CN104917628A - Automatic diagnosis method of Ethernet router/switch packet loss fault - Google Patents

Abstract

The invention discloses an automatic diagnosis method of an Ethernet router/switch packet loss fault. The method comprises performing timing message samplingfirstly, extracting register information of a router/switch, performing software simulation of a router/switch hardware chip forwarding process, and counting a message forwarding simulation result into statistical information. When a fault occurs, fault maintenance personnel output the software simulation statistical information out, a packet loss reason is searched in a packet loss count table according to a message type of a business fault, and then a fault reason can be found. The method can rapidly diagnose Ethernet forwarding packet loss faults, and is especially and particularly effective for a small-repetition-probability and bad-influence fault. A dotting positioning version does not need to be written, and positioning version production time is reduced; and software can continuously work for a long time, manual intervention attendance does not need, manpower investment is greatly reduced, and fault positioning efficiency is improved.

Description

A kind of ethernet router/switch packet loss automatic fault diagnosis method

Technical field

The present invention relates to the method for the automatic collection of ethernet router/switch forwarding data packet loss fault, analysis and statistics, belong to ethernet router/switch field.

Background technology

Ethernet switch and router are very complicated systems, bear the function of bottom transmission data in actual applications.In the transfer of data of reality, message packet loss fault is the phenomenon of objective reality.Some message packet loss is the normal phenomenon of transmission, and this kind of message does not form impact to upper-layer service; The transmission of some message abandons has considerable influence for upper-layer service, and fault shows as the performance bottleneck of business, burr, time extend, even interrupt between business.When upper-layer service goes wrong, usually require that route switch module attendant works in coordination with investigation problem.Especially, when the business of operator breaks down, how fast and accurately localizing faults reason becomes a urgent realistic problem.

Class fault location is forwarded for ethernet router/switch, the method of tradition investigation problem is artificial investigation, the first approach of analytical system networking and business datum forwarding theoretically, then the software and hardware information that specific aim collects each correlation module comprises statistical information, register information, software state machine etc., then the manual analysis link that may go wrong, then the analysis verification that progressively reduces the scope.Due to the system complex come into operation, network node is numerous, and the process of this investigation orientation problem often wastes time and energy, and efficiency is extremely low.

In order to reduce the workload of Breakdown Maintenance, release manpower, improves fault location efficiency, the present invention proposes a kind of based on the emulation of chip hardware Framework Software, and software acquisition message descriptor, emulation message is in the handling process of hardware chip.When emulating Message processing flow process, do not consider chip how hardware implementing, only realize chip according to black box principle and forward result, namely forward principle according to chip, the processing links of specifying, arranges the hardware condition of forwarding according to current hardware register information, Ethernet message should do forwarding or discard processing or give next link process under this condition, corresponding result stamped by corresponding message descriptor software, and associated counter does accumulation operations, gives next link to emulate.Said process carries out iteration repeatedly, until the final simulation result of message out.

Chip software emulation mode is based on the hardware chip forwarding plane of router/switch, according to the forward process flow process of chip register information emulator message in hardware chip of Real-time Collection, and the result of emulation is counted statistical information.Information due to hardware register is real-time, the message descriptor obtained also is real-time, hardware chip forwarding process is fixing, with software, message descriptor is substituted into chip and forward register and forwarding process emulation, be equal to chip oneself the hardware effect that E-Packets, the message emulating result drawn is true and reliable.The statistical counting of relevant treatment link help system attendant can directly judge the ruuning situation of current system fast, and when system has packet loss fault, associated count information directly can point out register and the packet loss reason of packet loss.

Conventional failure location needs to judge traffic affecting message kind, manually gathers register information, in conjunction with the hardware structure that chip forwards, and artificial judgment packet loss reason.When Rule of judgment is unintelligible, making is usually needed to get version assist location problem ready, the large and inefficiency of workload.Compare conventional mapping methods, software emulation of the present invention not only accurate positioning, and convenient and swift.Because software processing cycle is very fast, can iterate, except for except conventional data acquisition analysis, when a failure occurs, during orientation analysis class message packet loss reason, also can by arranging filter condition, Fast Convergent orientation range.Particularly little for reproduction probability and the fault that makes a very bad impression, the present invention does not need again to write and gets location version ready, reduces location version Production Time; Simultaneously software can continuous firing for a long time, does not need manpower intervention on duty, greatly reduction human input improve fault location efficiency.

When external manifestation of the present invention is fault location, a key collection also analyzes fault packet loss reason, and fault automatic location efficiency is very high.

Summary of the invention

technical problem:the technical problem to be solved in the present invention utilizes software Real-time Obtaining message descriptor and by this message of software emulation handling process in the chips, result is counted statistics, substitute by the mode of software process the reason that packet loss fault appears in manual analysis message repeating, significantly can reduce the workload of attendant and improve fault location efficiency.

technical scheme:in order to solve the problems of the technologies described above, the invention provides descriptor, Real-time Collection chip forwarding register information that a kind of dynamic acquisition E-Packets, software makes forwarding according to chip forwarding process emulation message in the processing procedure of chip and determines, and count statistical information according to result, comprise the following steps:

1, the descriptor of message is extracted

Software provides and arranges packet sampling frequency interface, and system is sampled according to the message of sample frequency to each forwarding port arranged.The setting principle of sample frequency is that the speed of sampling can not impact the performance of CPU, can not affect the process of CPU regular traffic simultaneously.According to the CPU CODE of message up sending, CPU identifies which message is the message of sampling, which message is the message normally giving CPU process.Message of wherein sampling delivers to dedicated cache district buffer memory, can not walk the passage of CPU process regular traffic message, prevents the process of sampling message interference operation layer and protocol layer.

The message of sampling, at each forwarding port up-sampling message, is given CPU and is described symbol extraction by software timing.After CPU receives sampling message, first by source mac, object mac, VLAN, protocol number, CRC check and extraction descriptor information.Then specify that the field information used extracts related data by this agreement further according to protocol type, such as: message protocol number is 0x0800, represent that message is IP type, extraction source IP, object IP, TTL, packet header School Affairs, further extract the field informations such as corresponding TCP, UDP according to the sub-protocol number of ip; Such as message protocol type is 0x8847-0x8848, and expression is MPLS type, extracts label information etc.Each field information classification of extracting is write in the descriptor of message.The principle extracted is that each protocol fields of the Ethernet message that message is related to all extracts, except calculating message checking and except, can be indifferent to the data content that message carries.

In addition, software also needs to provide packet sampling filter interface, arranges rear CPU and only to sample the message of specified type, for very effective during certain Class Type message packet loss failure diagnosis.

2, CPU extracts switch/router hardware register information

Each forwarding register information of the whole chip of CPU timing acquisition, due to the hardware chip often more complicated that switch/router uses, register is of a great variety, list item is very large, if all extract and be used for software emulation, although accuracy is very high, larger pressure can be caused to the CPU overhead of Installed System Memory and sample register.Accuracy for the performance of system, internal memory and emulation considers, the critical registers used of each forwarding process of generally only sampling.

The register information gathered includes but are not limited to the related register information such as VLAN, MPLS/MinM/TRILL, L2, L3, ACL, TRUNK, INGRESS MIRROR, MMU, L3 TUNNEL, VLAN/STP, EGRESS MIRROR, PORT.

3, software emulation message is in the forwarding process of hardware chip

1) descriptor of message is substituted into the forwarding process of hardware chip;

2) according to the concrete data simulation message of the descriptor of the message processing links through router/switch forwarding chip;

3) in the router/switch chip processing links of emulation message process, according to the hardware register information obtained, the result of message under Current hardware state is judged above, and by result statistical counting;

4) according to 3) result, message is delivered to next link and processes;

5) iterate 2), 3), 4) until message is forwarded away or discard processing;

4, the forward process information of message is counted in statistical form

1) set up respectively and forward according to the type of message, length, protocol type, the port numbers entered, the register type etc. that forwards link and abandon two kinds of counters;

2) the duration unit arranging statistics is provided, and the interface of the type of statistics, the personnel of maintaining easily read statistics;

3) log recording function is provided, by relevant statistical information write daily record, for historical analysis;

4) online query statistic information interface is provided, the forwarding statistical information that online query is current.

beneficial effect:the present invention compared with prior art, has the following advantages:

A) can by arranging filter condition, Fast Convergent orientation range.

B) do not need again to write and get location version ready, reduce location version Production Time;

C) software can continuous firing for a long time, does not need manpower intervention, greatly reduces human input and improve fault location efficiency

Accompanying drawing explanation

Fig. 1 is the software frame figure based on chip emulation automatic diagnosis ethernet router/switch packet loss fault.

Embodiment

Be described in further detail below in conjunction with the enforcement of accompanying drawing to the Ethernet message packet loss reason technical scheme that the present invention relates to:

The first step, arranges chip timing sampling message function, gives CPU, arrange the priority of packet sampling by each entering of forwarding port to message, requires that the speed of sampling can not impact the performance of CPU, can not affect the process of CPU regular traffic simultaneously.According to the CPU CODE of message up sending, CPU identifies which message is the message of sampling, which message is the message giving CPU process.Message of wherein sampling delivers to dedicated cache district buffer memory, can not walk the passage of CPU process regular traffic message, prevents the process of sampling message interference operation layer and protocol layer.

Software creation thread, resolves the message of buffer area, extracts the descriptor of message, as: the ingress port information etc. of object MAC, source MAC, vlan number, protocol number, CRC check, message.Extract concrete field according to protocol number, such as: message is 0x0800, represent that message is IP type, extraction source IP, object IP, further extract corresponding TCP, UDP field information; Such as message protocol type is 0x8847-0x8848, and expression is MPLS type, extracts label information etc.

Second step, software extracts forwarding chip register information or software processing links information, and the information of collection includes but are not limited to the related register information such as VLAN, MPLS/MinM/TRILL, L2, L3, ACL, TRUNK, INGRESS MIRROR, MMU, L3 TUNNEL, VLAN/STP, EGRESS MIRROR, PORT.

3rd step, message descriptor is substituted into chip forwarding process by software, according to the register information of each Real-time Obtaining judge message chip forwarding flow to.Because be by software emulation chip hardware forwarding process, so the ardware model number of software processes and chip and the handling process of chip series are strong correlations.Software is in concrete processing procedure, then the train type concrete according to chip carries out the segmentation of switching value.

This emulation is the principle based on Black-box Testing, switching value is set, eligible, pass through, ineligible, abandon, then count according to correlated results, this is a simple deterministic process, because hardware realizes forwarding capability, not needing software again to realize concrete forwarding capability, only needing black box to judge whether, by forwarding switch and statistics, therefore to implement simple and fast.Because software carries out message repeating judgement in strict accordance with the forwarding logic of chip, and the register information judged is Real-time Obtaining, and the data retransmission result therefore obtained also is true and reliable.

Below for common IP message repeating flow process, illustrate software how to substitute into chip forwarding process with descriptor; Other types message repeating emulation mode roughly the same, no longer describes in detail.

(1) first chip checks that whether message source MAC is legal, if source MAC is multicast or broadcast MAC, then abandons; If source MAC is set to DISCARD in L2, then abandon;

(2) secondly according to the object MAC+VLAN of message, next step forwarding process is judged; If object MAC is broadcast MAC, then broadcast in VLAN, now will check entrance broadcast speed limit mark, stamp the statistics that may abandon according to label information;

If object MAC is multicast MAC, then looks into multicast table L3 switch and whether open, if open away L3 Multicast Routing, now judge L3 interface table and L3 routing table information, if do not mated, then abandon; Coupling, then do IP verification, TTL subtracts 1, the operations such as source object MAC replacement, the value according to IP check results and TTL judges whether to abandon; Otherwise look into L2 multicast table, known multicast forwarding is walked in hit, miss walk unknown multicast forward (now needing to check that unknown multicast speed limit mark and VLAN forbid that unknown multicast forwards and mark), abandon/the statistics that may abandon or forward according to mark value;

If object MAC is clean culture MAC, then look into L2 table, if hit, check and answer the whether set of list item L3 field, set is walked L3 and is forwarded (handling process is with L3 multicast); Non-set is walked L2 known unicast and is forwarded, and is corresponding INGRESS L2 PASS and adds up; Otherwise walk L2 unknown unicast and forward (now check L2 unknown unicast whether speed limit, forward/may the statistics of speed limit).

(3) message port enters to MTU and stp state inspection, looks at whether message exceedes inbound port MTU, exceedes and does discard processing; If port STP is BLOCK, message object MAC be not bpdu type and acl rule do not arrange allow such message come in, then do discard processing.

(4) before message enters EGRESS passage, check the current configuration of MMU and take resource information, if when MMU resource occurs not enough, be MMU and abandon statistical counting.

(5) after message enters EGRSS passage, check that destination interface is common port or TRUNK port, if TRUNK port, according to after TRUNK HASH rule computation index, corresponding outbound port is found out; Will check VLAN STP whether BLOCK in addition, if BLOCK, and the non-bpdu type message of message does discard processing; If STP arranges FORWARDING, then do forward process; Check outbound port MTU configuration, if message length exceedes outbound port MTU, then message does discard processing.Then forwarding process emulation is terminated after relevant information being counted.

4th step, software, according to each message descriptor information, does detail statistics to the sample information that chip forwards.Such as divide broadcast, multicast, clean culture according to message MAC; Divide in detail according to message length, and statistic of classification.According to descriptor summary, each processing links packet loss of statistic of classification, possibility packet loss, counting of letting pass, and finally forward decision counting.When a failure occurs, software emulation statistical information exports out by Breakdown Maintenance personnel, and the type of message according to traffic failure searches packet loss reason in packet loss count table, can find failure cause.If failure cause can't be determined completely, can by arranging the filtering rule of software emulation message, specific aim sampling service message carries out software Packet Loss Simulation, directly orients packet loss concrete reason.Can avoid like this changing in office side operations such as getting version ready, greatly reduce difficulty and the workload of later stage Breakdown Maintenance, improve fault location efficiency.

Claims (4)

1. ethernet router/switch packet loss automatic fault diagnosis method, it is characterized in that, the method includes the steps of:

1) cpu extracts the descriptor of message;

2) the forward process result of message in hardware chip is emulated;

3) the forwarding result of the descriptor type of message, each link is carried out statistic of classification, and journalize information.

2. method according to claim 1, is characterized in that, in described step 1), extracts message descriptor in accordance with the following steps:

Timing sampling message, gives system CPU by this sampling message copy portion; What provide sample frequency arranges interface; Resampling filter interface is provided, accurately filters the message of specified type.

3. method according to claim 1, is characterized in that, step 2) in, the idiographic flow that emulation message forwards at hardware chip is:

1) descriptor of message is substituted into the forwarding process of hardware chip;

2) according to the concrete data simulation message of the descriptor of the message processing links through router/switch forwarding chip;

3) in the router/switch chip processing links of emulation message process, the hardware register information required for this link of Real-time Obtaining, judges the result of message under Current hardware state, and by result statistical counting;

4) according to 3) result, message is delivered to next link and processes;

5) iterate 2), 3), 4) until message is simulated forwards or discard processing.

4. method according to claim 1, is characterized in that, in step 3), the flow process of statistic of classification message repeating result is:

1) set up respectively and forward according to the type of message, length, protocol type, the port numbers entered, the register type etc. that forwards link and abandon two kinds of counters;

2) the duration unit arranging statistics is provided, and the interface of the type of statistics, the personnel of maintaining easily read statistics;

3) log recording function is provided, for historical analysis;

4) online query statistic information interface is provided, the forwarding statistical information that online query is current.