Background technology
The virtual network device technology be generally used for a plurality of physical equipments virtual be a network equipment, thereby improve the extensibility that the network equipment is disposed, reach the purpose of the passing investment of protection user.The IRF(Intelligent Resilient Framework that the applicant proposes, intelligent elastic framework) be exactly a kind of popular virtual network device.The IRF technology is by linking together multiple devices by the IRF physical port, carry out necessary configuration after, be virtualized into one " distributed apparatus ".IRF comprises a plurality of member devices, and member device has two kinds of roles: Master equipment (main equipment) and Slave equipment (from equipment).Master equipment is responsible for the Operations, Administration and Maintenance of whole IRF; Slave equipment is mainly for the treatment of professional and as the backup of Master equipment.This technology mainly contains following advantage:
1. management is simplified: the user can login the IRF system by the arbitrary port of any member device after IRF formed, and all member devices in the IRF are carried out unified management; Therefore on bookkeeping, more flexible compared to legacy network devices.
2. high reliability: in case the Master equipment fault, other Slave equipment can the new Master equipment of rapid automatic elections, does not interrupt to guarantee business, thereby realized the 1:N backup of equipment.IRF link between the member device is supported polymerizable functional, can backup each other between the multilink and also can carry out load balancing, thereby further improve the reliability of IRF.
3. extended capability is strong: by increasing member device, can improve rapidly the traffic handing capacity of whole IRF.
In actual moving process, the IRF system usually will experience the management of physical connection, collecting topology, role's election, IRF and safeguard four-stage.Need to set up first the IRF physical connection between the member device, then can automatically carry out collecting topology and role election, finish the foundation of IRF, after this enter the stage that administers and maintains of IRF.
The connection topology of IRF has two kinds: chain connects and the connection of being connected.The physical location that chain connects member device requires the connection of ratio annular low, is mainly used in the networking that the member device physical location disperses.Annular connects more reliable than the chain connection.Because when in chain connects, link failure occurring, can cause the IRF division; And annular connect in during certain bar link failure, can form chain to connect, the business of IRF can not be affected.
In IRF, the topology that each member device and neighbours' member device are collected whole IRF by mutual IRF hello packet.The IRF hello packet can carry topology information, specifically comprises the contents such as bridge MAC of IRF port annexation, member device numbering, member device priority, member device.Each member device participates in the operation of IRF control plane and is responsible for by the primary main control module in this locality.Each member device is in the known topology information of local record self.When equipment had just started, local primary main control module had only recorded the topology information of self.After the IRF port status became up, local primary main control module can start the collecting topology operation, and through collection after a while, all member devices all can be collected complete topology information.Can get inside the character that one is playing the election stage this moment.Role's election meeting is carried out in following situation among the IRF: IRF sets up, Master equipment leaves or fault, IRF merge etc.After role's election was finished, IRF formed, and entered IRF management and maintenance phase.After role election was finished, IRF can put into operation, and all member devices form virtual units and are present in the network, and the resource on all member devices returns this virtual unit to have and by the unified management of Master equipment.
If certain member device fault or IRF link failure, its neighbor device can be notified the information broadcasting of " member device leaves " to the miscellaneous equipment among the IRF immediately.The member device that gets access to leave group message can judge that leave is Master or Slave according to the IRF topology information table of local maintenance, if that leave is Master, then triggers new role's election, upgrades local IRF topology again; If that leave is Slave, then directly upgrade local IRF topology, can restrain rapidly to guarantee the IRF topology.
Yet because role election carries out automatically, when the fortuitous event such as link failure therefore occurring at IRF, might cause an IRF to be split into a plurality of new IRF.Such as chain topology opens from the centre division, and perhaps two link failures etc. occur a ring topology simultaneously.These new IRF can use original IRF configuration, and this will cause a plurality of new IRP to use identical three layers of configuration such as IP address, and obviously this can cause IP address conflict, so that Single Point of Faliure causes the overall network problem.
In order to address the above problem, prior art has proposed MAD(Multi-Active Detection, and multi-Active detects) technology deals with.The MAD technology can be passed through LACP(Link Aggregation Control Protocol usually; Link Aggregation Control Protocol), BFD(Bidirectional ForwardingDetection; the two-way converting detection), ARP(Address Resolution Protocol; Neighbor Discovery Protocol) etc. address resolution protocol) or ND(Neighbor Discovery Protocol, variety of protocol comes whether to exist in the Sampling network a plurality of IRF.
It is not only the challenge that IRF faces that multi-Active detects this problem, is other problems that face based on the virtual network device of other technologies framework equally.Although the mechanism than more rich solution Active collision detection and processing has been arranged at present, yet these existing technology also face following various predicaments (still take IRF as example):
Mutual IRF protocol massages between member device transmits via IRF link between the IRF member device, and this can take the bandwidth of cascade data channel on this link undoubtedly; When especially IRF link topology changed, a large amount of protocol massages can impact need to stride the forward process that frame is transmitted the data message of (namely striding member device transmits).
Although existing MAD detection technique can be avoided the problem of a plurality of new IRF conflicts (namely multi-Active conflict) as far as possible, but these MAD technology have the separately restriction of application scenarios usually, and are to consume possibly more multi-system resource (such as IP address etc.) or forwarding bandwidth.Please refer to Fig. 1, for LCAP MAD mode, need to use the intermediate equipment of supporting LACP protocol extension function in the networking.Please refer to Fig. 2, for BFD MAD, must have BFD MAD to detect link between all member devices, these link-attached interfaces must belong to same VLAN, dispose different IP addresses under the same network segment under this VLAN interface view different members equipment.Please refer to the networking mode of Fig. 1, for ARP MAD, different from the LCAP mode is that ARP MAD does not need intermediate equipment to support LCAP, but but need all to want on intermediate equipment, Master and the Slave configuration integrate tree function (STP), to prevent from forming loop.
In addition, when IRF detects the IRF of other Active states (professional effective status), for fear of conflict, need to close all service ports of some IRF member device, make it to be in Recovery state (professional disabled status), this handling process can avoid problem to spread in network, but present MAD art does not have algorithm preferably to select to be set to the IRF member device of Recovery state.For the IRF member device of Recovery state, must restart this IRF member device and just can rebulid the IRF connection simultaneously.
Summary of the invention
In view of this, the invention provides a kind of virtual network device management devices, be applied on the member device in the virtual network device, wherein member device comprises cascade data port and cascade control port, and the cascade data port of member device links to each other to provide the cascade data passage with neighbours member's cascade data port by the cascade data link; The cascade control port of this member device links to each other to provide the cascade that physically is independent of this cascade data passage control channel with neighbours member's cascade control port by the cascade control link; This device comprises:
Fisrt fault perception unit is used for perceiving at the cascade data passage that virtual network device breaks down and send the division notice to the first control unit when being split into a plurality of new virtual network device;
The first control unit is used for controlling alternately by cascade control channel and other member devices; Wherein
The first control unit is further used for when the main equipment of the new virtual network device of this member device conduct, notify according to the division that Fisrt fault perception unit sends, by the mutual separately validity control information of affiliated virtual network device of the main equipment of cascade control channel and other new virtual network devices; And the availability negotiation that basis is preset after receiving from the validity control information of other virtual network devices rule is determined the state of this virtual network device, if determine that this virtual network device is professional effective status, the traffic handing capacity that then keeps each member device in this virtual network device, if determine that the state of this virtual network device is professional disabled status, then the Business Processing of each member device in this virtual network device gone to enable.
The present invention provides a kind of virtual network device management method simultaneously, be applied on the member device in the virtual network device, wherein member device comprises cascade data port and cascade control port, and the cascade data port of member device links to each other to provide the cascade data passage with neighbours member's cascade data port by the cascade data link; The cascade control port of this member device links to each other to provide the cascade that physically is independent of this cascade data passage control channel with neighbours member's cascade control port by the cascade control link; The method may further comprise the steps:
Steps A, perceive at the cascade data passage that virtual network device breaks down and send the division notice when being split into a plurality of new virtual network device;
Step B, control alternately by cascade control channel and other member devices;
Step C, at this member device during as the main equipment of new virtual network device, according to the division notice, by the mutual separately validity control information of affiliated virtual network device of the main equipment of cascade control channel and other new virtual network devices;
Step D, the availability negotiation rule that basis is preset after receiving from the validity control information of other virtual network devices are determined the state of this virtual network device, if determine that this virtual network device is professional effective status, the traffic handing capacity that then keeps each member device in this virtual network device, if determine that the state of this virtual network device is professional disabled status, then the Business Processing of each member device in this virtual network device gone to enable.
Compared to existing technologies, the present invention has simplified the processing of multi-Active conflict greatly, in preferred mode, abundant compatible existing treatment mechanism also, backup each other with existing treatment mechanism, can also support that the regular new virtual unit of optimum of selecting is as effective virtual unit through consultation simultaneously in the virtual unit division.
Embodiment
The invention provides a kind of new solution and overcome the weak point of prior art on virtual network device division treatment mechanism.Below describe as an example of IRF example, to be applied in other virtual network devices technical but the present invention does not get rid of.
For IRF, the message that transmits between each member device mainly comprises two kinds: business datum message and control message.So-called service message is also referred to as data message, normally come in the automatic network, enter from certain service port of certain member device of virtual unit, this member device of process is determined port after tabling look-up, if this outbound port is positioned on other member devices, then need to pass on this member device by the cascaded link between the member device, the forwarding of data message can be understood as the business that whole virtual unit is born.So-called control message is also referred to as the IRF protocol massages usually, is mainly used between member device transmitting control information, and main effect allows to carry out between the member device coordination of the aspects such as professional.Also may relate to administrative message mutual between webmaster and the member device between ex-officio member's equipment, but this not outline of the present invention, is no longer discussed.
The present invention increases the independently cascade control link that the user is interconnected on the basis of cascade data link for each member device.In the preferred embodiment of the present invention, each member device includes master control borad and interface board, and master control borad is mainly in the control plane running, and interface board mainly operates in service layer.Member device links to each other with the cascade data port that makes up the cascade data link with neighbours' member device except comprising, and described master control borad comprises that at least one is used for linking to each other to make up the cascade control port of cascade control link with neighbours' member device.Generally, master control borad can comprise two such cascade control ports, can support simultaneously whole virtual unit annular networking or chain networking like this.Through after such connecting arrangement, the cascade control link between each member device can provide an independently IRF cascade control channel, and each cascade data link then can provide an independently IRF cascade data passage.
Please refer to Fig. 3 and Fig. 4; under above-mentioned hardware environment; be embodied as example with software; the invention provides a kind of virtual device management device; it operates on the master control borad of member device in the virtual unit; this master control borad except aforementioned management port also, usually also comprise CPU, internal memory, nonvolatile memory and other necessary hardware.This device comprises at logic level: neighbours' interactive unit, topological initialization unit, Fisrt fault perception unit, the second fault perception unit, the first control unit and the second control unit.In this device running, it can relate to following handling process usually.
Step 101, neighbours' interactive unit are respectively by cascade control channel and cascade data passage and the mutual neighbours' handshake message of neighbours' member device;
Step 102, the topology initialization unit is initialized the time, from neighbours' handshake message that the cascade data link receives, obtain the first initial topology structured data, from neighbours' handshake message that the cascade control link receives, obtain the second initial topology structured data; And in the first initial topology structure and the second initial topology structure not simultaneously to manager's alarm;
Step 103, the second fault perception unit is revised as enabled state with the first control unit from enabled state when cascade control channel perceives fault, and the second control unit is enabled to substitute the first control unit from going enabled state to be revised as;
Step 104, it is mutual that the second control unit carries out IRF control by cascade data passage and other member devices;
Step 105, it is mutual that the first control unit carries out IRF control by cascade control channel and other member devices;
Step 106, Fisrt fault perception unit sends the division notice to the first control unit when data channel perceives IRF fault and IRF and is split into a plurality of new IRF;
Step 107, if this member device becomes the Master equipment of new IRF, the first control unit is according to the division notice of fault perception unit, by the mutual separately Active control information of affiliated IRF of the Master equipment of cascade control channel and other new IRF;
Step 108, the first control unit is determined the state of this IRF according to default Active negotiation rules after receiving from the Active control information of other IRF, if determine that the state of this IRF is the Discovery state, then the Business Processing of each member device among this IRF gone to enable;
Step 109, when Fisrt fault perception unit when the cascade data passage perceives the fault recovery that causes IRF division, send division to the first control unit and recover notice;
Step 110, if this member device is the Master equipment of Active IRF, the first control unit is after receiving that described division recovers notice, send the recovering state notice by cascade control channel to each member device among the IRF that is in the Discovery state, to indicate those member device respective enable Business Processing.
Step 111, if this member device is the Discovery state, the first control unit enables the Business Processing of this member device after receiving that described division recovers notice.
The present invention introduces independently cascade control channel on the basis of the original cascade data passage of IRF; The great convenience that it brings alternately for the internal control of IRF.The at first introducing of this cascade control link (hereinafter to be referred as cascade control channel), make the related software exploitation of IRF agreement more convenient, the developer can separate cascade control channel with service channel clearly, the too many unexpected factor that need to consider when having avoided control and mix of traffic.Secondly, owing to introduced independently cascade control channel, the bandwidth pressure of cascade data passage significantly alleviates; Again, also there is control that the demand of interaction data/instruction between member device is also arranged owing to have the application of more high levels on the IRF member device, use such as safety or quality of service guarantee contour level, the cascade control channel of IRF fully can be simultaneously by these use utilize, so can further alleviate again the pressure of cascade data passage.Certainly, it should be noted that the application has very large lifting in the reasonability of improving IRF division treatment mechanism, is described in detail below in conjunction with accompanying drawing most.
In preferred mode, put into operation more reliably in order to guarantee IRF, under initial condition, member device can carry out alternately with other IRF by neighbours' handshake message (IRF hello packet), thereby collects the topology data of whole IRF.Can continue to use this collection process in the present invention, but different is, the topology initialization unit can be from respectively obtaining a topology data from cascade control channel and cascade data passage respectively, then two parts of topology data are compared, if two topological structures are consistent, illustrate that then cascade control channel and cascade data passage all are complete trouble-free, whole IRF is healthy and strong.But if there is inconsistent, then illustrating has cascade control link or cascade data link to break down (such as still not carrying out physical connection) among the IRF, and can it solve as early as possible faulty link or the malfunctioning node of existence this moment to manager's alarm prompt.
If IRF cascade control channel is consistent with IRF cascade data passage, this moment, IRF can continue follow-up normal operation.In operation, in case the situation of cascade control channel fault occurs, break down such as certain cascade control link, this moment, the second fault perception unit perceived fault, such as N continuous in the time cycle not from cascade control channel receive that neighbours' handshake message just can be determined and neighbours' member device between link disconnect, perhaps receive linkdown notification of other member devices reports etc.After perceiving fault, the second fault perception unit can enable the second control unit.The second control unit can come to control alternately by the cascade data passage in the conventional mode.The at this moment control of IRF inside has been got back to alternately traditional mode and has been suffered.
The second fault perception unit enable the control interaction mechanism that controlling mechanism can allow the present invention to take full advantage of to have existed.From the another one aspect, the cascade data passage can be considered as a backup of cascade control channel.Only this backup is not identical with traditional backup, when this is embodied in cascade control channel and breaks down, is not only the switching of passage, also relates to the switching of control unit.Use the cascade data passage control mutual in, the present invention can enable the second control unit.In the present invention, there is difference in the treatment mechanism of the first control unit and the second control unit, wherein the most just comprises for the processing of IRF division and the processing of IRF division recovery, this part follow-up will more description.Can find out that from step 103 mutual employed passage and treatment mechanism can be switched simultaneously.Same reason is if after the fault recovery of cascade control channel gets up, can carry out opposite switching; The first control unit will be enabled again, then uses cascade control channel to control alternately, and the second control unit is then gone to enable again.
In cascade control channel under normal circumstances, the control of whole IRF can be carried out by the first control unit of each member device alternately.Whole IRF internal control can relate to a lot of aspects alternately, some routines control alternately can be with reference to existing techniques in realizing.Typical advantage of the present invention is embodied in the processing of IRF division, in case the cascade data passage breaks down, cause IRF to be split into two new IRF such as one or more cascade data link fracture, this moment, Fisrt fault perception unit can perceive such fault, and the mechanism of fault perception can be realized with reference to example or the various prior art of before this second fault perception.Fisrt fault perception meeting sends the division notice to the first control unit that is in enabled state.Certainly, if current what be in enabled state is the second control unit (being that cascade control channel has also broken down), then the second control unit can be notified in Fisrt fault perception unit, the second control unit can adopt traditional mode to carry out respective handling, such as the various MAD mechanism of mentioning in the background technology.
Please refer to the example of Fig. 5, suppose that IRF comprises 4 member devices (D1 to D4), four cascade control links (CL1 to CL4) and four cascade data links (DL1 to DL4).Suppose that sometime DL1 and DL3 disconnect, this moment, IRF can divide two new IRF, and each member device can re-elect Master.Suppose that the IRF that D1 and D4 form is an IRF, suppose that the IRF that D2 and D3 form is the 2nd IRF.Because former IRF is being to be a network equipment for other equipment in the network, it only has an external mutual identity (such as IP address) usually, this moment is because two IRF have appearred in division, if participating in, two IRF carry out data message forwarding in the network, then not only the identity conflict can be occurred, more whole network failure can be caused.So in fact, the IRF that only withs a hook at the end substitutes former IRF as Active IRF and continue to provide service in network.
Suppose that D1 and D2 are respectively the Master equipment of an IRF and the 2nd IRF, can carry out at this moment the mutual of Active control information (validity control information) between the first control unit of D1 and D2, then negotiate which IRF according to the Active negotiation rules and keep the Active state, which IRF need to be placed in the Discovery(forbidding) state.In preferred mode, Active control information and corresponding Active negotiation rules can be defined as required by the developer.Below providing some examples describes.Suppose that the Active control information is outlet bandwidth, negotiation rules is that the IRF of outlet bandwidth maximum is the Active state, and other IRF are the Discovery state.If there are a plurality of outlet bandwidths identical, can select Master device mac address minimum as Active IRF.
Above example is to determine that according to the ability parameter of a key which IRF as Active IRF, in fact can also introduce more ability parameter, such as forwarding chip ability, list item specification and cpu performance etc.For different ability parameters, can define weight separately, the mode by weighting calculates capacity of water, ability large preferentially as Active IRF.Except ability parameter, can also select to introduce the service range parameter, in addition the application function set, such as have the preferential of safe handling function, such as the more Internet users' of service is preferential.As can be seen from the above description, the developer can define very reasonably Active control information and the Active negotiation rules reasonably selects a new IRF as Active IRF, and such flexibility is that various MAD technology are difficult to realize in the prior art.Before fault restoration, the IRF that is in the Discovery state can stop to carry out Business Processing.The first control unit that is to say the Master equipment of this IRF can be cut off service processing function, is that the service port that all are external all is set to the Down state such as the simplest mode.
After breaking down, the manager can receive the alarm from equipment usually, then orients guilty culprit by various means, and then carries out fault recovery.Still with the above-mentioned example that is exemplified as, suppose that the fault person of being managed of DL1 and DL3 has repaired, this moment, the second fault perception unit can perceive such variation equally, and it can send division to the first control unit that is in enabled state and recover notice.If what enabled this moment certainly is the second control unit, then it can corresponding second control unit of notifying.Suppose this moment the first control unit enable, that is to say cascade control channel be do not have out of order.Because fault is recovered, the mutual neighbours' handshake message of the DL1 that neighbours' interactive unit just can be passed through and DL3, for D2 and D3, it can join an IRF as member device.After adding was finished, after namely topology settled out, the first control unit can be notified other member devices that are in the Discovery state (the namely member device among the 2nd IRF) cancellation Discovery state.The first control unit of receiving the member device of this notice can correspondingly all be set to the UP state with service port.Can find out that from the description of this section owing to introduced cascade control link and corresponding treatment mechanism, division Restoration Mechanism of the present invention seems more succinct compared to existing technologies.
The above only is preferred embodiment of the present invention, and is in order to limit the present invention, within the spirit and principles in the present invention not all, any modification of making, is equal to replacement, improvement etc., all should be included within the scope of protection of the invention.