked port automatically, updates its MAC address table entries and AR/ND entries and transits to the link-up state if it fails to receive any COMLETE-FLUSH-FDB messages from the master node within the period of time specified by the Fail timer Multi-Domain Single Ring If traffic of multiple VLANs exists on an RR ring, you can configure multiple RR domains on the RR ring, with each domain transmitting traffic for different VLANs (protected VLANs). In this way, data traffic of different VLANs is transmitted along different paths on the ring, thus achieving load sharing. Hangzhou H3C Technologies Co., Ltd. 17/30

18 Figure 8 Multi-domain single ring As shown in Figure 8, Ring 1 is configured as the primary ring in both Domain 1 and Domain 2. Domain 1 and Domain 2 have different protected VLANs. In Domain 1, Device A is configured as the primary node on Ring 1, while in Domain 2, Device B is configured as the primary node on Ring 1. Under such configurations, traffic of different VLANs is transmitted along different paths. 2.4 Intersecting Ring Fundamentals Single-Domain Intersecting Rings In a single-domain intersecting rings topology, the implementation of the primary ring and the fault detection mechanism used by the subrings master nodes are the same as those used on a single-ring topology. The difference is that the SRT state detection mechanism is used on a multi-ring topology to prevent data loops on the subrings when both SRTs are disconnected. Before the master node on a subring unblocks its secondary port, the edge node blocks its edge port, thus preventing broadcast loops on the subrings. For more information about the SRT state detection mechanism, refer to SRT State Detection Mechanism. Additionally, when the nodes on the primary ring receive COMMON-FLUSH-FDB or COMLETE-FLUSH-FDB messages from a subring, they must update their MAC address table entries and AR/ND entries. However, the COMLETE-FLUSH-FDB messages from a subring cannot make a transit node on the primary ring to unblock its temporarily-blocked port. Hangzhou H3C Technologies Co., Ltd. 18/30

19 Domain 1 Device A Device B Edge node node Transit node Ring 1 Ring 2 node Device E Device D Device C Assistant edge node Figure 9 Single-domain intersecting rings Multi-Domain Intersecting Rings If traffic of multiple VLANs exists in an intersecting RR rings topology, you can configure multiple RR domains for the intersecting rings, with each domain transmitting traffic for the specified protected VLANs. Each RR domain in the multidomain topology works in the same way as a single domain does. In this way, data traffic of different VLANs is transmitted along different paths on the ring, thus achieving load sharing. Device A Device B Device E Domain 1 Ring 1 Ring 2 Domain 2 Device D Device C Figure 10 Multi-domain intersecting rings As shown in Figure 10, Ring 1 and Ring 2 are respectively configured as the master ring and subring in both Domain 1 and Domain 2. They have different protected Hangzhou H3C Technologies Co., Ltd. 19/30

20 VLANs. In Domain 1, Device A is configured as the master node of Ring 1, while in Domain 2, Device D is configured as the master node of Ring 1. In both Domain 1 and Domain 2, Device E is configured as the master node of Ring 2. However, for Domain 1 and Domain 2, the primary and secondary ports on Device E are different. In this way, traffic of different VLANs can be transmitted along different paths in the primary ring and subrings, thus achieving load sharing on the intersecting rings SRT State Detection Mechanism 1. Introduction to SRT state detection mechanism SRTs are tunnels for subring packets on the primary ring. The primary ring as a whole serves as a logical node on the subrings, and protocol packets from the subrings are transparently transmitted through the primary ring. The primary ring forwards protocol packets (except EDGE-HELLO packets) from the subrings as data packets. Each subring has two SRTs. As shown in Figure 11, the two SRTs for subrings Ring 2 and Ring 3 are S3-S2 and S3-S4-S1-S2. When the primary ring is healthy, the secondary port of its master node is blocked, in which case only the S3-S2 tunnel is available. When fault occurs on the S3-S4-S1-S2 tunnel, the S3-S2 tunnel is available; when fault occurs on the S3-S2 tunnel, the S3-S4-S1-S2 tunnel is available. In other words, of a subring s two SRTs, only one of them is available at any point of time, thus preventing data loops on the primary ring for subring protocol packets. When both the SRTs of a subring fail, and the subring s master node fails to receive its own Hello packets within the period of time specified by the Fail timer, the subring s master node unblocks its secondary port. In this way, the subring can restore communication to the maximum extent without creating data loops. This works fine in a common RR network topology but not in a dual-homed RR ring topology. As shown in Figure 11, the two subrings Ring 2 and Ring 3 are interconnected through the edge node and assistant-edge node and form a loop naturally. Therefore, data loops are unavoidable when the secondary ports on the master nodes of the subrings are unblocked after the two SRTs on Ring 1 went down. The arrows in the figure indicate traffic directions. Hangzhou H3C Technologies Co., Ltd. 20/30

21 Figure 11 Loops in a dual-homed ring topology without SRT state detection To address the problem, the SRT state detection mechanism is introduced, which is carried out by the edge and assistant-edge nodes together. When the edge node detects that both the SRTs are down, it blocks the edge ports on the edge nodes before the secondary ports on the master nodes of the subrings are both unblocked. Thus, loops are avoided. Figure 12 shows how loops are removed with the SRT state detection mechanism. Figure 12 Loop removal in a dual-homed ring topology with SRT state detection Hangzhou H3C Technologies Co., Ltd. 21/30

22 2. SRT state detection process In the SRT state detection mechanism, the edge node initiates SRT state detection and determines the actions to take; the assistant-edge node monitors SRT state and notifies the edge node of SRT state changes, if any. The following subsections describe how the mechanism works. 1) Detecting SRT state The edge node of a subring periodically sends EDGE-HELLO messages destined for the assistant-edge node out the two ports connecting the subring to the SRTs. The EDGE-HELLO messages travel through each node on the SRTs, as shown in Figure 13. If at least one SRT is normal for transmitting subring protocol packets, the assistant-edge node can receive these EDGE-HELLO messages within the specified period of time. If the two SRTs are both disconnected and the subring protocol packets cannot travel through the primary ring, the assistant-edge node cannot receive these EDGE-HELLO messages within the specified period of time. EDGE-HELLO Edge B S Major ring Sub ring B S Assistant - rimary ort S - Secondary ort B - Blocked ort Data acket Control acket Figure 13 The edge node sends EDGE-HELLO messages 2) Blocking the edge port of the edge node when the SRTs are disconnected Upon detecting that the two SRTs are both disconnected, the assistant-edge node sends MAJOR-FAULT messages out the edge port to the edge node through the subring. If the subring is normal, the edge node can receive these MAJOR-FAULT Hangzhou H3C Technologies Co., Ltd. 22/30

23 messages. Upon receiving the MAJOR-FAULT messages, the edge node blocks its edge port as shown in Figure 14. If the subring fails, the edge port of the edge node will not be blocked. MAJOR-FAULT messages are sent periodically. If the edge node receives them, its edge port stays blocked; if the edge node fails to receive any MAJOR-FAULT message within the specified period of time, its edge port is unblocked automatically. Edge B B S Major ring Sub ring S Assistant MAJOR-FAULT - rimary ort S - Secondary ort B - Blocked ort Data acket Control acket Figure 14 The edge node blocks its edge port upon receiving MAJOR-FAULT messages 3) Transitioning to the failed state when the subring fails As the master node on the subring cannot receive the HELLO messages it sent out due to disconnection of the two SRTs, it unblocks its secondary port and transitions to the failed state, as shown in Figure 15. Hangzhou H3C Technologies Co., Ltd. 23/30

24 Edge B S Major ring Sub ring S Assistant - rimary ort S - Secondary ort B - Blocked ort Data acket Control acket Figure 15 The subring s master node transitions to the failed state due to SRT disconnection 4) SRT recovery When the primary ring recovers, the SRTs also recover. Therefore, the assistantedge node stops sending MAJOR-FAULT messages. In this case, if the subring is normal, its master node can receive its own Hello packets again, and thus blocks its secondary port and transitions to the complete state, as shown in Figure 16. Hangzhou H3C Technologies Co., Ltd. 24/30

25 Edge HELLO B B S Major ring Sub ring S Assistant - rimary ort S - Secondary ort B - Blocked ort Data acket Control acket Figure 16 SRT recovery After the subring is recovered, its master node sends COMLETE-FLUSH-FDB messages out the primary port. Upon receiving the COMLETE-FLUSH-FDB messages, the edge node unblocks its edge port if the port is blocked, as shown in Figure 17. Thus, communication in the entire network is recovered. Edge COMLETE-FLUSH-FDB B S Major ring Sub ring S Assistant - rimary ort S - Secondary ort B - Blocked ort Data acket Control acket Figure 17 The edge node of the subring unblocks the edge port Hangzhou H3C Technologies Co., Ltd. 25/30

26 If a fault is present on the subring when the SRTs recover, the subring cannot be recovered. In this case, the master node of the subring does not send COMLETE- FLUSH-FDB messages and the blocked edge port of the edge node will be unblocked only after the Fail timer expires. 3. Ring group In SRT state detection, the edge node and assistant-edge node of subrings respectively sends and receives EDGE-HELLO packets frequently. As shown in 1. Figure 11 in the multi-ring dual-homed topology, if you configure S2 and S3 as the edge node and assistant-edge node of both Ring 2 and Ring 3, S2 needs to send EDGE-HELLO packets for both Ring 2 and Ring 3, while S3 needs to receive EDGE- HELLO packets for both Ring 2 and Ring 3. The more subrings are configured, the more EDGE-HELLO packets will occur in the network, thus increasing the CU load of the devices. To reduce Edge-Hello traffic, you can configure a group of subrings on the edge node or assistant-edge node. A ring group configured on the edge node is called an edge node ring group, and a ring group configured on an assistant-edge node is called an assistant-edge node ring group. In an edge node ring group, only an active subring with the smallest domain ID and ring ID sends EDGE-HELLO packets; in an assistant-edge node ring group, a random active subring receives the EDGE-HELLO packets and passes the information to other active subrings. In this way, after subring groups are configured on the edge node and assistant-edge node, only one subring in each group sends/receives EDGE-HELLO packets, thus reducing the device CU load significantly. You must configure a device as the edge node of these subrings, and another device as the assistant-edge node of these subrings. Additionally, these subrings must have the same SRTs. Hangzhou H3C Technologies Co., Ltd. 26/30

27 3 Application Scenarios 3.1 Single-Domain Single Ring Domain 1 Device A Device B node Transit node Ring 1 Transit node Transit node Device D Device C Figure 18 Network diagram for a single-domain single-ring network A single-domain single-ring network has only one ring. Therefore, you need to define only one RR domain and one RR ring. A single-domain single-ring network features fast response to topology changes and fast topology convergence. 3.2 Multi-Domain Single Ring Figure 19 Network diagram for a multi-domain single-ring network Hangzhou H3C Technologies Co., Ltd. 27/30

28 A multi-domain single-ring network has only one ring, but multiple VLANs for the purpose of load sharing. Multiple RR domains are configured in the network, each having its own protected VLANs. In addition, an RR ring has different master nodes in different RR domains or has the same master node but different primary/secondary ports in different RR domains. Therefore, the protected VLANs of different RR domains have different logical topologies. 3.3 Tangent Rings node Device E node Transit node Ring 2 Domain 2 Device A Device B Device F Ring 1 Transit node Domain 1 Device D Transit node Device C Transit node Figure 20 Network diagram for a tangent-ring network A tangent-ring network contains two or more rings. Every two rings have only one common node. For each ring, you are required to define an independent RR domain. The tangent-ring topology is suitable for large-scale networks that require networks at the same level to be managed as independent domains. Hangzhou H3C Technologies Co., Ltd. 28/30

29 3.4 Single-Domain Intersecting RR Rings Domain 1 Device A Device B Edge node node Transit node Ring 1 Ring 2 node Device E Device D Device C Assistant edge node Figure 21 Networking diagram for single-domain intersecting rings A single-domain intersecting-ring network contains two or more rings. Every two rings have two common nodes. In this case, you can define one RR domain. In the domain, configure a ring as the primary ring and the other rings as subrings. A typical single-domain intersecting-ring topology is dual-homed rings where the master node of a subring is dually uplinked to the primary ring through the edge node and the assistant-edge node for uplink backup. 3.5 Multi-Domain Intersecting RR Rings Device A Device B Device E Domain 1 Ring 1 Ring 2 Domain 2 Device D Device C Figure 22 Networking diagram for multi-domain intersecting rings Hangzhou H3C Technologies Co., Ltd. 29/30

30 A multi-domain intersecting-ring network contains two or more rings. Every two rings have two common nodes. To achieve load sharing when data traffic of multiple VLANs exists in the network, you can configure multiple RR domains in the network, each having its own protected VLANs. In addition, an RR ring has different master nodes in different RR domains or has the same master node but different primary/secondary ports in different RR domains. In this way, the protected VLANs of different RR domains can have different logical topologies. 3.6 RR in Combination with ST RR is mutually exclusive with ST on a port because RR may conflict with ST in port state calculation. An RR ring can be connected to an ST ring only in the tangent mode, where no common ports exist between rings. Figure 23 RR in combination with ST Copyright 2008 Hangzhou H3C Technologies Co., Ltd. All Rights Reserved Extracting and copying partial or whole contents of the document without H3C s written permission is prohibited. The information is this document is subject to due modification. Hangzhou H3C Technologies Co., Ltd. 30/30