CN1791064A

CN1791064A - Stack manager protocol with automatic set up mechanism

Info

Publication number: CN1791064A
Application number: CNA2005101170301A
Authority: CN
Inventors: 伊格内修斯·桑托索; 罗伯特·亨里克·雅各布·达·席尔瓦; 蒂莫西·希克斯; 文森特·马格雷特
Original assignee: Alcatel NV
Current assignee: Alcatel CIT SA; Alcatel Lucent NV
Priority date: 2004-10-28
Filing date: 2005-10-28
Publication date: 2006-06-21
Anticipated expiration: 2025-10-28
Also published as: CN101499976B; CN100481814C; CN1770739A; CN100481815C; CN100477638C; CN1770738A; CN101499976A

Abstract

A stack switch for performing pass-through forwarding operations, assigning stack switch identifiers to a plurality of stack switches, temporarily disabling supervision messages exchanged by stack switches, and electing a primary master stack switch is disclosed. In the pass-through mode, the stack switch automatically forwards packets between stack ports independent of their destination address to maintain connectivity between adjacent stack switches while conventional bridging and routing operations are disabled. The sequential identifiers are also preferably assigned in accordance with the sequential position of the various stack switches to better reflect the topology of those switches. The switch is also preferably adapted to temporarily disable the transmission of supervision messages between stack switches to prevent re-assignment of network management functions when switch resources are low. The switch may execute a stack switch identifier numbering scheme based on the time the switch became operational.

Description

Possess create a mechanism automatically pile up the manager agreement

The cross reference of related application

The U.S. Provisional Patent Application series number No.60/623 that the application requires to be attained at is that on October 28th, 2004 submitted to, be entitled as " STACKMANAGER PROTOCOL WITH AUTOMATIC SET UP MECHANISM (possess create a mechanism automatically pile up manager agreement) ", 370 rights and interests are included it in here by reference.

Technical field

Present invention relates in general to the management of the stacked switchboard system in a data communication network.More particularly, the present invention relates to a kind of fault-tolerant stacked switchboard system, it is suitable for detection, processing switch fault and recovers from exchange fault.

Background technology

In the stacked switchboard system in a packet switching network, respective switch functionally interconnects via a pair of appointment stack port that exists on each switch.Stacked switchboard system generally connects successively, and the feature of system topological generally is a closed-loop path, is called ring, or one opened the switch string, is called chain here.Each stacked switch is suitable for exchanging by utilizing the stacked switch identifier to transmit grouping via stack port between the FPDP of its FPDP and other stacked switch, and wherein the use of stacked switch identifier helps these suitable stacked switch that divides into groups effectively to transmit and exchange to.At present, can for example distribute switch identifier to stacked switch by user's manual configuration.But still having simultaneously needs stacked switch to be suitable for automatically with certainty ground and in turn, that is to say that according to the order of switch in ring or chain, distribution marker is given the switch that participates in.

At present, stacked switchboard system also is easy to occur identifier and assigns situation about repeating, and wherein when two stacked switch of combination from different system, above-mentioned situation will occur when perhaps user has been assigned same identifier mistakenly for twice.In current system, stacked switch automatically shuts down, and the connectedness that this can cut off system has unfortunately formed the stacked switch isolated island.Have a kind of needs, require stacked switch to be suitable for keeping two communication links between neighboring switch, and the switch with duplicate-identifier can not be transmitted grouping in stacked switchboard system.In this case, also need which exchange opportunity success of (1) definition election, and can will enliven but idle switch remains on the standard of certain state, this switch can also the receiving management order under aforesaid state; And (2) manage conflict and do not need artificial intervention.

At present, stacked switchboard system also adopts at the first main stacked switch and is carried out message between the stacked switch by pipe and determine separately operating state.Yet in some cases, first main frame may not have the computational resource status checkout that makes an immediate response, and infers the managed device possible errors that in this case first main frame breaks down, and begins to elect first a new main frame.As a result, stacked switchboard system may have two first main frames, thereby causes whole system to break down.Therefore, need stacked switch to be suitable for handling intermittent interruption, and the not change of start-up system management too early.

Summary of the invention

Preferred implementation of the present invention is characterised in that a kind of switching equipment, is used for carrying out in a stacked switchboard system that comprises a plurality of stacked switch penetrating forwarding operation.This switching equipment preferably includes: a plurality of ports and a routing engine are used for transmitting grouping between port.These a plurality of ports comprise a plurality of FPDP and a plurality of stack port, and just first stack port and second stack port are used for switch is connected to other stacked switch.This routing engine is suitable for realizing alleged penetrating pattern here.If penetrating pattern is under an embargo, routing engine for example according to the destination address information of packet associated, between a plurality of ports, transmit grouping.Yet, if penetrating pattern by gating, the packet forward that routing engine automatically will receive on first stack port is given second stack port, and the packet forward that automatically will receive on second stack port is given first stack port.The grouping that receives on FPDP generally is dropped.Like this, switch can be kept the minimum connective of stacked switchboard system, and does not need fully to operate or fully be integrated in the stacked switchboard system.For example when two stacked switch have identical identifier, can adopt penetrating pattern.

Preferred implementation of the present invention also is characterised in that a kind of switching equipment, is used for automatically distributing the stacked switch identifier.In a plurality of stacked switch each--all of responsibility that comprise first a main stacked switch that be responsible for to distribute and be suitable for bearing first main frame are managed stacked switch--comprises that two stack ports and one pile up manager, the latter is suitable for discerning a stacked switch that needs work stacked switch identifier, and automatically distributes an identifier in the work stacked switch set of identifiers to each of described one or more stacked switch.This work stacked switch set of identifiers preferably includes orderly integer, although person of skill in the art will appreciate that, because some identifiers were distributed to one or more stacked switch in the past, so may occur in this sequence at interval.Distribute orderly integer according to described one or more stacked switch in the orderly position of a sequence subsequently.Suppose that all stacked switch all do not have the identifier that existed in the past, the switch that obtains so may be " 1 " since first main frame for example with the ascending order number consecutively.If stacked switch has a ring topology, suppose that first main frame has distributed identifier, from this set of identifiers, the minimum available identifiers of distribution is given the adjacent stacks switch on its first stack port, and the maximum available identifiers of distribution is given the adjacent stacks switch on its second stack port.If in a chain topology, minimum available identifiers generally can be distributed to first main frame, the perhaps switch that can arrive by a stack port of first main frame, and available identifiers is distributed to respective switch when needed successively, just as stacked switch connects ringwise.The identifier allocation that obtains has like this reflected the orderly position of stacked switch better, and is on sensation, more more user friendly than the allocative decision of the prior art of present employing.

Preferred implementation of the present invention also is characterised in that a kind of switching equipment, be used for forbidding automatically and provisionally by the stacked switch exchange with state that monitors pile system and the supervision messages that detects change in topology.This system generally comprises one or more by pipe stacked switch and one the first main stacked switch that is used to manage these equipment.Be suitable for sending supervision messages one or more in a plurality of stacked switch by the pipe stacked switch, comprise first main frame.First main frame then under possible situation, is suitable for supervision messages is made a response, and it is a plurality of by the one or more transmission supervision messages in the pipe stacked switch to stop provisionally perhaps to send a supervision suspending message.First main frame also can be removed the supervision suspending message and recover standard operation.In some embodiments, if first main frame can't turn back to normal operating conditions, managed device also can automatically recover the regular transmission of supervision messages after a predetermined amount of time.

Preferred implementation of the present invention also is characterised in that a kind of being used at system initialization, and perhaps first main frame of appointment becomes when not working, the method for one first main stacked switch of election from a plurality of stacked switch.This method may further comprise the steps: be an election of each generation key in a plurality of stacked switch, the election key of each stacked switch of exchange between a plurality of stacked switch; The stacked switch that will have minimum election key is appointed as interim first stacked switch.The election key of switch comprises the start-up time of related stack switch--just for example date, hour, minute and second--, and switch arrives the required time of predetermined work state after starting.In a preferred embodiment, start-up time and switch are suitable for comprising that the network topology that has other stacked switch begins to carry out the time correlation connection of entry evaluation.

Description of drawings

The present invention illustrates by example, but is not limited to accompanying drawing, in the accompanying drawings:

Fig. 1 is a kind of functional block diagram of the integrated stacked switchboard system (ISS) according to the preferred embodiment for the present invention;

Fig. 2 is a kind of functional block diagram of the stacked switch that adopts in the ISS system according to the preferred embodiment for the present invention;

Fig. 3 is a state diagram according to each stage of stacked switch in the representative starting process of the preferred embodiment for the present invention;

But Fig. 4 is a flow chart of determining state according to the stackability corresponding to Fig. 3 of the preferred embodiment for the present invention;

Fig. 5 is a flow chart corresponding to the Discovery Status of Fig. 3 according to the preferred embodiment for the present invention;

Fig. 6 A and 6B are the flow chart corresponding to the election state of Fig. 3 according to the preferred embodiment for the present invention together;

Fig. 7 A and 7B are the flow chart corresponding to the unit number code assignment state of Fig. 3 according to the preferred embodiment for the present invention together;

Fig. 8 A and 8B are the ISS example system that corresponds respectively to number assignment front and back, unit according to the preferred embodiment for the present invention;

Fig. 9 A and 9B are the ISS example system that corresponds respectively to number assignment front and back, unit according to the preferred embodiment for the present invention;

Figure 10 A and 10B are the ISS example system that corresponds respectively to number assignment front and back, unit according to the preferred embodiment for the present invention;

Figure 11 is the representational election message of piling up according to the preferred embodiment for the present invention; And

Figure 12 is a representational supervision messages of piling up according to the preferred embodiment for the present invention.

Embodiment

Figure 1 illustrates the functional block diagram of a kind of integrated stacked switch (ISS) system in a data communication network.Network 120 comprises a plurality of stacked switch 100-103, and they functionally connect successively, and by means of piling up link 110-113, for example twisted-pair feeder or fiber optic cables for example form a chain or ring topology.Switching equipment 100-103 is stackable switch preferably, can interconnect by the special-purpose port that one or more those skilled in the art are referred to as stack port.A plurality of stacked switch 100-103, here be also referred to as stackable unit and unit, be suitable for other switches of ISS 120 and by for example one or more Local Area Network, wide area network (WAN), metropolitan area network (MAN) or Internet connection to the data that transmit packetizing between one or more end points station of ISS or other addressable entity.

In a preferred embodiment, stacked switch 100-103 is suitable in the data cell (PDU) of being located to carry on an agreement by the second layer (data link layer) of OSI(Open Systems Interconnection) reference model definition and the 3rd layer (network layer), preferably frame and grouping, exchange and the multilayer switch of routing operations, although they also can finish 4-7 layer swap operation.Each of stacked switch 100-103 generally can be served as independently bridge, switch or a router.But by the integrated manager agreement (ISMP) of piling up of preferred implementation, stacked switch 100-103 cooperates with each other and only carries out simple comprehensive function with the respective switch 100-103 that simulates a comparison independent operation and want much better than single switching equipment.The ISS120 that obtains is suitable for providing the fault-tolerance with enhancing and the extendible network capabilities of user friendliness.Specifically, the ISS 120 of preferred implementation can dynamically expand to comprise other stacked switch, detect idle stacked switch and pile up link, dynamically carry out route, and automatically configuring stacking switch 100-103 operates to help transmitting between stacked switch to produce a nonredundant identifying schemes around--passing through in some cases--idle stacked switch.

Fig. 2 shows the functional block diagram of a stacked switch that adopts in the ISS of preferred implementation system.This stacked switch comprises one or more Network Interface Modules (NIM) 204, one or more exchange control unit (network processing unit) 206, and administration module 220, all these preferably cooperate with each other, to receive data business volume and to send out data business volume via in the outside port 202 each.For present embodiment, the data that flow into switch 200 from another network node are called into data, and it includes protocol Data Unit.On the contrary, be sent to outside port 202 internally, be called data in order to the data that send to another network node, it comprises PDU.Each of a plurality of outside ports 202 all is a duplex port, is suitable for receiving data and sends out data.

NIM 204 and 204S preferably include and are suitable for exchanging for example one or more physical layer interfaces and Media Access Controlled (MAC) interface of the PDU of ethernet frame and IP grouping etc. and so on via the network communication link (not shown).In a plurality of outside ports 202, two exemplary stack port 202S are used for this given stack switch 200 is integrated into ISS 120.The NIM 204S that is associated with two stack port 202S for example is a standard ethernet port, be suitable for according to the stack arrangement pattern with communicate by letter between other stacked switch piles up and with various compatible nodes exchange traditional data traffic carrying capacity PDU.Go into PDU by means of one or more data/address bus 205A of going into, be transmitted to exchange control unit 206 from a plurality of NIM 204 and 204S.Similarly, go out PDU and be sent to a plurality of NIM 204 and 204S from exchange control unit 206 via one or more data/address bus 205B of going out.

Administration module 220 generally comprises a policy manager 224, is used for keeping and realizing the traffic carrying capacity strategy.The tactful the pick of of being realized by policy manager 224 is based on the second layer and/or the 3rd layer of addressing information from the source learning manipulation, the routing iinformation that receives from other routing device, and the filtering rule that utilizes Simple Network Management Protocol (SNMP) message 226 for example to upload by the network manager by a configuration manager 222.Transmitting the traffic carrying capacity strategy of concentrating expression to obtain from source study, other network nodes and keeper in 254, routing engine 230 can utilize them.

Except the traffic carrying capacity strategy, administration module 220 also comprises a central management module (CMM) 210, piles up function of exchange in order to realize ISS discussed in more detail below.The CMM 210 of preferred implementation comprises that a port status module 212 and one pile up manager 214.Port status module 212 is suitable for utilizing and for example keeps active signal, monitors the operating state of stack port 202S, and identification is connected to the existence of the adjacent stacks switch of stack port 202S.

CMM 210 in the preferred implementation comprises that is also piled up a manager 214, be suitable for participating in determining the election of each stacked switch management responsibility, processing is in order to the supervision messages of the state of monitoring other switches, and if desired, also serve as its responsibility and can comprise first host exchange (PMS) or second host exchange (SMS) that distributes the only element identifier to give the network processing unit of stacked switch 100-103 to one or more stacked switch 100-103 and distribution processor identifier.In a preferred embodiment, piling up that manager 214 has can be based on the communication protocol of only working fully with the message of adjacent neighbors.This be because operate on the stacked switch 100-103 pile up manager 214 examples must in addition before the topology of knowing exchange piling ISS, with other instance communications that operate on the distant-end switch.

In addition, pile up manager 214 and be suitable for determining ISS stacked switch topology, handle the relevant message of topology that exchanges between the ISS120 stacked switch.Specifically, pile up manager 214 and send the ISS topology request, send known ISS topology information, safeguard one or more local topology tables to other stacked switch.In some embodiments, pile up manager 214 and also be responsible for detecting losing of a unit, a unit is removed in the insertion of an other unit (thereby producing a trap) from pile up, and the operating state of determining relevant CMM 210.

Piling up manager 214 maintenances and upgrading one or more stacked switch neighbor table 218, preferably a table that is associated with each stack port 202S in the preferred implementation.Neighbor table 218 is tabulations of a stacked switch that can arrive by the stack port 202S that is associated.In a preferred embodiment, neighbours' stacked switch is arranged in order, wherein with the row of cells of this port direct neighbor first.Each of two stack ports is all safeguarded the neighbor table 218 of all adjacent stacks switches of having listed can arrive by this stack port.If the topology of ISS 120 is a loop configurations, these two neighbor table 218 have identical adjacent stacks switch tabulation so, but order is opposite.

Switch 200 preferably includes at least one network processing unit 206, and it can carry out but be not limited to the swap operation of the defined second layer (data link) and the 3rd layer (network) in the OSI(Open Systems Interconnection) reference model.Comprise IEEE (IEEE) 802.3 and I EEE 802.11 standards in order to outside port 202 is connected to a possible second layer assembly of protocols wired and/or wireless communication link, and the 3rd possible layer protocol set comprises the Internet engineer's task groups (IETF) request comment (RFC) 791 defined Internet protocol (IP) edition 4s and IETF RFC 1883 defined IP versions 6.

Network processing unit 206 preferably includes a routing engine 230 and a queue management device 240.Routing engine 230 comprises a grader 232, it receives PDU from data/address bus 205A, check one or more fields of this PDU, utilize Content Addressable Memory 233 that this PDU is categorized in a plurality of streams one, and retrieve forwarding information in 254 from transmitting of high-speed memory, keeping.Preferably include but be not limited to a flow identifier from transmitting the forwarding information that retrieves 254, be used to specify to preparing out necessary those of this specific PDU and transmit operation, wherein this flow identifier can comprise next hop address, and service type (COS) or service quality (QOS) regulation.

Forwarding processor 234 receives the PDU that goes into that has relevant forwarding information, and before sending to suitable outbound port, carries out one or more forwarding operations.These transmit that operation preferably includes but the conversion that is not limited to again encapsulation of data, one or more VLAN marks are appended to the VLAN mark push operation of a PDU, the VLAN that removes one or more VLAN marks from a PDU marks operation, the service quality (QoS) that keeps Internet resources, the charging and the book keeping operation of monitoring client traffic, the multiprotocol label switching (mpls) management, optionally filter the authentication of PDU, access control, comprise the high level study of ARP(Address Resolution Protocol) control, be used to regenerate and redirected PDU is used for the Port Mirroring of traffic analysis, source study, be used for determining the service type (CoS) of relative priority level, wherein distribute switch resource to give PDU by this relative priority level, be used to generate strategy and the color mark of traffic shaping, and be used for for example annotation management between the stacked switch of distributing PDU between the switch 100-103 of ISS 120 effectively.

After forwarding processor 234, PDU is sent to and is stored in the queue management device 240, sends PDU to suitable outbound port until there being bandwidth can be used for.Particularly, going out PDU is buffered among a plurality of priority queries in the buffer 242 one or more, until via output data bus 205B they being sent to outside port 202 by scheduler 244.

Network processing unit 206 can be equipped with an inner exchanging structure (not shown), each outside port 202 of switch can be connected to each other port on this switch.In a preferred embodiment, this inner exchanging structure can be configured to each the traffic carrying capacity of going into from two stack port 202S is sent to other stack ports by channel, bypass grader 232 and forwarding processor 234, with the penetrating pattern that realizes below will going through.

Fig. 3 shows a state diagram, and it is represented by an ISS stacked switch from starting to each stage that creates a mechanism automatically that complete mode of operation adopts.After initialization, but entering stackability, stacked switch 200 determines state 302, this switch determines that it is to be configured to serve as a stand alone exchange therein, still a stacked switch.But can there be to determine stackability according to physics and the work of stack port 202S.In certain execution mode of the present invention, in a switch, can there be stack port yet.If this switch is configured to carry out independent operation 304, pile up manager 214 and be under an embargo, this switch is according to the multilayer switch work with all FPDP 202.Yet when being configured to as a stacked switch, port status module 212 monitoring stack ports see if there is other message that keep alive message or indication adjacent cells to exist.Under the situation that does not have the adjacent stacks switch, this switch determines that it is one and piles up 306, and enter into forwarding state 308, it receives and sends data business volume on normal data port 202 at this state, monitor stack port 202S simultaneously, to introduce one or more other stackable unit.

If but and determine on stack port 202S, to have detected in the state 302 one or more switches at stackability, switch 200 enters into Discovery Status 310, to determine the topology of ISS 120.Stacked switch 200 can enter into election state 312 subsequently, determine to handle in the stacked switch executive responsibility of this state I SS 120, serve as first host exchange (PMS) and second host exchange in order to discern which unit, also be called first main frame and second main frame here respectively.

First main frame is responsible for the ISS management function, comprise for example distribution and the distribution of element identifier (element ID), all stacked switch among the ISS are by the mutual addressing of element identifier (element ID), and the processing of all command line interfaces input and synchronous reflection--that is to say, synchronously the different software versions on the stacked switch.It is the successor of this first main frame that second main frame is appointed as, if first hostdown or become is not worked, it bears the responsibility of first main frame.Although each of the stacked switch of the ISMP agreement of support preferred implementation can be born the work of first and second main frames, but remaining stacked switch is accepted host exchange, handles being chosen as main frame afterwards according to level discussed in more detail below election until wherein any one.

First main frame enters into unit number code assignment state 314 from election state 312, and each other stacked switch enters into forwarding state 308.In unit number code assignment state 314, a unit-of-work identifier of first host assignment (OEID) of selecting is given each still unallocated stacked switch of giving an element identifier (element ID).If effectively, OEID is arranged to equal the element identifier (element ID) of in the switch permanent memory, preserving (SEID).Though stacked switch uses OEID messaging between other stacked switch, OEID is only effective in the stacked switch life cycle, that is to say until restarting.If a stacked switch has an effective SEID when starting, OEID is configured to equal SEID.Otherwise, distribute to OEID of stacked switch by first main frame by unit number distribution mechanism, be arranged to equal OEID with SEID this moment.

When being operated in forwarding state 308, switch 200 is suitable for being transferred to supervision state 316 and penetrating (PT) state 320 and goes back to from supervision state 316 and penetrating (PT) state 320.In supervision state 316, switch 200 send supervision messages to its two adjacent neighbors to supervise, be similar to and be used to exchange the active mechanism of the maintenance that keeps alive message.For example, when a new stacked switch was inserted into ISS 120 or removes an existing switch, switch 200 automatically exchanged topology information with other stacked switch, and upgrades its stacked switch neighbor table.If first and second main frames break down simultaneously, remaining stacked switch--it is probably in forwarding state 308--enters into election state 312, to select a new main frame.If second main frame breaks down, not election, but first main frame is selected a responsibility of serving as second main frame of free cells.In case selected this unit, first main frame is informed whole piling up with new appointment by an election Indication message that is awarded the highest weight limit.If first main frame breaks down, do not have real election, but second main frame becomes the first new main frame with oneself rising to, and select one of free cells to become the second new main frame.In case selected this unit, the first new main frame is informed whole piling up with new appointment by an election Indication message that is awarded the highest weight limit.

In penetrating (PT) state 320, the FPDP 202 of stacked switch is under an embargo fully, and routing engine 230 is configured to and will be sent to opposite stack port from the data business volume of any one among its two stack port 202S.In PT state 320, routing engine 230 has been simulated between the stack port of two adjacent stacks switches one fixedly wired connection effectively, like this, has prevented the continuity of stacked switchboard system 120 otherwise an interruption will be arranged.Penetrating can being used for keeps, rather than close continuity with a common unit adjacent stacked inter-exchange, thus ISS 120 safeguarded, otherwise, wherein do not have the stacked switchboard system of realization example execution mode will cut off their ring topology, perhaps generate two independently chains.

In a preferred embodiment, this penetrating pattern can be got back in a unit, and wherein for example two unit have identical OEID.A plurality of stacked switch can have identical OEID at first, and they have disposed the default EID that for example identical manufacturer is provided with or have remained on identical user-defined SEID in the permanent memory.When two or more unit have identical OEID, one in two or more stacked switch, all stacked switch are all transferred to PT state 320, give new unique OEID of each stacked switch until first host assignment, and this moment, they can turn back to forwarding state 308.Particularly, first main frame adopts a global unit identifier (GID), sends its a new OEID to the unit in penetrating state 320, and reconfigures this unit and make it turn back to other administration orders of forwarding state 308 necessity.Global unit identifier (GID) comprises OEID and one or more attach identifier, the Media Access Controlled of stacked switch (MAC) address for example, the perhaps series number of switch for example, thus even the OEID of two switches is identical, also can discern each stacked switch uniquely.

As shown, a stacked switch can shift on the either direction between Discovery Status 310 and the supervision state 316, because supervision needs, and the Discovery Status 310 that detects neighbours as far back as one of them stacked switch is carried out, so, must carry out the task of supervision of describing in detail below.

A. but stackability is determined state

But Fig. 4 shows the flow chart that a stackability corresponding to Fig. 3 is determined state.After the stackable unit initialization, port status module 212 determines whether this switch is configured to as a stackable switch job by the network manager.If not, the answer of stack port state decision frame 402 is negated that switch carries out the forwarding operation of standard as a stand alone exchange, and does not possess the advantage of the integrated stacked switchboard system that adopts ISMP.If this switch is configured to as a stacked switch job, central management module 210 and pile up manager 214 and start (step 404) mode timers especially, this timer has defined distributes to switch, in order to the time of the connectedness that is determined to one or more other stacked switch.If in order to detect the stacked switch that these exist, pile up manager 214 and monitors on stack port 202S whether to exist maintenance alive message or indication to pile up link be other type signal that work.Other testing mechanisms can be monitored the signal of telecommunication that is produced by stacked switch, perhaps can exchange the maintenance alive message that is mutually related, their adopt message SN with send preceding maintenance alive message opening relationships, thereby guarantee two-way communication.After having confirmed an adjacent stacks switch, the answer of piling up link work test (step 408) is sure, and switch 200 enters into Discovery Status 310.At neither one adjacent stacks switch, and under the overdue situation of mode detection time, the answer of mode timers disappearance test 410 is sure, stacked switch 200 is initialized to one and piles up (state 306), thereby make the network manager can connect one or more stacked switch later on to stack port 202S, to construct an ISS 120.

B. Discovery Status

Fig. 5 is the flow chart of a Discovery Status corresponding to Fig. 3 (state 310).After entering Discovery Status 310, stacked switch 200 starts preliminary supervision action 501, and the topology of a plurality of stacked switch of definite ISS 120.Preliminary supervision action 501 comprises the beginning of a supervision timer and the initialization that keeps alive message.In start handling, send out the maintenance alive message as far as possible from stacked switch, with increase an adjacent stacks switch conclude do not exist by the accessibility switch of related stack port before, detect the probability of given stack switch.In some embodiments, a stacked switch also will be compared with local data in the information in the maintenance alive message that receives, and with preliminary change and the topology of determining in the responsibility appointment, just whether change first or second main frame.

It started from (step 502) and entered Discovery Status 310 by the discovery time period definition study topology institute's energy time spent.The topology request message of asking known stacked switch identity is sent (step 504) from each of the stack port 202S that detects an adjacent stacks subsequently.Similarly, switch 200 is also monitored (step 506) its stack port 202S, sees and whether sends topological related news from the adjacent stacks switch, and is over and done with at interval until discovery time.

If in office one piles up and detects a topology request on the port 202S, the answer that receives topology request test 514 is sure, and (step 516) replied with a topological Indication message in unit 200.The topology Indication message comprises the tabulation of the one or more stacked switch of the ISS 120 known to it, just piles up unit listed in the neighbor table 218 with opposite stack port is associated.When detecting the topology indication, answer is yes to receive topology indication test 512, and this switch 200 is compared the topology of (step 518) known stacked switch with the topology of this stacked switch tabulation of reception.If topology is different, if and be unique (test 522) in ISS 120 by the stacked switch identifier that each stacked switch provided, update content that stacked switch 200 upgrades in (step 524) its neighbor table 218--just that is associated with the stack port that receives topological Indication message thereon--is the identity of the stacked switch of study recently, preferably includes OEID, SEID, MAC Address, exchange hardware type and start-up time.In a preferred embodiment, if the adjacent stacks switch exists, stacked switch automatically sends (step 526) and has the renewal topology indication of another topology learnt recently from two stack port 202S to this adjacent stacks switch.

If by the neighbor table of piling up that is associated with first stack port, be called first and pile up neighbor table, given topological sum piles up neighbor table by what second stack port with stack port 202S centering was associated, be called second and pile up neighbor table, given is topological identical and opposite, then the answer of loop detection test 528 is sure, and switch 200 enters into the election stage 312.Yet, if stacked switch detects the identical of unit-of-work identifier that its uses and another switch--for example, when this may occur in wherein a plurality of stacked switch and has identical default value or user-defined stacked switch identifier--the switch with low priority identifier is transferred to penetrating (PT) state 320 immediately, and is uncertain to prevent addressing.

Suppose and in ISS 120, do not detect loop, discovery time will disappear, under the situation of not finding the adjacent stacks switch, stacked switch 200 enters into forwarding state 308, if perhaps in ISS 120, detect one or more adjacent stacks switches (test 530), then transfer to election state 312.

After having finished Discovery Status 310, all stacked switch have preferably exchanged topological Indication message and have learnt the OEID of all other stacked switch of existence in ISS 120.In a preferred embodiment, stacked switch is also propagated the start-up time and the MAC Address of each stacked switch, and they and OEID have constituted an election key together, are used for according to the process that describes below, and assist and pile up election.

C. elect state

Fig. 6 A and 6B have provided the flow chart corresponding to the election state of Fig. 3 together.After entering election state 312, stacked switch 200 is determined the management responsibility of stacked switch, and just this switch serves as first stacked switch, second stacked switch still is an idle stacked switch.First main frame, it comprises main CMM according to definition, be to be responsible for sharing out the work the switch of element identifier (element ID), and second stacked switch and first main frame is synchronous, becomes when not working at first main frame, prepares to bear the responsibility of first main frame.

If discerned first main frame, the answer of the known test 602 of first main frame is sure.If stacked switch 200 itself is first main frame (test 604), if second main frame still unidentified (test 610), this switch enters definite (step 612) second main frame.In a preferred embodiment, second main frame of selection is the stacked switch with the minimum election key that is only second to first main frame.In a preferred embodiment, the stacked switch that oneself is considered as first main frame is sent (step 614) election indication, declare that to all other switches in ISS 120 it is first main frame itself.If first main frame is known, but be different from current stacked switch 200, this switch enters into forwarding state 308, and accepts first main frame and carry out necessary ISS 120 bookkeepings, comprises the appointment of the new stacked switch identifier that for example below will go through.

Under the situation that does not have known first main frame, stacked switch 200 determines whether (step 606) second main frame is known.If switch 200 self is second stacked switch, this switch will oneself upgrade to first main frame, and sends (step 614) election Indication message before transferring to unit number code assignment state 314.If it is not it oneself that current stacked switch is known second main frame, this stacked switch is thought and oneself is born idle responsibility, enters into forwarding state 308.

Under the situation of not knowing first and second main frames, the process that stacked switch 200 provides according to Fig. 6 B attempts determining a unit of just electing ISS 120.Specifically, a plurality of stacked switch are preferably deferred to a level number scheme and are discerned the unit with minimum election key (to give a definition), although various other precedence scheme well known by persons skilled in the art also may be suitable for.This election key preferably includes start-up time, OEID and the MAC Address of switch.But represented switch power-up start-up time and entered the time of stackability determining step 302.OEID is one 8 bit number preferably, in order to discern each stacked switch uniquely from all other stacked switch of ISS 120.OEID is generally automatically assigned by first main frame, and is one or more although the network manager also can define.

All stacked switch have been found, and the election key of each switch is propagated to after all other switches among the ISS120, at first by comparing the election key of (step 620) its election key and all stacked switch, this locality determines whether it should bear the responsibility of first main frame to current stacked switch 200.If current switch 200 determines that it has the election key of a lower priority, the answer of election contest winner's test (step 622) is negated that this switch is transferred to forwarding state 308.Yet, if current switch 200 determines that it has the highest election key of priority, and therefore be the election contest winner, this switch is born the responsibility of first main frame temporarily so, start (step 624) and stipulated to distribute to the election timer that switch is finished election and converged to the time cycle of NextState, preferably three seconds.Stacked switch 200 in the preferred implementation is also distributed one or more election request messages (step 626), and it bears the responsibility of first main frame temporarily with other unit among the notice ISS.

Those skilled in the art will appreciate that a purpose of election request message and election timer is to prevent that calamitous unintentionally ground, two or more unit is the PMS that oneself upgrades among the same ISS 120.If for example in a stackable unit during the discovery stage (state 310) or pile up that link breaks down or a new unit powers up after finishing in the discovery stage immediately, interim like this ISS 120 is divided into two or more separated portions, each part is selected first a different main frame, and the one or more stacked switch among the ISS 120 believe that it is first main frame so.In order to prevent that this situation from causing ISS 120 paralysis, determine that it is each stacked switch distribution (step 626) one or more election request message of the preferred implementation of first main frame itself, preferably three, give all other stacked switch of temporarily bearing the responsibility of first main frame, like this, offer exemplary mechanism of competition switch, in order to relatively to come to assist the arbitration of first main frame by notice and key.

As mentioned above, if disappear (test 632) before at the election timer, switch 200 detects (test 636), and another elect request, and this switch is the election key of (step 644) its election key and interim other stacked switch with first host work relatively.Ask to show again that by repeating transmission (step 648) its election it is first main frame itself based on the switch that limit priority election key is selected to win, the switch of failure then enters into forwarding state 308.First main frame distribution (step 642) the election Indication message of winning is given all stackable unit, to inform the identity of first and second main frames that they are new.Except second main frame, idle responsibility is all born in other unit of not specifying responsibility at present after receiving this message.

Before the election efflux, the stacked switch that receives an election indication (test 634) does not send any further election request message by entering into forwarding state 308, implements this message.Generally speaking, this message must always be accepted in each unit that receives the election Indication message, unless the responsibility of first main frame has been born in another unit.This exception needs, because pile up combination situation, perhaps ring is gone up two faults of appearance and caused a plurality of first main frames to appear at same situation of piling up to coordinate.

Do not serve as under the situation of first main frame there being other stacked switch to attempt, the election timer disappears (test 632), and stacked switch 200 is assigned (step 640) one second main frames, sends an election and indicates (step 642); And enter into unit number code assignment state 314.

As discussed above, in a preferred embodiment, election state 312 adopts two classes election message, just (a) is used for asking the first main frame responsibility by the unit or asks the election request message of the identity of current main frame, and (b) be used for to its identity of other stack members broadcasting, and the election Indication message of the identity of second main frame by first main frame.Figure 11 has provided the representational election message format that piles up that adopts in the preferred implementation.Pile up election message 1100 and comprise a message type field 1102, in order to distinguish an election request and an election indication; An originator identity key 1104 comprises that an OEID and MAC Address are right, its this unit of best unique identification, with its with ISS 120 in all other cellular zones separate, and no matter two unit whether have identical OEID; A hop count field 1106, in order to following the tracks of when arriving the purpose stacked switch, this message propagation the number of unit of process; One first election key field 1108 comprises start-up time, OEID and MAC Address by first stacked switch of initiating to discern the unit; And one second election key field 1110, comprise start-up time, OEID and MAC Address by second stacked switch of initiating to discern the unit.Hop count field 1106 is defined according to a predetermined number by the stacked switch that sends this message at first.Send from the unit to the unit in the process of message in the stack port via them, hop count is stacked switch place decrement in the middle of each, is 0 until hop count, deletes this message this moment.

D. unit number code assignment state

Fig. 7 A and 7B have provided the flow chart corresponding to the unit number code assignment state of Fig. 3 together.After entering unit number code assignment state 314, first stacked switch of ISS 120 determines that those go back the unique slot numbering of neither one or have a unspecified slot numbering, for example the OEID of the unit of the default value of manufacturer's definition.As previously mentioned, OEID is used for addressing PDU and control messages between stacked switch.Under possible situation, first main frame preferably attempt with start from 1 or minimum available integer-valued sequential distributions OEID to continuous switch.

Shown in first stacked switch (PSS) OEID test 702, its unit-of-work identifier of PSS assessment, when needed, (step 704) oneself assigns minimum available OEID for it.For convenience of the network manager, first main frame has preferably been assigned element identifier (element ID) 1, if be not assigned to another unit before it, it should be available so.First main frame successively to each by with first stack port neighbor table associated index _ A, and with second stack port neighbor table associated index _ B and neighbor table 218 that two stack port 202S are associated in each stacked switch of providing operate accordingly.As mentioned above, index _ A be initialised (step 706) be 1, corresponding to the first module that provides in first neighbor table, increment (step 714) index _ A afterwards.If provided one or more adjacent stacks switches in first neighbor table, the answer of neighbours _ A test 708 is sure, and assigns (step 712) OEID who begins from the available cell label of minimum for each has the neighbours that do not specify OEID (710).Unit is starting or its configuration information does not have OEID when being eliminated usually for the first time.If this unit started in the past, and be assigned an OEID, this unit has an OEID usually so, and it is retained in the permanent memory with the form of SEID.Each port _ A neighbours' stacked switch just by the given sum of quantity _ neighbours _ A, is assigned an OEID when needed, and this moment, index _ A was greater than quantity _ neighbours _ A, and the answer of neighbours _ A test 708 is negated.If the feature of ISS 120 is loop topology for example, each stacked switch can be from the visit of first stack port, therefore, assigns an OEID under the given process in Fig. 7 A.In this case, the answer of the go-and-return test 720 of Fig. 7 B is sure, but and the OEID that assigns be sent to each applying unit.

Referring to Fig. 7 B, in that finish can be after the OEID appointment process of the unit of first stack port visit, first main frame be from the addressable unit appointment of second stack port OEID.In a preferred embodiment, the order of assigning depends on the topology of ISS 120, and from the quantity of the unit that each stack port can be visited.If the feature of ISS 120 topologys is chains, and the unit that can visit from two stack port 202S is arranged,, assign OEID to the unit that leaves second stack port with ascending order then from farthest unit.Therefore this OEID assignment scheme assigns OEID to give in the unit in a chain, just as these unit connect in a ring, thereby provides a kind of numbering plan directly perceived of the network manager's of being convenient to use.Specifically, if there is the unit to arrive by first stack port--that is to say, quantity _ neighbours _ A test 722 is negated that----index in by neighbor table _ B=quantity _ neighbours _ B provides--assigns (step 732) minimum available cell label from port _ B adjacent stacks switch farthest, if (test 730) OEID is not also assigned in this unit.Show as decrement step 734, reduce index _ B, assign OEID to give next nearest stacked switch as required, until handling the adjacent stacks switch that leaves this second stack port.After OEID had all been assigned in each unit, the answer of index _ B test 726 was sure, and element identifier (element ID) transmit (step 728) and give their stacked switch separately, and first main frame enters into forwarding state 308 with the form of OEID assignment messages.

Opposite with top description, do not having the adjacent stacks switch can be when first stack port is visited--just quantity _ neighbours _ A equals 0, and to test 722 be sure--start from the given adjacent neighbors in unit at first listed in neighbours _ B table, OEID assigns generally from the available OEID of minimum.Specifically, the stacked switch of index _ B indication position in the neighbours _ B table--just neighbours _ B[index _ B]--has been assigned (step 746) minimum available OEID, all assigned OEID (step 744) until each unit that needs OEID, and traveled through whole neighbours _ B table, this occurs in index _ B greater than quantity _ neighbours _ B, and regularly whether the answer of neighbours' testing procedure 742.

Fig. 8 A and 8B had provided respectively before the assignment unit number is given the left side, and the assignment unit number is given after the right the example of ISS system 800.In these examples, ISS 800 comprises four stacked switch 801-804, with the form connection of loop or ring.Before unit number code assignment state 314, all there is a unspecified OEID each unit, and the 3rd switch 803 is taken on the responsibility of first main frame.According to the present invention, the 3rd switch 803 has been assigned minimum available OEID to oneself, and just one (1), and assign in the mode that increases progressively successively can be from first stack port, port _ A just, the unit of visit.Shown in Fig. 8 B, assigned OEID two (2) with PSS 803 ports _ A adjacent stacked switch 804, next nearest stacked switch 801 has been assigned OEID three (3), has assigned OEID four (4) from first main frame, 803 ports _ A stacked switch 802 farthest.After the message of the OEID that receives all unit of broadcasting from first main frame, each stacked switch 801-804 stores it in volatile memory new OEID assigns, its SEID is arranged to equate with its new OEID, its SEID is stored in the permanent memory.Also comprise the MAC Address that is associated with each OEID from message from the broadcasting OEID of first main frame.

Fig. 9 A and 9B had provided respectively before the assignment unit number is given the left side, and the assignment unit number is given after the right the example of ISS system 900.In these examples, ISS 900 comprises four stacked switch 901-904, with the form connection of chain.Before unit number code assignment state 314, all there is a unspecified OEID each unit, and the 3rd switch 903 is taken on the responsibility of first main frame.According to the present invention, the 3rd switch 903 has been assigned minimum available OEID to oneself, and just one (1), and assign in the mode that increases progressively successively can be from first stack port, port _ A just, the unit of visit.Shown in Fig. 9 B, assigned OEID two (2) with first main frame, 903 ports _ A adjacent stacked switch 904.Owing to exist the stack port _ A of the one or more stacked switch and first main frame 903 adjacent, first main frame is assigned OEID from from port _ B stacked switch farthest with ascending order.Particularly, first main frame 903 is assigned next minimum available element number, and just OEID three (3) gives far-end stacked switch 901, and next stacked switch 902 has then been assigned OEID four (4) in the first main frame direction.Like this, by the automatic unit number of PSS 903 assign caused a unit number go up ascending chain successively, the stacked switch chain seems a ring, thus for the network manager automatically provide a kind of convenience and intuitively numbering reach an agreement on.

Figure 10 A and 10B had provided respectively before the assignment unit number, and after the assignment unit number, the example of ISS system 1000.In these examples, ISS 1000 comprises four stacked switch 1001-1004, with the form connection of ring.Before unit number code assignment state 314, some but be not the element number that there is former appointment whole unit, just first module has OEID two (2), there is OEID one (1) Unit the 3rd.As before, the 3rd switch 1003 is taken on the responsibility of first main frame.According to the present invention, be assigned minimum available cell numbering with first main frame, 1003 ports _ A adjacent stacked switch 1004, just OEID three (3).Proceed to next stacked switch that can arrive by first main frame, 1003 ports _ A, first main frame is assigned available OEID with the form of ascending order.Specifically, first main frame 1003 assigns next minimum available cell numbering four (4) to give stacked switch 1002.

E. forwarding state

Referring to Fig. 3, after unit number code assignment state 314, the stacked switch of ISS is suitable for transferring to forwarding state 308 once more, and respective switch 100-103 is suitable for sending data business volume to all other switches therein.In a preferred embodiment, communication between the stacked switch can be adopted an inter-exchange label, be used for transmitting effectively PDU through ISS120, and each receiving element is not carried out and is independently classified and transmit operation, these are operated very consumption of natural resource and sometimes are unnecessary, suppose that wherein stacked switch is consecutively connected to each other.The inter-exchange label that adopts in the preferred implementation comprises a protocol identifier, the OEID of source unit, and the OEID of purpose unit, their prefixes are to the PDU that transmits through ISS.After the stacked switch between the source and destination unit received PDU, this temporary location was checked the protocol identifier label, determines it oneself is not the purpose unit, and PDU is forwarded to the port opposite with the port that receives this PDU.Label is removed in the purpose unit from PDU after receiving, this PDU is transmitted to FPDP 202, thereby leads to final purpose node specified in this PDU.

F. penetrating state

As shown in Figure 3, one or more stacked switch of preferred implementation are suitable for comprising forwarding state 308 from one or more other states, transfer to penetrating state 320.Penetrating (PT) state 320 provides a kind of mechanism, and when forbidding the function of exchange of a unit, it can prevent that data business volume from interrupting on ISS 120, still allows the network manager to monitor concurrent delivery simultaneously and puts order and give the unit of forbidding.Particularly, for example forbid carrying out the unit of exchange and Route Selection operation and transfer to penetrating state 320, wherein the traffic forwarding that receives from a stack port 202S is to opposite stack port.202 of FPDP are opposite, under a kind of blocking model of forbidding the traffic carrying capacity transmitting-receiving.Can not carry out second layer exchange and the 3rd layer of Route Selection even skilled person in the art will appreciate that a discrete cell, PT state 320 still can be safeguarded the transmission continuity on the ISS.

In a preferred embodiment, in following situation one or more time of unit, enter PT state 320: exchange control unit identifier (SCID) is not enough; The unit is too many in piling up; Identical OEID, just same OEID is used by two or more unit; And the request of the user by administration order.In a preferred embodiment, first main frame is under an embargo and is operated in PT state 320.Having only responsibility is that second main frame, free time or the stacked switch of not assigning could be worked under this state.Second main frame or idle responsibility can be born in unit in PT state 320, and do not need to restart.

In a preferred embodiment, the unit in PT state 320 comes unique identification by the penetrating element identifier (element ID) (PTID) of a special range.PTID is similar to OEID, and just PTID is an operation time, in order to allow this pile up in all software applications can both be identified in unit in the penetrating state, and OEID is all other states, comprises the formant identifier that forwarding state 308 is for example used.The scope of PTID preferably and the OEID scope is same prolongs, and has the digimigration amount.If for example element number changes between 1 and 100, the scope of PTID can from 101 to 200, and just the maximum quantity of unit adds 100.PTID is assigned by the penetrating assignment messages of first host computer using, and this penetrating assignment messages comprises the tabulation of one or more unit, and--by MAC Address and OEID identification--these are the reasons that enter PT state 320 and change.

Before entering PT state 320, the manager 214 that piles up that enters the switch of penetrating pattern must be with status change notification to interested software application, this unit of notification application is not re-used as the folded cell operation of standard pile, and allows switch to forbid all network user's ports 202.In addition, pile up manager 214 and also notify this switch of all unit among the ISS 120 to change PT state 320 over to, and the reason that enters this state is provided, and together with the former OEID of switch.A unit preferably produces suitable daily record and management trap, to give the network manager with this event alarm.

When PT state 320 is transferred in a unit, assign a PTID by first main frame, this PTID is sent to this unit, and the mutual middle use PTID of what is the need breath in office, rather than former OEID.Unit in PT state 320 is general only to be communicated with the management application program by piling up the manager protocol transmission.

From topological angle,, regard the unit in the PT state 320 as a jumping in order to determine hop count.For example, in a stacked switch sequence that is provided by 1-101-4-5-6, wherein the unit that is associated with 101 is in penetrating pattern, and unit 4 differs double bounce with unit 1.Therefore, piling up before manager 214 transmitting of the unit under the penetrating state, be responsible for the hop count field in the message that receives is carried out the phase reducing.In the variation of internal communication mechanism, wherein do not adopt hop count field, opposite stack port 202S is always given with the forwards that receives in this unit.

A unit in penetrating state 320 is suitable for sending data business volume between stack port, and receives the ISS administrative messag that is mail to PTID by first main frame.In a preferred embodiment, penetrating unit receive and handle following receive pile up protocol message: supervision messages, topology request, topological Indication message, election Indication message, the SCID identifier is assigned and the heavy duty request, just restarts the request of a unit.Penetrating unit preferably also is suitable for producing and piles up below the transmission protocol message: supervision messages, topology request message and topological Indication message.A unit in penetrating state does not generally participate in election as a candidate and handles.Yet it preferably can visit the identity of first and second main computer units.It preferably also participate in identification repeat main frame, repeat OEID and even repeat the processing of PTID, pile up merging at two, and each piles up when the PT unit with identical PTID is arranged, above-mentioned situation may take place.

Following paragraph has been summed up a unit and how have been responded SCID by outspent situation, stacked switch many ISS situation about can't support, the situation that repetition OEID is arranged, and the user asks the situation of penetrating state indirectly by administration order, arrives penetrating operator scheme.The numbering of summing up in the paragraph is for each example of classifying, and does not reflect unique sequence arrangement of illustrative steps.

SCID consumes excessively:

1. the decision of a unit being put into PT state 320 is made by first main frame.One of first main frame distribution has the penetrating assignment messages that is entered the reason of the unit list of PT state and each unit by instruction.Each unit that enters PT state 320 for needs distributes a PTID.

2. after receiving a PT assignment messages, whether unit inspection it transferred to penetrating state by instruction.If the PTID of appointment in this message is changed into its OEID in this unit, and writes down the reason that enters penetrating pattern.This unit is subsequently:

3. remove a market bit that is associated with the SCID request message, to stop to ask a SCID;

4. the responsibility of upgrading it is penetrating;

5. notify interested application program; And

6. its switching fabric (not shown) of network processing unit 206 is programmed, make the legacy user's traffic carrying capacity that receives from stack port _ A send from stack port _ B, perhaps opposite.

The unit is too much in piling up

3. the responsibility of upgrading it is penetrating;

4. notify interested application program; And

5. its switching fabric (not shown) of network processing unit 206 is programmed, make the legacy user's traffic carrying capacity that receives from stack port _ A send from stack port _ B, perhaps opposite.

Repeat OEID

1. the decision that enters PT state 320 is at first initiated by local stackable unit.When a unit knew that another stackable unit has identical OEID, it compared its election key and the election key of far-end repetitive.If its election key is lower than the election key priority of this another unit, then should abandon this OEID in unit, this locality.

2. this unit changes over a special use (special) value with its OEID, is called here and piles up the repetitive identifier, for example is worth 254, and it is an interim ID, uses before receiving PTID.

3. work as first main frame by for example topology or supervision messages, when the PTID that detects a unit equals this and piles up the repetitive identifier, it distributes a unique PTID to give this unit, and is set to indicate the PT assignment messages of repetition OEID that this new PTID is distributed to this unit by a field reason wherein.Each unit that enters penetrating pattern for needs distributes a new PTID.

4. after receiving a PT assignment messages, whether unit inspection it transferred to PT state 320 by instruction.If this unit changes over the PTID of appointment in this message with its OEID, and write down the reason that enters penetrating pattern.This unit is subsequently:

5. the responsibility of upgrading it is penetrating;

6. notify interested application program; And

7. its switching fabric (not shown) of network processing unit 206 is programmed, make the legacy user's traffic carrying capacity that receives from stack port _ A send from stack port _ B, perhaps opposite.

When asking by the administration order user

In some cases, even without conflict occurring or lack resource situation, PT state 320 still comes in handy.Embodiments of the present invention provide a kind of administration order, and the network manager can come " removing " SEID, the element number of just storing by it in permanent memory.Although needn't adopt PT state 320 to upgrade the element identifier (element ID) of reservation, when element number reaches maximum, penetrating state be put in a unit can help to renumber to piling up.For example, the maximum quantity of supposing stackable unit is 4, and the scope of the OEID of permission is [1..4], and each unit initial configuration becomes a 1-4-3-2 ring, and these unit need renumber, and makes that final configuration is 1-2-3-4.

The new element identifier (element ID) that is kept in the permanent memory if this realization needs reenable becomes unit-of-work identifier, so generally do not cut apart this pile up and the unique method that reaches this point preferably by restarting whole piling up.Can adopt following process:

A) piling up removing OEID=2 is penetrating state, is 1-4-3-101 thereby cause work topology: OEID, and SEID is 1-4-3-0;

B) being stacked as the unit with OEID=4, SEID=2:OEID is set is 1-4-3-101, and SEID is 1-2-3-0;

C) heavily loaded unit OEID=4, thus after restarting, OEID is 1-2-3-101, and SEID is 1-2-3-0;

D) heavily loaded unit PTID=101, thus after restarting, OEID is 1-2-3-4, and SEID is 1-2-3-4;

In this way, in fact just restarted the unit that need renumber.

It is as follows that the sequence of events of penetrating state is transferred in user request:

1. the decision that enters penetrating pattern is at first made by first main frame.One of first main frame distribution have by instruction enter the unit list of PT state and wherein the field reason be set to manage the penetrating assignment messages of removing slot.Each unit that enters the PT state for needs distributes a PTID.

2. after receiving a PT assignment messages, whether unit inspection it transferred to penetrating state 320 by instruction.If the PTID of appointment in this message is changed into its OEID in this unit, and writes down the reason that enters penetrating pattern.This unit is subsequently:

3. from permanent memory, remove this SEID;

4. the responsibility of upgrading it is penetrating;

5. notify interested application program; And

G. supervise state

Referring to Fig. 3, all stacked switch of preferred implementation all are suitable for producing regular supervision messages once more, and they are sent to and their adjacent neighbors.In this way, the variation that each stacked switch can the monitor network topology.Specifically, the exchange of supervision messages make the unit can: when detect an adjacent neighbors and be inserted into and pile up, detecting the physical connection of when leading to adjacent neighbors is changed, detect with communicating by letter of neighbours and whether lose, even and the physical link between these neighbours still work.In a preferred embodiment, each stacked switch of ISS 120 is termly from one or more other states, comprise Discovery Status 310, election state 312, unit number code assignment state 314, forwarding state 308 and penetrating state 320, transfer to supervision state 316.Yet, skilled person in the art will appreciate that topological supervision state 316 and forwarding state 308 can work simultaneously, therefore can be merged into single status.

Figure 12 has provided the content of piling up supervision messages according to preferred implementation.This supervision messages comprises a type of message 1202, and it is identified as a supervision messages with this message; An originators information field 1204 comprises OEID, the MAC Address of transmitting element, the stack port that starts and unit hardware type; One first identification key field 1206 comprises the OEID and the MAC Address of first main frame of being discerned by transmitting element; One second identification key field 1208 comprises the OEID and the MAC Address of second main frame of being discerned by transmitting element; A separate marking field 1210 shows when it is set up that transmitting element does not have the adjacent stacks switch on opposite stack port; A SCID request matrix field 1212 provides its processor ID, for example the tabulation of exchange control unit ID stackable unit unknown or that do not assign.

501 supervision timer sends supervision messages from stack port 202S according to being initialised after entering Discovery Status 310.In a preferred embodiment, the supervision messages per second sends termly, although the time can be according to for example topology change.As following detailed description, for example, can adopt the supervision messages exchange: (1) is as the active mechanism of a kind of maintenance; (2) in ISS120, detect (a) topology change, (b) have the unit of repetition OEID, (c) repetition among the ISS first or second main frame; (3) activate the processing that integrality is piled up in checking; And (4) if desired, and the topological sum stackable unit information of piling up takes action to correct and upgrade.

In a preferred embodiment, stacked switch 200 is suitable for monitoring and piles up Link State, piles up link with the physics that is determined to neighbours and when works or do not work.When one was piled up link and starts working, this unit sent a topology request message to these neighbours that detect recently.When one was piled up link and quits work, this unit carried out: (a) reset corresponding to the out-of-work neighbor list that piles up link; (b) send a topological Indication message from opposite stack port, change in topology is propagated to other neighbours; (c) handle responsibility and change, comprise and select second a new main frame, take over the responsibility of first main frame,, ask the responsibility of first main frame if no longer include first or second main frame on the ISS 120 if known first main frame has broken down.

As mentioned above, supervision messages is sent to the adjacent neighbors on its both sides termly by each stackable unit.This exchange makes that each stacked switch can be soon: (a) detect when an adjacent neighbors is inserted into ISS 120; (b) detect the physical connection of when having revised with an adjacent neighbors, for example when changed one and piled up link cables; (c) confirm that existing neighbours are still active; And (d) detect--have one or more continuous supervision messages based on expection, even but do not receive from neighbours--when and the physical link between neighbour still work, but lose with communicating by letter of these neighbours.If do not receive one or more continuous supervision messages, then stacked switch 200 these neighbours of supposition no longer work, and under the situation that known second main frame breaks down, activate responsibility and change processing, to select second a new main frame, the responsibility of this first main frame of situation lower linking tube that breaks down at known first main frame if when perhaps no longer including first or second main frame on piling up, is asked the responsibility of first main frame.

The topology change of particular type is depended in the action of being taked by stacked switch 200.If receive a supervision messages, and detect any ISS topology change, stacked switch 200 generates a topological Indication message, then it is sent to the opposite link that piles up.If receive a supervision messages, and detect new neighbours, stacked switch 200 sends to the known ISS topology of these new neighbours at the opposite link that piles up, just with neighbor table that opposite stack port is associated in the stacked switch neighbours that list.If receive a supervision messages that is provided with a separate marking, stacked switch 200 upgrades the neighbor table that is associated by removing any neighbours that can arrive by the unit that sends this supervision messages.If received a supervision messages that separate marking is not set, and the neighbor table that is associated with this stack port shows and has single neighbours that stacked switch 200 sends the originators of a topology request message to this supervision messages.If receive a supervision messages, and the cable line that piles up link changed, stacked switch 200 reset or otherwise remove and the neighbor table that link is associated of piling up that receives this message, and by a known topology of topology indication transmission to these neighbours.After having carried out appropriate responsive, stacked switch 200 shifts back into going into supervision state 316 state before.

Persons skilled in the art can understand that the supervision messages exchange has many advantages.Specifically, because neighbours' monitoring features can detect neighbours' fault apace, and when one is piled up link and quits work, when perhaps an adjacent neighbors is not made a response to a plurality of continuous supervision messages, handle responsibility immediately and change the fast quick-acting pipe that obtains.In addition, the stacked switch in some execution mode can be forbidden supervision state 316 provisionally, just for example prevents the transmission of supervision messages.If fault has for example taken place, and first main frame predicts shutdown and restarts, and when preparing afterwards an investigation and analysis file, first main frame can be distributed a supervision suspending message and give all stackable unit.As response to this supervision suspending message, the unit of ISS 120 stops the transmission of supervision messages, if it does not meet with a response, can cause stacked switch to be supposed mistakenly and lose first main frame, and turn to second main frame, thereby after former first main frame is finished the preparation of investigation and analysis file afterwards, cause having occurred two first main frames.After first main frame is finished the preparation of investigation and analysis file afterwards, can be by its gating supervision state again automatically.In some embodiments, after defined predetermined amount of time,, wherein supervise state not by the situation of first main frame gating again to handle by unit oneself gating supervision state 316 again automatically by timeout parameter.This timeout parameter should be enough must be long, make the main frame of winning can finish the preparation of investigation and analysis file afterwards.

In some embodiments, after receiving a supervision messages, the SCID matrix of its this locality is also checked in a unit.If a unit requests SCID, the local copy of its local SCID matrix is checked in this unit, looks at whether it knows that the SCID of this discrete cell assigns.If, it sends a SCID assignment messages to all stack members by internal communication mechanism.If the recipient of supervision messages does not know that SCID assigns, it will be provided with a mark corresponding to the requesting party unit in overall SCID matrix, and the matrix after will upgrading sends opposite neighbours unit to.

Although above description has comprised a lot of regulations, these regulations are not to be construed as limiting scope of the present invention, but the explanation to some current preferred implementations of the present invention only is provided.

Therefore, the present invention comes open by example, but is not limited to these examples, should determine scope of the present invention with reference to appending claims.

Claims

1. stacked switchboard system that comprises a plurality of stacked switch, described a plurality of stacked switch comprise:

One or more by the pipe stacked switch, wherein said one or more each by the pipe stacked switch all are suitable for sending supervision messages one or more in described a plurality of stacked switch with the form that keeps active indication, and

One first main stacked switch is suitable for sending a supervision suspending message, and is described a plurality of by the supervision messages of the one or more expectations in the pipe stacked switch from other stacked switch temporarily to prevent.

2. according to the stacked switchboard system of claim 1, the wherein said first main stacked switch is suitable for removing this supervision suspending message.

3. according to the stacked switchboard system of claim 2, wherein after described first main frame is finished the preparation of investigation and analysis file afterwards, automatically remove described supervision suspending message by described first main frame.

4. according to the stacked switchboard system of claim 1, wherein saidly one or morely be suitable for selecting first a new main stacked switch by the pipe stacked switch, if the current first main stacked switch can not be finished:

Send described supervision suspending message; And

Supervision messages to one or more transmissions responds.

5. according to the stacked switchboard system of claim 1, wherein said one or morely be suitable for after a predetermined amount of time by the pipe stacked switch recovered the transmission of supervision messages.

6. according to the stacked switchboard system of claim 1, wherein saidly one or morely be suitable at timeout parameter after expired and supervision suspending message is removed the transmission that recovers supervision messages by the pipe stacked switch.

7. according to the stacked switchboard system of claim 1, wherein said supervision messages comprises the stacked switchboard system topology information.

8. according to the stacked switchboard system of claim 1, wherein said stacked switchboard system topology information comprises: the work stackable unit identifier of the stacked switch that starts, and the work stackable unit identifier with respect to the first main stacked switch of the described stacked switch that starts.

9. according to the stacked switchboard system of claim 1, wherein said supervision messages is notified:

A new stacked switch is inserted into described stacked switchboard system;

The modification of a physical connection between two in described a plurality of stacked switch; And

Communication failure between two in described a plurality of stacked switch.