Virtual link aggregations across multiple fabric switches

Abstract

One embodiment of the present invention provides a switch. The switch is configurable to be a member of a first fabric switch. The switch includes a link aggregation module. During operation, the link aggregation module marks an ingress-switch field of a frame with a virtual switch identifier. This virtual switch identifier is associated with the switch and a second switch, which is a member of a second fabric switch, and is from a range of identifier associated with the first fabric switch and the second fabric switch. Each of the first fabric switch and the second fabric switch is operable to accommodate a plurality of switches and operate as a single switch.

Description

BACKGROUND

Field

The present disclosure relates to network management. More specifically, the present disclosure relates to a method and system for facilitating link aggregation from one device to multiple fabric switches.

Related Art

The relentless growth of the Internet has brought with it an insatiable demand for bandwidth. As a result, equipment vendors race to build larger, faster, and more versatile switches to move traffic. However, the size of a switch cannot grow infinitely. It is limited by physical space, power consumption, and design complexity, to name a few factors. More importantly, because an overly large system often does not provide economy of scale due to its complexity, simply increasing the size and throughput of a switch may prove economically unviable due to the increased per-port cost.

A flexible way to improve the scalability of a switch system is to build a fabric switch. A fabric switch is a collection of individual member switches. These member switches form a single, logical switch that can have an arbitrary number of ports and an arbitrary topology. As demands grow, customers can adopt a “pay as you grow” approach to scale up the capacity of the fabric switch.

Meanwhile, layer-2 (e.g., Ethernet) switching technologies continue to evolve. More routing-like functionalities, which have traditionally been the characteristics of layer-3 (e.g., Internet Protocol or IP) networks, are migrating into layer-2. Notably, the recent development of the Transparent Interconnection of Lots of Links (TRILL) protocol allows Ethernet switches to function more like routing devices. TRILL overcomes the inherent inefficiency of the conventional spanning tree protocol, which forces layer-2 switches to be coupled in a logical spanning-tree topology to avoid looping. TRILL allows routing bridges (RBridges) to be coupled in an arbitrary topology without the risk of looping by implementing routing functions in switches and including a hop count in the TRILL header.

As more mission-critical applications are being implemented in data communication networks, high-availability operation is becoming progressively more important as a value proposition for network architects. It can be desirable to divide a conventional aggregated link (from one device to another) among multiple network devices, often belonging to different fabric switches, such that unavailability of one fabric switch would not affect the operation of the multi-homed device.

While a fabric switch brings many desirable features to a network, some issues remain unsolved when end devices are coupled to multiple fabric switches. Particularly, when an end device is coupled to multiple fabric switches using link aggregation, existing technologies do not provide a scalable and flexible solution that takes full advantage of a fabric switch.

SUMMARY

One embodiment of the present invention provides a switch. The switch is configurable to be a member of a first fabric switch. The switch includes a link aggregation module. During operation, the link aggregation module marks an ingress-switch field of a frame with a virtual switch identifier. This virtual switch identifier is associated with the switch and a second switch, which is a member of a second fabric switch, and is from a range of identifier associated with the first fabric switch and the second fabric switch. Each of the first fabric switch and the second fabric switch is operable to accommodate a plurality of switches and operate as a single switch.

In a variation on this embodiment, the switch also includes a packet processing module which identifies the virtual switch identifier as the egress switch identifier of a frame and, in response, determines that the switch is the egress switch for the frame.

In a variation on this embodiment, the switch also includes a suppression module which identifies the virtual switch identifier as egress switch identifier of a broadcast, unknown unicast, or multicast frame, and, in response, precludes the switch from determining a port associated with a local end device as the output port for the frame.

In a variation on this embodiment, the switch also includes a multicast module which identifies a frame to be a broadcast, unknown unicast, or multicast frame and, in response, mark an egress-switch field of the frame with a virtual root switch identifier. This virtual root switch identifier is associated with respective multicast root switches of the first fabric switch and the second fabric switch.

In a further variation, the switch also includes a failure detection module which, in response to detecting unavailability of the multicast root switch of the first fabric switch, identifies a new multicast root switch and associates the virtual root switch identifier with the new multicast root switch of the first fabric switch.

In a variation on this embodiment, the switch is a routing bridge (RBridge) operable in accordance with Transparent Interconnection of Lots of Links (TRILL) protocol.

In a further variation, the switch also includes a TRILL-supported edge port and a TRILL primary module. The TRILL-supported edge port is operable as an output port for frames destined outside of the first fabric switch. During operation, in response to determining a TRILL-encapsulated frame to be a broadcast, unknown unicast, or multicast frame, the TRILL primary module precludes the switch from removing TRILL encapsulation from the TRILL-encapsulated frame and determines the TRILL-supported edge port as the output port for the TRILL-encapsulated frame.

In a variation on this embodiment, the switch also includes a failure detection module which disassociates the switch from the virtual switch identifier in response to detecting unavailability of the second switch and marks an egress-switch field of a frame, which is received after detecting unavailability of the second switch, with a switch identifier of the local switch.

In a variation on this embodiment, the switch also includes a control module operable, which runs a control plane with automatic configuration capabilities based on a protocol associated with the first fabric switch and operate the first fabric switch as a single Ethernet switch based on the automatic configuration capabilities of the control plane. The control module also receives an automatically assigned identifier corresponding to the Ethernet switch and joins the first fabric switch via the control plane.

One embodiment of the present invention provides a switch. The switch is configurable to be a member of a first fabric switch. The switch includes a link aggregation module. During operation, the link aggregation module operates a first trunked link of the switch in conjunction with a second trunked link of a second switch of the first fabric switch as a virtual link aggregation. The virtual link aggregation is mapped to the switch and the second switch. A second fabric switch is reachable via the first and second trunked links. Each of the first fabric switch and the second fabric switch is operable to accommodate a plurality of switches and operate as a single switch.

In a variation on this embodiment, the virtual link aggregation is mapped to the switch based on a hash function.

In a variation on this embodiment, the switch selects respective output ports of the switch for the frames destined to the second fabric switch such that the frames are distributed across links of the first trunked link, thereby spraying the frames across the links of the first trunked link.

BRIEF DESCRIPTION OF THE FIGURES

FIG. 1A illustrates an exemplary virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention.

FIG. 1B illustrates an exemplary virtual link aggregation across multiple fabric switches coupled via TRILL forwarding link(s), in accordance with an embodiment of the present invention.

FIG. 2A presents a flowchart illustrating the process of forwarding a frame received via an edge port at a partner routing Bridge (RBridge) which participates in a virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention.

FIG. 2B presents a flowchart illustrating the process of an RBridge forwarding a TRILL-encapsulated unicast frame, in accordance with an embodiment of the present invention.

FIG. 3A presents a flowchart illustrating the process of forwarding a frame belonging to broadcast, unknown unicast, and multicast (BUM) traffic received via an edge port at a partner RBridge which participates in a virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention.

FIG. 3B presents a flowchart illustrating the process of an RBridge forwarding a TRILL-encapsulated frame destined to a virtual root RBridge, in accordance with an embodiment of the present invention.

FIG. 3C presents a flowchart illustrating the process of an RBridge forwarding a TRILL-encapsulated frame belonging to BUM traffic, in accordance with an embodiment of the present invention.

FIG. 4A illustrates exemplary failure scenarios associated with a virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention.

FIG. 4B presents a flowchart illustrating the process of handling a failure that affects a partner RBridge which participates in a virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention.

FIG. 4C presents a flowchart illustrating the process of handling a failure that affects a root RBridge which is associated with a virtual root RBridge, in accordance with an embodiment of the present invention.

FIG. 5A illustrates an exemplary hybrid virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention.

FIG. 5B illustrates an exemplary packet forwarding via a hybrid virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention.

FIG. 6 illustrates an exemplary RBridge which supports virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention.

In the figures, like reference numerals refer to the same figure elements.

DETAILED DESCRIPTION

The following description is presented to enable any person skilled in the art to make and use the invention, and is provided in the context of a particular application and its requirements. Various modifications to the disclosed embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the present invention. Thus, the present invention is not limited to the embodiments shown, but is to be accorded the widest scope consistent with the claims.

Overview

In embodiments of the present invention, the problem of providing a scalable and flexible way of provisioning a virtual link aggregation across multiple fabric switches is solved by forming a logical, virtual switch and assigning a virtual switch identifier from a global identifier range known to the fabric switches. For example, if the communication within the fabric switches is based on Transparent Interconnection of Lots of Links (TRILL) protocol, when an end device is coupled to two separate routing bridges (RBridges) belonging to two separate fabric switches and the links to these RBridges form an aggregate link, a virtual RBridge identifier (ID) is generated, and the end device is considered to be logically coupled to the virtual RBridge. In the following description, RBridges which participate in a virtual link aggregation and form a virtual RBridge are referred to as “partner RBridges.”

If the virtual RBridge identifier associated with the virtual link aggregation is from a local identifier range of a fabric switch, partner RBridges, which are members of other fabric switches, do not recognize that virtual RBridge identifier as its own. As a result, these partner RBridges may not be able to perform operations associated with efficient deployment of a link aggregation, such as source suppression of BUM traffic. Furthermore, to distribute of BUM traffic in a fabric switch, a partner RBridge sets the destination RBridge identifier to be the RBridge identifier of the root RBridge (i.e., the root of a multicast tree) of the fabric switch. In some embodiments, this root RBridge is responsible for distributing BUM traffic in a fabric switch and usually is specific to the fabric switch. Hence, for a virtual link aggregation spanning multiple fabric switches, a fabric-specific root RBridge may not support BUM traffic distribution in the multiple fabric switches.

In embodiments of the present invention, the aforementioned problems are solved by assigning the virtual RBridge identifier, which is associated with a virtual link aggregation spanning multiple fabric switches, from a global identifier range. Identifiers belonging to this global identifier range are known to the multiple fabric switches. This allows partner RBridges in different fabric switches to recognize the same virtual RBridge identifier and perform source suppression for BUM traffic originating from the end device associated with the virtual link aggregation. Furthermore, a virtual root RBridge is created for forwarding BUM traffic in the fabric switches. This virtual root RBridge is associated with the root RBridge of respective fabric switch. A partner RBridge forwards BUM traffic toward the virtual root RBridge, which, in turn, is received by respective root RBridge in respective fabric switch. This enables a respective root RBridge to distribute BUM traffic within corresponding fabric switch.

In some embodiments, the fabric switch is an Ethernet fabric switch. In an Ethernet fabric switch, any number of switches coupled in an arbitrary topology may logically operate as a single switch. Any new switch may join or leave the fabric switch in “plug-and-play” mode without any manual configuration. A fabric switch appears as a single logical switch to an external device. In some further embodiments, the fabric switch is a TRILL network and a respective member switch of the fabric switch is a TRILL RBridge.

It should be noted that a fabric switch is not the same as conventional switch stacking. In switch stacking, multiple switches are interconnected at a common location (often within the same rack), based on a particular topology, and manually configured in a particular way. These stacked switches typically share a common address, e.g., IP address, so they can be addressed as a single switch externally. Furthermore, switch stacking requires a significant amount of manual configuration of the ports and inter-switch links. The need for manual configuration prohibits switch stacking from being a viable option in building a large-scale switching system. The topology restriction imposed by switch stacking also limits the number of switches that can be stacked. This is because it is very difficult, if not impossible, to design a stack topology that allows the overall switch bandwidth to scale adequately with the number of switch units.

In contrast, a fabric switch can include an arbitrary number of switches with individual addresses, can be based on an arbitrary topology, and does not require extensive manual configuration. The switches can reside in the same location, or be distributed over different locations. These features overcome the inherent limitations of switch stacking and make it possible to build a large “switch farm” which can be treated as a single, logical switch. Due to the automatic configuration capabilities of a fabric switch, an individual physical switch can dynamically join or leave the fabric switch without disrupting services to the rest of the network.

Furthermore, the automatic and dynamic configurability of a fabric switch allows a network operator to build its switching system in a distributed and “pay-as-you-grow” fashion without sacrificing scalability. The fabric switch's ability to respond to changing network conditions makes it an ideal solution in a virtual computing environment, where network loads often change with time.

Although the present disclosure is presented using examples based on the TRILL protocol, embodiments of the present invention are not limited to networks defined using TRILL, or a particular Open System Interconnection Reference Model (OSI reference model) layer. For example, embodiments of the present invention can also be applied to a multi-protocol label switching (MPLS) network. In this disclosure, the term “fabric switch” is used in a generic sense, and can refer to a network operating in any networking layer, sub-layer, or a combination of networking layers.

In this disclosure, the term “end device” can refer to a physical or virtual device coupled to a fabric switch. An end device can be a host, a server, a conventional layer-2 switch, a layer-3 router, or any other type of device. Additionally, an end device can be coupled to other switches or hosts further away from a network. An end device can also be an aggregation point for a number of network devices to enter the network. The terms “device” and “machine” are used interchangeably.

The term “frame” refers to a group of bits that can be transported together across a network. “Frame” should not be interpreted as limiting embodiments of the present invention to layer-2 networks. “Frame” can be replaced by other terminologies referring to a group of bits, such as “packet,” “cell,” or “datagram.”

The term “RBridge” refers to routing bridges, which are bridges implementing the TRILL protocol as described in Internet Engineering Task Force (IETF) Request for Comments (RFC) “Routing Bridges (RBridges): Base Protocol Specification,” available at http://tools.ietf.org/html/rfc6325, which is incorporated by reference herein. Embodiments of the present invention are not limited to application among RBridges. Other types of switches, routers, and forwarders can also be used.

The term “edge port” refers to a port in a fabric switch which exchanges data frames with an end device outside of the fabric switch. The term “inter-switch port” refers to a port which couples a member switch of a fabric switch with another member switch and is used for exchanging data frames between the member switches.

The term “RBridge identifier” refers to a group of bits that can be used to identify an RBridge. Note that the TRILL standard uses “RBridge ID” to denote a 48-bit intermediate-system-to-intermediate-system (IS-IS) System ID assigned to an RBridge, and “RBridge nickname” to denote a 16-bit value that serves as an abbreviation for the “RBridge ID.” In this disclosure, “RBridge identifier” is used as a generic term and is not limited to any bit format, and can refer to “RBridge ID” or “RBridge nickname” or any other format that can identify an RBridge.

The term “dual-homed end device” refers to an end device that has an aggregate link to two or more switches belonging to one or more fabric switches, where the aggregate link includes multiple physical links to the different RBridges. The aggregate link, which includes multiple physical links, functions as one logical link to the end station. Although the term “dual” is used here, the term “dual-homed end device” does not limit the number of physical RBridges sharing the aggregate link to two. In various embodiments, other numbers of physical RBridges can share the same aggregate link. Where “dual-homed end device” is used in the present disclosure, the term “multi-homed end device” can also be used.

Network Architecture

FIG. 1A illustrates an exemplary virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention. As illustrated in FIG. 1A, a fabric switch 110 includes member switches 112, 114, 116, and 118, and a fabric switch 120 includes member switches 122, 124, and 126. In some embodiments, one or more switches in fabric switches 110 and 120 can be virtual switches (e.g., a software switch running on a computing device).

In some embodiments, fabric switches 110 and 120 are TRILL networks and a respective member switch of fabric switches 110 and 120, such as switches 112 and 122, is a TRILL RBridge. RBridges in fabric switches 110 and 120 use edge ports to communicate to end devices and inter-switch ports to communicate to other member switches. For example, RBridge 118 is coupled to end device 106 via an edge ports and to RBridges 112, 114, and 116 via inter-switch ports and one or more links. Communication via an edge port can be based on Ethernet and via an inter-switch port can be based on TRILL protocol. Note that the link between fabric switches 110 and 120 is established via edge ports of switches 114 and 124.

RBridges 112 and 118 are configured to operate in a special “trunked” mode for multi-homed end device 104 and form a virtual link aggregation 150. End device 104 view these partner RBridges 112 and 118 as a common virtual RBridge 152, with a corresponding virtual RBridge identifier. Dual-homed end device 104 is considered to be logically coupled to virtual RBridge 152 via logical links represented by dotted lines. Virtual RBridge 152 is considered to be logically coupled to partner RBridges 112 and 118, optionally with zero-cost links (also represented by dotted lines). Incoming frames from end device 104 is marked with virtual RBridge 152's identifier as their ingress RBridge identifier. As a result, other RBridges in fabric switch 110 can learn that end device 104 is reachable via virtual RBridge 152. Furthermore, RBridges 112 and 118 can advertise their respective connectivity (optionally via zero-cost links) to virtual RBridge 152. Hence, multi-pathing can be achieved when other RBridges choose to send frames to virtual RBridge 152 (which is marked as the egress RBridge in the frames) via partner RBridges 112 and 118.

Since partner RBridges 112 and 118 function as a single logical RBridge 152, the MAC address reachability learned by a partner RBridge is shared with the other partner RBridge. For example, during normal operation, end device 104 may choose to send its outgoing frames only via the link to RBridge 118. As a result, only RBridge 118 learns end device 104's MAC address (and the corresponding port on RBridge 118 to which end device 104 is coupled). This information is then shared by RBridge 118 with RBridge 112. Since the frames coming from end device have virtual RBridge 152's identifier as their ingress RBridge identifier, when RBridges in network fabric switch 110 send frames back to end device 104, these frames have virtual RBridge 152's identifier as their egress RBridge identifier, and these frames can be sent to either RBridge 112 or 118.

In virtual link aggregation 150, RBridges 112 and 118 can forward BUM traffic to each other. If the BUM traffic is originated from end device 104, the RBridge receiving the BUM traffic performs source suppression by precluding the BUM traffic from being forwarded to end device 104 (i.e., by precluding the RBridge from determining an output port associated with end device 104 for the BUM traffic). For example, if RBridge 112 receives a frame belonging to BUM traffic, RBridge 112 checks the source RBridge identifier of the frame. If the source RBridge identifier is associated with virtual RBridge 152, RBridge considers the frame to be from end device 104 and suppress source forwarding by not forwarding the frame to end device 104.

In addition, ingress RBridge 112, 118, or both can receive a frame belonging to BUM traffic via virtual link aggregation 150 (i.e., from end device 104). Upon receiving the frame, ingress RBridge 112 and/or 118 forward the received frame to a root RBridge of a multicast tree in fabric switch 110. In this example, RBridges 116 and 126 can be the root RBridges of fabric switches 110 and 120, respectively. Ingress RBridge 112 and/or 118 set the destination RBridge identifier to be the RBridge identifier of root RBridge 116 and send the frame toward RBridge 116. Root RBridge 116, in turn, distributes the frame to a respective RBridge in fabric switch 110 via the multicast tree.

Similar to end device 104, end device 102 is also multi-homed and coupled to RBridges 112 and 122 of fabric switches 110 and 120, respectively. Hence, RBridges 112 and 122 form a virtual link aggregation 130 across fabric switches 110 and 120. End device 102 view RBridges 112 and 122 as a common virtual RBridge 132, with a corresponding virtual RBridge identifier. Dual-homed end device 102 is considered to be logically coupled to virtual RBridge 132 via logical links represented by dotted lines. Virtual RBridge 132 is considered to be logically coupled to partner RBridges 112 and 122, optionally with zero-cost links (also represented by dotted lines). Details about virtual link aggregation and virtual RBridge assignment can be found in U.S. patent application Ser. No. 12/725,249, the disclosure of which is incorporated herein.

Other RBridges in fabric switches 110 and 120 view end device 102 to be coupled to virtual RBridge 132. Incoming frames from end device 102 is marked with virtual RBridge 152's identifier as their ingress RBridge identifier. When RBridges in network fabric switches 110 and 120 send frames back to end device 102, these frames have virtual RBridge 152's identifier as their egress RBridge identifier. Frames from RBridges in fabric switch 110 are received by RBridge 112 and frames from RBridges in fabric switch 120 are received by RBridge 122.

However, source suppression of virtual link aggregation 150 may not be applicable to virtual link aggregation 130 because virtual link aggregation 130 spans two fabric switches. Typically, a fabric switch has a local identifier range associated with the fabric switch. A respective RBridge, physical or virtual, of that fabric switch is assigned an RBridge identifier from that corresponding local identifier range. As a result, virtual RBridge 152 is assigned an identifier from the local identifier range associated with fabric switch 110. Furthermore, the virtual RBridge identifier assigned by fabric switch 110 does not go beyond the boundaries of fabric switch 110. If virtual RBridge 132 is assigned an identifier from the local identifier range of fabric switch 110, RBridge 122 would not recognize the identifier and would not perform source suppression because virtual RBridge identifier is unknown to RBridge 122. Similarly, if virtual RBridge 132 is assigned an identifier from the local identifier range of fabric switch 120, RBridge 112 would not perform source suppression based on the unknown virtual RBridge identifier.

In addition, forwarding of BUM traffic from virtual link aggregation 150 may not be applicable to virtual link aggregation 130 because root RBridge is typically fabric specific. For example, to forward a frame belonging to BUM traffic, if RBridge 116 is assigned as the root RBridge, the frame is going to be distributed in fabric switch 110. On the other hand, if RBridge 126 is assigned as the root RBridge, the frame is going to be distributed in fabric switch 120. Hence, for virtual link aggregation 130 that spans multiple fabric switches, a fabric-specific root RBridge is not suitable for forwarding BUM traffic.

In embodiments of the present invention, the aforementioned problems are solved by assigning an identifier to virtual RBridge 132 from a global identifier range, which is known, recognizable, and unique in both fabric switches 110 and 120. This allows partner RBridges 112 and 122, which are in different fabric switches, to recognize the virtual RBridge identifier and perform source suppression for BUM traffic originating from end device 102 associated with virtual link aggregation 130. Furthermore, a virtual root RBridge 136 is created, which is associated with root RBridges 116 and 126. Virtual root RBridge 136 is considered to be logically coupled to root RBridges 116 and 126, optionally with zero-cost links (represented by dotted lines). Virtual root RBridge 136 is also assigned an identifier from the global identifier range, which is known, recognizable, and unique in both fabric switches 110 and 120. Upon receiving a frame belonging to BUM traffic, partner RBridges 112 and 122 set the identifier of virtual root RBridge 136 as the egress RBridge identifier of the TRILL encapsulation of the frame.

Because virtual RBridge 136 is considered to be logically coupled to root RBridges 116 and 126, virtual root RBridge 136 is mapped to the respective root RBridge of fabric switches 110 and 120. For example, in fabric switch 110, upon receiving a frame belonging to BUM traffic via virtual link aggregation 130, RBridge 112 encapsulates the frame in a multicast TRILL header and sets the egress RBridge identifier of the TRILL header to be virtual root RBridge 136's identifier. RBridge 112 then forwards this TRILL-encapsulated frame (can be referred to as a TRILL packet) toward virtual root RBridge 136. In this disclosure, the terms “TRILL-encapsulated frame” and “TRILL packet” are used interchangeably. Because virtual root RBridge 136 is logically coupled to root RBridge 116, the TRILL packet then reaches root RBridge 116. Upon receiving the TRILL packet, root RBridge 116 forwards the frame to respective RBridge in fabric switch 110. Similarly, in fabric switch 120, RBridge 122 sends a TRILL-encapsulated frame belonging to BUM traffic toward virtual root RBridge 136. Root RBridge 126 receives the frame and forwards the frame to respective RBridge in fabric switch 120.

For a respective frame from end device 102, the corresponding ingress RBridge (i.e., RBridge 112 if the traffic enters fabric switch 110, or RBridge 122 if the traffic enters fabric switch 120) attaches a TRILL header to the frame and sets the identifier of virtual RBridge 132 as the ingress RBridge identifier in the TRILL header. If the frame belongs to BUM traffic, regardless of whether the frame is sent to RBridge 112 or 122, the corresponding ingress RBridge sets the identifier of virtual RBridge 132 as the ingress RBridge identifier and the identifier of virtual root RBridge 136 as the egress RBridge identifier in the TRILL header.

Because virtual RBridge 132's identifier is known to both fabric switches 110 and 120, partner RBridges 122 and 122 can facilitate source suppression for BUM traffic from end device 102. For example, when RBridge 112 or 122 receives from another RBridge a TRILL packet with a TRILL ingress RBridge identifier set to be virtual RBridge 132's identifier, RBridge 112 or 122 recognizes the packet to be from end device 102 and drops the packet, thereby preventing undesired looping. Furthermore, because virtual root RBridge 136's identifier is known to both fabric switches 110 and 120, when a TRILL-encapsulated frame belonging to BUM traffic reaches an RBridge in either fabric switch, the RBridge forwards the packet toward the physical root RBridge of the fabric switch in which the RBridge is a member.

In some embodiments, there are at least two links between fabric switches 110 and 120. FIG. 1B illustrates an exemplary virtual link aggregation across multiple fabric switches coupled via TRILL forwarding link(s), in accordance with an embodiment of the present invention. In this example, RBridge 114 of fabric switch 110 and RBridge 124 of fabric switch 120 have two links 142 and 144 between them. Link 142 is referred to as a TRILL primary link. Link 142 is used for forwarding TRILL-encapsulated frame belonging to BUM traffic while retaining the TRILL encapsulation. Typically, TRILL encapsulation is specific to a fabric switch, and when a frame leaves a fabric switch, the TRILL encapsulation is removed. However, a TRILL-encapsulated frame belonging to BUM traffic forwarded via TRILL primary link 142 (i.e., via a TRILL-supported edge port of RBridge 114 coupled to link 142) retains the TRILL encapsulation. This allows the TRILL header to retain virtual RBridge 132's identifier as the ingress RBridge identifier and virtual root RBridge 136's identifier as the egress RBridge identifier.

During operation, end device 102 generates a frame belonging to BUM traffic and sends the frame to RBridge 112. Upon receiving the frame, RBridge 112 encapsulates the frame with a TRILL header with virtual RBridge 132's identifier as the ingress RBridge identifier and virtual root RBridge 136's identifier as the egress RBridge identifier. Subsequently, RBridge 112 forwards the TRILL packet toward virtual root RBridge 136. Because root RBridge 116 is logically coupled to virtual root RBridge 136, root RBridge 116 receives the TRILL packet and distributes this TRILL packet along its multicast tree to respective RBridge of fabric switch 110.

When this TRILL packet reaches RBridge 114, RBridge 114 forwards this TRILL packet via link 124 with the same ingress and egress RBridge identifiers. When RBridge 124 receives this TRILL packet, RBridge 124, in turn, forwards the packet toward virtual root RBridge 136. Because root RBridge 126 of fabric switch 120 is logically coupled to virtual root RBridge 136, root RBridge 126 receives the TRILL packet and distributes this TRILL packet along its multicast tree to respective RBridge of fabric switch 120. When this packet reaches RBridge 122, RBridge 122 prevents the packet from being forwarded to end device 102 because the packet's ingress RBridge identifier is virtual RBridge 132's identifier.

If end device 102 generates an Ethernet frame destined for an end device 108, which is coupled to fabric switch 120 via RBridge 126, and sends this frame to RBridge 112, RBridge 112 generates a TRILL header with an egress RBridge identifier set to be RBridge 114's identifier. Consequently, when the TRILL packet reaches RBridge 114, RBridge 114 decapsulates the packet's TRILL header and forwards the Ethernet frame via link 144. This link 144 can be referred to as an Ethernet primary link. When the Ethernet frame reaches RBridge 124 via Ethernet primary link 144, RBridge 124 generates a new TRILL header with RBridge 126's identifier as the egress RBridge identifier of the new TRILL header.

In some embodiments, RBridge 114 determines whether to forward via Ethernet primary link 144 based on the identifier range of the ingress RBridge identifier of a TRILL packet. For example, if the TRILL packet includes an Ethernet frame from end device 106, the ingress RBridge identifier of the TRILL packet is RBridge 118's identifier. Similarly, if the TRILL packet includes an Ethernet frame from end device 104, the ingress RBridge identifier of the TRILL packet is virtual RBridge 152's identifier. Because RBridge 118's identifier and virtual RBridge 152's identifier are from the local identifier range associated with fabric switch 110, RBridge 114 removes the TRILL encapsulation of the TRILL packet and forwards the Ethernet frame via link 144. However, if the identifier range of the ingress RBridge identifier of a TRILL packet corresponds to a global identifier range, RBridge 114 checks whether the packet belongs to BUM traffic, and forwards accordingly.

Frame Forwarding

In the example in FIG. 1B, partner RBridges 112 and 122 are responsible for forwarding frames received via local edge ports from end device 102. On the other hand, a respective RBridge in fabric switches 110 and 120 can forward a TRILL-encapsulated frame destined for end device 102. FIG. 2A presents a flowchart illustrating the process of forwarding a frame received via an edge port at a partner RBridge which participates in a virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention. Upon receiving a frame via a local edge port (operation 202), the partner RBridge identifies the destination media access control (MAC) address of the received frame (operation 204). The partner RBridge then identifies the egress RBridge identifier corresponding to the destination MAC address (operation 206). For example, if the destination MAC address corresponds to an end device coupled to another fabric switch, the egress RBridge identifier is the identifier of an RBridge coupled to the other fabric switch, as described in conjunction with FIG. 1B.

The partner RBridge encapsulates the frame in a TRILL header and sets the identified RBridge identifier as the egress RBridge identifier (operation 208) and sets a virtual RBridge identifier associated with the virtual link aggregation as the ingress RBridge identifier of the TRILL header (operation 210). The partner RBridge determines the next-hop RBridge based the egress RBridge identifier (operation 212), and sets the outer destination MAC address (MAC DA) corresponding to the next-hop RBridge and outer source MAC address (MAC SA) corresponding to the local RBridge (operation 214). The partner RBridge then determines the output port corresponding to the outer destination MAC address (operation 216) and transmits the TRILL-encapsulated frame via the determined output port (operation 218).

FIG. 2B presents a flowchart illustrating the process of an RBridge forwarding a TRILL-encapsulated unicast frame, in accordance with an embodiment of the present invention. Upon receiving a TRILL-encapsulated frame (operation 252), the RBridge checks whether the egress RBridge identifier is local (i.e., whether the egress RBridge identifier is associated with the local RBridge) (operation 254). In the example of FIG. 1A, virtual RBridge 132's identifier and RBridge 112's identifier both are associated with RBridge 112. Hence, if the egress RBridge identifier of the TRILL header corresponds to a virtual or physical RBridge identifier associated with the local RBridge, the RBridge determines that the egress RBridge identifier is local.

If the egress RBridge identifier is not local, the RBridge forwards the frame to the next-hop RBridge based on the egress RBridge identifier (operation 262). If the egress RBridge identifier is local, the RBridge removes the TRILL encapsulation (operation 256), determines the output port corresponding to the frame's inner destination MAC address (operation 258), and transmits the frame via the determined output port (operation 260).

In the example in FIG. 1B, partner RBridges 112 and 122 are responsible for forwarding frames belonging to BUM traffic received via local edge ports from end device 102. On the other hand, a respective RBridge in fabric switches 110 and 120 can forward a TRILL-encapsulated frame belonging to BUM traffic. FIG. 3A presents a flowchart illustrating the process of forwarding a frame belonging to BUM traffic received via an edge port at a partner RBridge which participates in a virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention. Upon receiving a frame belonging to BUM traffic via a local edge port (operation 302), the partner RBridge forwards the frame to other local edge ports associated with the BUM traffic (operation 304). For example, if the frame is a multicast frame of a multicast group, the partner RBridge forwards the frame to the local end devices, which are members of the multicast group.

The partner RBridge then identifies the virtual root RBridge identifier (operation 306), and encapsulates the frame in a TRILL header and sets the virtual root RBridge identifier as the egress RBridge identifier of the TRILL header (operation 308). The partner RBridge sets a virtual RBridge identifier associated with the virtual link aggregation as the ingress RBridge identifier of the TRILL header (operation 310). The partner RBridge determines the next-hop RBridge based the virtual root RBridge identifier (operation 312), and sets the outer destination MAC address corresponding to the next-hop RBridge and outer source MAC address corresponding to the local RBridge (operation 314). The partner RBridge then determines the output port corresponding to the outer destination MAC address (operation 316) and transmits the TRILL-encapsulated frame via the determined output port (operation 318).

FIG. 3B presents a flowchart illustrating the process of an RBridge forwarding a TRILL-encapsulated frame destined to a virtual root RBridge, in accordance with an embodiment of the present invention. Note that a frame destined to virtual root RBridge belongs to BUM traffic and the egress RBridge identifier of the TRILL header of the frame is the identifier of the virtual root RBridge. Upon receiving a TRILL-encapsulated frame destined to virtual root RBridge (operation 352), the RBridge checks whether the virtual root RBridge identifier is local (i.e., the virtual root RBridge identifier is associated with the local RBridge) (operation 354). In the example of FIG. 1A, virtual root RBridge 136's identifier is local to both RBridges 116 and 118.

If the virtual root RBridge identifier is not local, the RBridge forwards the frame to the next-hop RBridge based on the virtual root RBridge identifier (operation 358). Otherwise, the RBridge is the root RBridge of the corresponding fabric switch. Hence, the RBridge forwards the frame to respective other RBridges of the fabric switch (operation 356).

FIG. 3C presents a flowchart illustrating the process of an RBridge forwarding a TRILL-encapsulated frame belonging to BUM traffic, in accordance with an embodiment of the present invention. Upon receiving a TRILL-encapsulated frame belonging to BUM traffic from virtual root RBridge (operation 362), the RBridge checks whether the egress RBridge identifier is local (i.e., the egress RBridge identifier is associated with the local RBridge) (operation 364). If the egress RBridge identifier is not local, the RBridge forwards the frame to the next-hop RBridge based on the egress RBridge identifier (operation 374).

If the egress RBridge identifier is local, the RBridge checks whether the local RBridge has a TRILL primary link (operation 366). In the example of FIG. 1B, RBridges 114 and 124 have a TRILL primary link. If the RBridge has a TRILL primary link, the RBridge forwards the TRILL-encapsulated frame via the TRILL primary link without decapsulating the TRILL header (operation 368), as described in conjunction with FIG. 1B. If the RBridge does not have a TRILL primary link (operation 366) or has forwarded the TRILL-encapsulated frame via the TRILL primary link (operation 368), the RBridge removes the TRILL encapsulation (operation 370) and forwards the frame to local edge ports associated with BUM traffic (operation 372). For example, if the frame is a multicast frame of a multicast group, the RBridge forwards the frame to the local end devices, which are members of the multicast group.

Failure Handling

FIG. 4A illustrates exemplary failure scenarios associated with a virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention. Suppose that failure 412 disrupts the availability of RBridge 122 to end device 102 (e.g., reachability between RBridge 122 and end device 102). Examples of failure 412 include, but are not limited to, a failure of link 402 between RBridge 112 and end device 102, an interface card failure in RBridge 112 and/or end device 102 associated with link 402, node failure of RBridge 112, unavailability of fabric switch 110. Examples of a cause for unavailability of fabric switch 110 include, but are not limited to, software update, maintenance, and wiring updates to inter-switch links. Consequently, RBridge 112 can no longer couple end device 102 via virtual link aggregation 130.

As a result, RBridge 122 starts receiving frames from end device 102. Upon detecting failure 412, instead of virtual RBridge 132's identifier, RBridge 122 starts setting RBridge 122's identifier as the ingress RBridge identifier in the TRILL encapsulations of the frames from end device 102. In other words, since end device 102 no longer has virtual link aggregation 130 to both RBridges 112 and 122, virtual RBridge 132 no longer exists for end device 102. After the TRILL-encapsulated frames from end device 102 reach other egress RBridges in fabric switch 120, these RBridges learn that the MAC address corresponding to end device 102 is associated with RBridge 122, instead of virtual RBridge 132. Subsequent frames destined to end device 102 are sent to RBridge 122.

Note that, during the topology convergence process, other RBridges can continue to send frames to virtual RBridge 132. If RBridge 112 is available (e.g., has not suffered a node failure), RBridge 112 may continue to receive frames destined to end device 102. RBridge 112 can flood these frames to all the ports (except the ports from which the frames are received), or optionally forward these frames to RBridge 122 so there is minimal data loss.

Suppose that failure 414 disrupts the availability of root RBridge 126. Examples of failure 414 include, but are not limited to, one or more link failures disconnecting RBridge 126 from fabric switch 120, one or more interface card failures in RBridge 126 disconnecting RBridge 126 from fabric switch 120, node failure of RBridge 126, unavailability of fabric switch 120. Examples of a cause for unavailability of fabric switch 120 include, but are not limited to, software update, maintenance, and wiring updates to inter-switch links. Consequently, RBridge 126 can no longer serve as the root RBridge for fabric switch 120. In some embodiments, upon detecting the unavailability of RBridge 126, the active RBridges of fabric switch 120 elect another root RBridge among the active RBridges and associate virtual root RBridge 136 with the newly elected root RBridge of fabric switch 120. This newly elected root RBridge can be logically coupled to virtual root RBridge 136, optionally with zero-cost links.

FIG. 4B presents a flowchart illustrating the process of handling a failure that affects a partner RBridge which participates in a virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention. During operation, an RBridge detects unavailability of its partner RBridge (operation 452). The RBridge then disassociates the RBridge and the end device with the corresponding virtual RBridge (operation 454), and returns to the normal forwarding and/or flooding operation as for non-trunked ports. Furthermore, the RBridge configures local RBridge to set its own RBridge identifier as the ingress RBridge identifier in the TRILL header of the ingress frames from the end device (operation 456). Optionally, the RBridge can broadcast the MAC reachability of the end device via its own RBridge identifier to other RBridges in the corresponding fabric which (operation 458).

FIG. 4C presents a flowchart illustrating the process of handling a failure that affects a root RBridge which is associated with a virtual root RBridge, in accordance with an embodiment of the present invention. Upon detecting the unavailability of a root RBridge (operation 462), the active RBridges of the corresponding fabric switch elect another root RBridge among the active RBridges (operation 464) and associate the virtual root RBridge with the newly elected root RBridge (operation 466). This newly elected root RBridge can be logically coupled to the virtual root RBridge, optionally with zero-cost links.

Hybrid Virtual Link Aggregation

FIG. 5A illustrates an exemplary hybrid virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention. As illustrated in FIG. 5A, a fabric switch 510 includes member switches 512, 514 and 516, and a fabric switch 520 includes member switches 522, 524, and 526. One or more switches in fabric switches 510 and 520 can be virtual switches (e.g., a software switch running on a computing device). In some embodiments, fabric switches 510 and 520 are TRILL networks and a respective member switch of fabric switches 510 and 520, such as switch 512 and 522, are TRILL RBridges. End devices 502 and 504 are coupled to RBridges 516 and 526, respectively. RBridges in fabric switches 510 and 520 use edge ports to communicate to end devices and inter-switch ports to communicate to other member switches. Communication via an edge port can be based on Ethernet and via an inter-switch port can be based on TRILL protocol.

RBridges 512 and 522 are coupled to each other via a plurality of links. These links can be configured to operate as a trunk 532. Similarly, RBridges 514 and 524 are coupled to each other via a plurality of links. These links can be configured to operate as a trunk 534. In some embodiments, trunks 532 and 534 can be configured to operate in a special “trunked” mode and form a hybrid virtual link aggregation 536. Even though RBridge 512 is not coupled to RBridge 524, and RBridge 514 is not coupled to RBridge 522, trunks 532 and 534 can operate as one hybrid virtual link aggregation.

RBridges 516 considers hybrid virtual link aggregation 536 as one single link associated with two RBridges 512 and 514. Similarly, RBridges 526 considers hybrid virtual link aggregation 536 as one single link associated with two RBridges 522 and 524. As a result, to forward a frame toward RBridge 526, RBridge 516 first determines an egress RBridge between RBridges 512 and 514. Suppose that RBridge 516 determines RBridge 512 as the egress RBridge, encapsulates the frame in a TRILL header, and sends the TRILL-encapsulated frame to RBridge 512. Upon receiving the frame, RBridge 512 removes the TRILL header, selects a physical link in trunk 532, and forwards the frame via the selected link.

FIG. 5B illustrates an exemplary packet forwarding via a hybrid virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention. During operation, end device 502 is in communication with end device 504 and forwarding a plurality of frames toward end device 504. Upon receiving a frame from end device 502, ingress RBridge 516 first determines that end device 504 is coupled to fabric switch 510 via hybrid virtual link aggregation 536, which is associated with RBridges 512 and 514. In some embodiments, a respective RBridge in fabric switch 510 maintains a mapping between hybrid virtual link aggregation 536, and associated RBridges 512 and 514.

RBridge 516 then uses an arbitration mechanism 552 to determine an egress RBridge from RBridges 512 and 514 for a respective frame. In some embodiments, arbitration mechanism 552 can be a hash function. This hash function can select the egress RBridge such a way that frames from 516 are distributed among RBridges 512 and 514 based on a policy. Examples of a policy include, but are not limited to, load balancing, quality of service, security, network availability, and computing resources. Suppose that arbitration mechanism 552 selects RBridge 514 as the egress RBridge. RBridge 516 then encapsulates the frame in a TRILL header and forwards the frame toward RBridge 514 based the TRILL encapsulation. RBridge 514 receives the frame and removes the TRILL encapsulation. RBridge 514 then determines the physical link in trunk 534 via which the frame should be forwarded.

In some embodiments, RBridge 514 can forward frames across the multiple links in trunk 534, thereby achieving a desired distribution among the links in trunk 534. Such transmission allows RBridge 514 to perform spray operation 554 on frames by selecting respective output ports for the frames such a way that the frames are distributed to different links of trunk 534. Upon receiving a frame, RBridge 524 identifies that end device 504 is coupled to RBridge 526, encapsulates the frame in a TRILL header, sets RBridge 526's identifier as the egress RBridge identifier in the TRILL header, and forwards the TRILL packet toward egress RBridge 526.

Exemplary RBridge

FIG. 6 illustrates an exemplary RBridge which supports virtual link aggregation across multiple fabric switches, in accordance with an embodiment of the present invention. In this example, an RBridge 600 includes a number of communication ports 602, a packet processor 610, a link aggregation module 630, a suppression module 632, a multicast module 634, and a storage 650. In some embodiments, packet processor 610 adds a TRILL header to a packet. RBridge 600 can also include a fabric switch management module 620, which maintains a membership in a fabric switch. Switch 600 maintains a configuration database in storage 650 that maintains the configuration state of a respective switch within the fabric switch. Switch 600 maintains the state of the fabric switch, which is used to join other switches. Under such a scenario, communication ports 602 can include inter-switch communication channels for communication within a fabric switch. This inter-switch communication channel can be implemented via a regular communication port and based on any open or proprietary format (e.g., TRILL protocol).

During operation, link aggregation module 630 marks an ingress RBridge identifier field of a frame received via one of the communication ports 602 with a virtual RBridge identifier. Upon receiving a frame, packet processor 610 identifies the virtual RBridge identifier as the egress RBridge identifier of a frame and, in response, determines that RBridge 600 is the egress RBridge for the frame. If the frame belongs to BUM traffic, suppression module 632 precludes RBridge 600 from determining an output port associated with a local end device for the frame, as described in conjunction with FIG. 1A. If RBridge 600 receives a frame belonging to BUM traffic via an edge port, multicast module 634 identifies the frame and marks an egress-switch field of the frame with a virtual root RBridge identifier.

In some embodiments, RBridge 600 also includes a failure detection module 642. Failure detection module 642 can detect unavailability of the physical root RBridge of the fabric switch and, in response, identifies a new multicast root RBridge and associates the virtual root RBridge identifier with the new multicast root RBridge, as described in conjunction with FIG. 4C. Failure detection module 642 can also detect unavailability of a partner RBridge and, in response, disassociates the RBridge from the virtual switch identifier. For any frame received after detecting unavailability of the partner RBridge, failure detection module 642 marks an egress RBridge field of the frame with an RBridge identifier of RBridge 600, as described in conjunction with FIG. 4A.

In some embodiments, RBridge 600 also includes a TRILL-supported edge port in communication ports 602 and a TRILL primary module 644. The TRILL-supported edge port is determined as an output port for frames destined outside of the fabric switch. During operation, TRILL primary module 644 determines a TRILL-encapsulated frame to be belonging to BUM traffic. In response, TRILL primary module 644 precludes RBridge 600 from removing TRILL encapsulation from the TRILL-encapsulated frame and determines the TRILL-supported edge port as the output port for the TRILL-encapsulated frame, as described in conjunction with FIG. 3C.

In some embodiments, link aggregation module 630 operates a trunked link, which is coupled to RBridge 600 via a number of ports in communication ports 602, in conjunction with a second trunked link of a second RBridge of the fabric switch as a virtual link aggregation. The virtual link aggregation is mapped to RBridge 600 and the second RBridge. A second fabric switch is reachable from the fabric switch via the trunked links coupled to RBridge 600 and the second trunked links. In some embodiments, this virtual link aggregation is mapped to RBridge 600 based on a hash function. RBridge 600 can select respective output ports for the frames destined to the second fabric switch such that the frames are distributed across links of the trunked link, thereby spraying the frames across the links of the trunked link, as described in conjunction with FIG. 5B.

Note that the above-mentioned modules can be implemented in hardware as well as in software. In one embodiment, these modules can be embodied in computer-executable instructions stored in a memory which is coupled to one or more processors in RBridge 600. When executed, these instructions cause the processor(s) to perform the aforementioned functions.

In summary, embodiments of the present invention provide a switch and a method for providing virtual link aggregation across multiple fabric switches. In one embodiment, the switch is configurable to be a member of a first fabric switch. The switch includes a link aggregation module. During operation, the link aggregation module marks an ingress-switch field of a frame with a virtual switch identifier. This virtual switch identifier is associated with the switch and a second switch, which is a member of a second fabric switch, and is from a range of identifier associated with the first fabric switch and the second fabric switch. Each of the first fabric switch and the second fabric switch is operable to accommodate a plurality of switches and operate as a single switch.

The methods and processes described herein can be embodied as code and/or data, which can be stored in a computer-readable non-transitory storage medium. When a computer system reads and executes the code and/or data stored on the computer-readable non-transitory storage medium, the computer system performs the methods and processes embodied as data structures and code and stored within the medium.

The methods and processes described herein can be executed by and/or included in hardware modules or apparatus. These modules or apparatus may include, but are not limited to, an application-specific integrated circuit (ASIC) chip, a field-programmable gate array (FPGA), a dedicated or shared processor that executes a particular software module or a piece of code at a particular time, and/or other programmable-logic devices now known or later developed. When the hardware modules or apparatus are activated, they perform the methods and processes included within them.

The foregoing descriptions of embodiments of the present invention have been presented only for purposes of illustration and description. They are not intended to be exhaustive or to limit this disclosure. Accordingly, many modifications and variations will be apparent to practitioners skilled in the art. The scope of the present invention is defined by the appended claims.

Claims

1. A switch, comprising: fabric circuitry configured to maintain a membership in a first network of interconnected switches, wherein the first network of interconnected switches is identified by a first fabric identifier; andaggregation circuitry configured to: identify a virtual link aggregation group comprising a second switch and a third switch in the first network of interconnected switches, wherein a second network of interconnected switches is reachable via the virtual link aggregation group, and wherein the second network of interconnected switches is identified by a second fabric identifier; andmaintain a mapping between the second and third switches and the virtual link aggregation group; andforwarding circuitry configured to select, for a packet destined to the second network of interconnected switches, an egress switch between the second and third switches based on the mapping.

2. The switch of claim 1, wherein the virtual link aggregation group further comprises a first trunked link of the second switch and a second trunked link of the third switch.

3. The switch of claim 1, wherein the egress switch is selected based on an arbitration policy of the first network of interconnected switches, wherein the arbitration policy is based on one or more of: load balancing, quality of service, security, network availability, and computing resources.

4. The switch of claim 2, wherein the arbitration policy is based on a hash function.

5. The switch of claim 1, wherein the forwarding circuitry is further configured to select a port via which the egress switch is reachable for the packet destined to the second network of interconnected switches.

6. The switch of claim 1, wherein the forwarding circuitry is configured to select the egress switch such that packets are distributed between the second and third switches.

7. The switch of claim 1, wherein the forwarding circuitry is further configured to obtain an inner frame by decapsulating an encapsulation header of an encapsulated packet.

8. A method, comprising: maintaining, for a switch, a membership in a first network of interconnected switches, wherein the first network of interconnected switches is identified by a first fabric identifier;identifying a virtual link aggregation group comprising a second switch and a third switch in the first network of interconnected switches, wherein a second network of interconnected switches is reachable via the virtual link aggregation group, and wherein the second network of interconnected switches is identified by a second fabric identifier;maintaining a mapping between the second and third switches and the virtual link aggregation group; andselecting, for a packet destined to the second network of interconnected switches, an egress switch between the second and third switches based on the mapping.

9. The method of claim 8, wherein the virtual link aggregation group further comprises a first trunked link of the second switch and a second trunked link of the third switch.

10. The method of claim 8, wherein the egress switch is selected based on an arbitration policy of the first network of interconnected switches, wherein the arbitration policy is based on one or more of: load balancing, quality of service, security, network availability, and computing resources.

11. The method of claim 10, wherein the arbitration policy is based on a hash function.

12. The method of claim 8, further comprising selecting a port via which the egress switch is reachable for the packet destined to the second network of interconnected switches.

13. The method of claim 8, further comprising selecting the egress switch such that packets are distributed between the second and third switches.

14. The method of claim 8, further comprising obtaining an inner frame by decapsulating an encapsulation header of an encapsulated packet.

15. A computer system; comprising: a processor;a storage device coupled to the processor and storing instructions that when executed by the processor cause the processor to perform a method, the method comprising:maintaining a membership in a first network of interconnected switches, wherein the first network of interconnected switches is identified by a first fabric identifier;identifying a virtual link aggregation group comprising a second computer system and a third computer system in the first network of interconnected switches, wherein a second network of interconnected switches is reachable via the virtual link aggregation group, and wherein the second network of interconnected switches is identified by a second fabric identifier;maintaining a mapping between the second and third computer systems and the virtual link aggregation group; andselecting, for a packet destined to the second network of interconnected switches, an egress system between the second and third computer systems based on the mapping.

16. The computer system of claim 15, wherein the virtual link aggregation group further comprises a first trunked link of the second computer system and a second trunked link of the third computer system.

17. The computer system of claim 15, wherein the egress system is selected based on an arbitration policy of the first network of interconnected switches, and wherein the arbitration policy is based on a hash function.

18. The computer system of claim 15, wherein the method further comprises selecting a port via which the egress system is reachable for the packet destined to the second network of interconnected switches.

19. The computer system of claim 15, wherein the method further comprises selecting the egress switch such that packets are distributed between the second and third computer systems.

20. The computer system of claim 18, wherein the method further comprises obtaining an inner frame by decapsulating an encapsulation header of an encapsulated packet.