CN101083567A - Performance monitoring of frame transmission in data network oam protocols - Google Patents

Abstract

Performance measurements for a network, such as delay, delay variation, and loss, are performed over layer-2. A PM frame is sent from a source MEP (Maintenance End Point) to a destination MEP. The PM frame is timestamped immediately prior to transmission, and the local clock at the destination MEP is noted immediately upon arrival. MIPs (Maintenance Intermediate Points) may also time stamp the PM frame. Preferably, each PM frame has an interval (session) identifier and a sequence number within the interval. The PM frames are received by and sent by PM modules within the nodes of the network. The PM modules include a Sender function for sending an interval of one or more PM frames and a Receiver function for receiving PM frames and computing measurements.

Description

In data network operation and maintenance agreement to performance monitoring of frame transmission

The cross reference of related application

The application advocates the common unsettled U.S. Provisional Application No.60/691 about " in data network OAM (operation and maintenance) agreement to performance monitoring of frame transmission (PERFORMANCEMONITORING OF FRAME TRANSMISSION IN DATA NETWORKOAM PROTOCOLS) " by name submitted on June 17th, 2005, the priority of 493 the applying date.

Technical field

Relate generally to data communication of the present invention, and relate more particularly in data network performance monitoring of frame transmission.

Background technology

The demand of carrying out network performance monitoring in data network is increased.Up to date, this network performance monitoring is mainly used in network failure and gets rid of (network management, state and warning) except that being used for conventional continuity testing method.At present, inspection, charging and the network security of SLA (service-level agreement) (detection that the type that floods is attacked etc.) also need network performance monitoring.Therefore, the status of performance monitoring just becomes necessary and essential behavior from the behavior based on interest.

Although have the 3rd layer of limited accuracy and more the network performance measurement on upper strata is former is enough, yet need higher accuracy now.Aspect network management and agreement, next generation network needs higher accuracy and stronger comprehensive.Postpone, delay distortion (" shake ") and Loss Rate are three key metric parameter (metrics) of assessing the quality that voice and video stream for example etc. mainly serves.These metric parameter must can be measured by operator and provider, thereby examine in fact they provide SLA from agreement to its client, and can detect and make fast reaction when these metric parameter are brought down below the agreement rank.

Generally speaking, performance monitoring can realize (that is, by observing actual customer service and may adding some specific markers information) passively or realize (that is the extra control business that, is used for the performance measurement purpose by injection) on one's own initiative.Common preferred initiatively solution, reason is that it disposes manyly easily, although because injected extra business in the identical network of its measurement, therefore accuracy is lower in itself for the active solution.Passive solution is more accurate, but needs more complicated architecture, needs basically being correlated with in the grouping of entrance and exit place observation.

A kind of form of passive performance monitoring solution comprises the local counter in the Maintenance Point (information trap (information trap)).RMON (far-end monitoring) is based on the example of the solution (at the IETF RFC2819 (internet engineering task group draft 2819) of RMON-1, at the IETF RFC2021 of RMON-2) of counter.With storage in MIB (management information bank), and usually by SNMP (Simple Network Management Protocol) swap data.Data counter records such as byte (octet) number, frame number, number of bit errors, number of collisions, type (broadcasting, multicast), big keg (size bucket) (frame between 64 bytes and 127 bytes etc.) or the more customizing messages, these customizing messages for example by the terminal node of its MAC (media interviews control) address designation to the byte count between (pair of endnodes) (counter of " matrix " type).Based on threshold value or filter that the user determines, can produce alarm by the RMON detector: for example, if counter exceeds certain threshold value, then detector will produce alarm.Filter can be caught the frame that is complementary with certain filter (for example be used for later stage playback (post-replay), or be logout).

The problem of usage counter is the basis that local counter is not provided for measuring delay or delay distortion.In Loss Rate is measured, need between sending node and receiving node, exchange counter information, so that be correlated with between the frame that observes, this has increased one deck complexity for measuring.

Most of current performance monitoring systems are gone up and are realized the 3rd layer (IP (Internet protocol) level).IPPM (IP performance metric parameter) is responsible for the definition metric parameter and the IETF working group that realizes guidance is provided.The theoretical foundation of the statistical significance that is used to measure is provided especially.IPPM has issued one group of RFC (draft): 2330,2678,2679,2680,2681,3148,3357,3393,3442, the unidirectional agreement of initiatively measuring) and new metric parameter (in many ways and reorder) and drafting signaling protocol (OWAMP:.ITU (International Telecommunication Association) has issued ITU-T-International Telecommunication Association TC Telecommunication Committee) Y.1540 and Y.1541 file.Exist a large amount of Software tools that is used for IP level performance measurement ( Www.caida.org, Www.ripe.net/test-traffic/), and some providers measure the customer service counter commonly used of the error rate, bandwidth availability ratio etc. except being provided for, and the software or the hardware that are used for the performance monitoring of IP level of certain form also is provided.U.S. Patent No. 6,868,094 and No.6,868,068 have described IP level measuring system.

Provide statistical significance preferably from the method for the IPPM of IETF (analysis of measurement errors of calibration is adopted in randomization), but regrettably its layer that is suitable for has reduced the measurement quality.Therefore in the IP level, the stream of packets from the source to the destination can be used different paths, even under all identical situation of other conditions, the measured value of acquisition is also inequality.And general implementation rule is, measurement module off-line road interface is far away more, and the accuracy of mark time stamp (timestamping) is poor more.In addition, the method for measuring grouping must be configured respectively, because it is not a part that is used for the standard OAM infrastructure of IP: this is a problem for operator, operator's preference simplicity and automation (direct integrated performance monitoring in the OAM framework).In addition, IPPM need create session for measurement performance.

A series of in addition solutions comprise the measuring equipment (use is with the independent box of several modules, or third party's plate/module of the compatibility of the slot in the use manufacturer node, even mobile hand-held device) that adds stack to existing network.This optional equipment can comprise other functions, typical protocols having compliance test, physical signalling quality inspection (electricity or light) etc.Use the performance monitoring of superimpose device to carry out on random layer, these layers comprise application layer (for example, generating the performance of false call with monitoring VoIP (IP-based voice)).

Yet the performance measurement that uses superimpose device can increase the cost of monitoring network widely.

Therefore, need be used for data network carry out performance monitoring accurately, cost effective method and equipment.

Summary of the invention

In the present invention, by the performance monitoring frame execution performance monitoring in network that generates the second layer (layer-2) by first node, wherein to have frame identification be one or more type fields of performance monitoring frame to the performance monitoring frame, specify the initiation time stamp field of the time of transmit frame, and wherein one or more features of performance monitoring frame change.In predefined interval, send the performance monitoring frame by network.By the Section Point receptivity in network monitoring frame, and the arrival time stamp that is associated with the reception of performance monitoring frame of storage.Initiation time stamp and arrival time stamp according to the performance monitoring frame are determined performance index.

Description of drawings

For a more complete understanding of the present invention and advantage, with reference now to the description of carrying out below in conjunction with accompanying drawing, wherein:

Fig. 1 shows the diagrammatic sketch of simplifying very much that can realize network environment of the present invention;

Fig. 2 shows the general diagrammatic sketch of the preferred structure that is used for PM (performance measurement) frame;

Fig. 3 shows the block diagram of PM module;

Fig. 4 shows maintenance association body (Maintenance Association);

Fig. 5 shows the state diagram of the transmitter of PM module;

Fig. 6 show two of transmit frame measure at interval at random regularly and the example of length;

Fig. 7 shows the state diagram of the receiver of PM module;

Fig. 8 shows the calculating of delay distortion.

Embodiment

Fig. 1-Fig. 8 in conjunction with the accompanying drawings can understand the present invention best, and similar label is used for the similar units of each accompanying drawing.

Fig. 1 shows the diagrammatic sketch of simplifying very much that can realize environment of the present invention.Provider's network 10 connects one or more position of customer network.In an illustrated embodiment, customer network is depicted as has 4 positions (being shown as position A, B, C and D), each position has its oneself local area network (LAN) 12.Fringe node 14 provides between provider's node 16 of local area network (LAN) 12 and provider network 10 and is connected.Therefore, provider's network 10 connects various local area network (LAN)s 12, to form wide area network.

In reality realized, provider's network 10 can be supported a plurality of clients, and it can use other third party provider network so that essential connectedness to be provided.

At present, aforesaid situation provides SLA (service-level agreement), and SLA has stipulated the performance class of provider's network.Can comprise the index that is used for throughput, maximum delay, delay distortion and data loss rate among the SLA.This index may depend on the use of wide area network, and for example, network can move the application that is used for VOIP (IP-based voice) phone and video conference.In this case, delay, delay distortion (being sometimes referred to as " shake ") and Loss Rate may compare key.

The preferred embodiments of the present invention are used for the specific OAM frame (" PM frame ") of the second layer (data link layer) performance monitoring.For the efficient and the simplicity of network management, the PM frame is integrated among the OAM (operation and maintenance).Preferably this PM frame is appended on other OAM frames that develop at present in 802.1ag series draft, these other OAM frames are CC (connectivity checks) frame, LB (loopback) frame, LT (link trace) frame and AIS (AIS) frame particularly.Although the part of PM frame as Ethernet second layer environment can be described, the PM frame also can be used in other the second layer environment such as MPLS (multiprotocol label switching).

Figure 2 illustrates the general diagrammatic sketch of the preferred structure that is used for PM frame 20.Except that standard header information 22, PM frame 20 comprises type field 24, time stamp field 26 and id field 28 and fills field 30.If suitable, these fields can be arranged in the header or the payload of frame.

Type field 24 comprises and is used to differentiate the mark that PM frame 20 is OAM frames of PM type.Type field 24 can comprise those supervisory levels (eight ranks of from 0 to 7 are used for client, provider and operator) that for example define in 802.1ag.In existing C FM (Connectivity Fault Management) 802.1ag, these two clauses and subclauses are called " CFM " type and " ME (maintenance entity) rank ".In addition, in this field, can increase version entry, to be used for expansion in the future.

The one or more time stamps of time stamp field store.For one-way delay measurements, it is essential having only a time stamp, and this is preferred.A plurality of time stamps are supported.Time quantum should be represented microsecond, and should make it can expand to expression nanosecond, thereby is keeping available in the future.At present, the ethernet frame of going up the minimum that sends at less LAN (short distance) with higher transmission speed (near the light velocity) and higher interface rate (10Gb/s) will experience the delay less than 1 microsecond.For the ease of using, time format can be followed for example existing standard NTP (NTP (Network Time Protocol)), and NTP uses 64 bits (be divided into two parts: 32 bits are used for second, and 32 bits are used for the mark (fraction of a second) of second).

Id field 26 comprises separation dot and sequence number (or monotonically increasing identification schemes) arbitrarily, distinguishes so that can carry out frame by processing capacity, and can carry out relevant between transmitter and receiver.Hereinafter will be described in more detail this.

Fill field 30 and make the size of frame 20 to change, to simulate actual Frame.

In an illustrated embodiment, there is not specific field to be used for classification in the PM frame 20.Be expected in the Ethernet, classification is with measured VLAN mark (C-mark, S-mark), based on the User Priority bit and/or based on loss priority bit (drop precedence bit).

Explain below in conjunction with the operation of PM module 31 and the use of PM frame in Fig. 3, to illustrate in greater detail the operation of PM module 31.The function of PM module 31 can be implemented as the routine in the bridge control centre (bridge brain) that must need any or all provider's nodes in the provider of the performance monitoring node 16.PM module 31 can also realize on client node 14 equally.On port 32, send and receive data, on port 32, form frame with these data from ground floor.Port 32 is coupled to transmitter 34 and receiver 36.Transmitter 34 and receiver 36 are coupled to management information bank (" MIB ") 38.NMS (network management system) interface 40 is coupled to NMS 42 (it is usually located in the Network Operation Centre or " NOC " of provider) with MIB 36.The order that NMS interface 40 receives from NMS 42, upgrade the operating variable that transmitter 34 and receiver 36 use, and can carry out the computing in early stage (pre-computation) such as data are classified and synthesized, with the workload of minimizing NMS, and the data volume of exporting to NMS 42 is minimized.Transmitter 34 generates the stream of randomized PM frame 20, and writes down the frame number that each sends at interval.Receiver 36 is intercepted and captured PM frame 20 from other OAM frame, and writes down the delay that measures.

In conjunction with Fig. 4 the operation of PM module 31 is described.Maintenance association body (" MA ") 48 (also is called Maintenance Entity (or " MEG ") by ITU-T, with the maintenance entity (or " ME ") of representing one group of point-to-point) be defined as the group that comprises two or more nodes 14 or 16, wherein at least two nodes comprise maintaining end point (" MEP ") 50, and can comprise zero or a plurality of Maintenance Intermediate Point (" MIP ") 52.According to type of module, be not all component all be essential for PM module 31.MIP can only realize receiver 36 and NMS interface 40, and does not have transmitter 34.On the other hand, the MEP of simplification (for example MEP that only realizes in residential gateway) can only realize the NMS interface 40 of transmitter 34 and more uncomplicated version, and does not have receiver 36.

Can use PM frame 20 to realize several different calculating.Can realize one-way latency calculating by sending one or more PM frames 20 to destination MEP 50 (spy is called MEP 50d) from source MEP 50 (spy is called MEP 50s).When source MEP 50s sends, in time stamp field 26, import time stamp.When receiving, PM frame 20 is identified as the maintenance frame of type specified in type field 24, and sends it to receiver 36 at MEP 50d place, destination.By from the current time of the clock that is associated with destination MEP 50d, deducting the time in the time stamp field 26, calculate delay.

Can carry out several distortion to this measurement.By sending a plurality of PM frames (preferably send, hereinafter will be described in more detail), can calculate average one-way delay measurements value to it with randomized interval, randomized loading.

In addition, can also be PM frame flag time stamp on each MIP 52 in the path.By the time stamp from each MIP 52 is provided, can be at each branch's computing relay of transmission.For more accurate, transmitter 34 and receiver 36 can all provide time stamp, deduct this time difference from postpone.

Can realize two-way delay measurements by sending one or more PM frames 20 to destination MEP 50d from source MEP 50s, wherein by the receiver 36 of destination MEP when receiving PM frame 20 to its mark time stamp, and by transmitter 34 when source MEP 50s beams back PM frame 20 to its mark time stamp.For more accurate, from latency measurement, deduct the time difference between destination transmitter 34 and the receiver 36.Should be noted that the feasible precise synchronization that no longer needs between source MEP 50 and the destination MEP 50 of two-way delay measurements.

Can carry out the calculating of one-way latency deviation by between source MEP 50s and destination MEP 50d, sending two or more PM frames 20.The delay difference of two PM frames from the source to the destination is defined as delay distortion.This delay distortion can also calculate based on the stream of PM frame, preferably with random interval, fill size at random and send the PM frame.

In the past, can be by on each MIP 52, PM frame 20 mark time stamps being come the delay distortion of each branch of calculating path.Because transmitter that can each MIP of cancellation in deviation calculation and the delay between the receiver, so on each MIP 52, only need an independent time stamp.By sending the PM frame to destination MEP 50d, and it is returned to source MEP 50s, can carry out two-way delay distortion and calculate from source MEP 50s.

Can be by carrying out Loss Rate calculating to the stream that destination MEP 50d sends the PM frame from source MEP 50s.The frame that to not receive in certain period is counted the frame of losing.As an alternative, the passive measurement of Loss Rate can be calculated from the count value of MEP 50s transmit frame and the count value of destination MEP 50d received frame by exchange.

NMS 42 is provided for some parameter of source MEP 50 and destination MEP 50.Average ETH-PM speed (bit/s) is made as makes the professional all maintenances of ETH-PM can ignore (for example, less than physics available bandwidth (on average) 0.1% or even 0.01%).Miss-threshold (second) is provided with the duration (from transmitting time), thinks that after this duration frame loses.Usually, miss-threshold (for example: 200 seconds) on the magnitude of a few minutes.The random-length of frame is calculated and is used minimum and maximum frame sign (byte).Usually, the payload magnitude range of Ethernet is 46 bytes-1500 byte (being 46 bytes-9000 bytes for jumbo frame perhaps).These parameters are stored in the public operating variable 44 of PM module 31.

The generation of PM frame 20 should be carried out randomization, and it should be ignored for the OAM business.For the frame relevant (postponing and delay distortion) with delay, should realize that the non-of certain form wilfully grow up to be a useful person at sending scheduling (typically being Poisson process), and frame length and filling also should be randomized, measure and (if measure frame is not randomized, and then node manufacturer can carry out optimization to its realization that is used to measure frame to the mistake optimization of service area to avoid wrong.)

For with lose relevant frame, between transmission PM module 31 and reception PM module 31, should there be a kind of mechanism, to reach an agreement (usually to sending scheduling, share the seed (seed) of random number generator, and the ID scheme is reached an agreement), make whether receive PM module 31 can judgment frame lose.The parameter that is associated (especially " miss-threshold ") should be configurable in NMS 42.

Preferably, storage/output function should be used existing standard (SNMP, TMN (telecommunication management network), IPFIX (output of IP flow information), CORBA (CORBA), XML (extend markup language) etc.), although under standardized interface, can use the proprietary realization of optimization.

Because needed randomization, this solution have guaranteed the statistical significance that the better second layer is measured; Because this solution is integrated among the OAM, do not need other superimpose device, so operator and provider is easier that it is managed; And this solution balance the advantage of second layer monitoring: time stamp mark (timestamping) and deterministic path more accurately.The raising of accuracy is because time stamp can more close physical layer (ground floor), has therefore observed the time departure of first bit of time of advent of last bit of frame and frame better.

In table 1, provided the more detailed description of PM frame below.Should notice that this only is an example of PM frame 20, this PM frame 20 uses the Ethernet OAM frame of current version as the basis; Suppose and to expand in the future, so that 802.1ah I-mark and B-mark and other headers can increase.

The field of table 1 PM frame

Parameter	Value	Index in MIB	At random
				DA (destination-address)	Two kinds of situations: 1) all MA are given in multicast: the multicast OAM address of reservation (having or do not have the MA rank of embedding) 2) clean culture (being determined by destination MEP)
SA (source address)	Determine by source MEP
				S-type (S-type) (if existence)	(802.1ad PB provider bridge)
S-tag (S-mark): abandon qualified bit	That be provided with or randomized	Be (1)	Possible (4)
				S-tag (S-mark): loss priority is other	That be provided with or randomized	Be (1)	Possible (4)
S-tag:S-VLAN (S-mark: the S-local area network (LAN))	Determine by source MEP
				C-type (C-type) (if existence)	(802.1Q)：0x8100
C-tag (C-mark): User Priority	That be provided with or randomized	Be (1)	Possible (4)

C-tag:C-VLAN (C-mark: the C-local area network (LAN))	Determine by source MEP
				Ethertype (EtherType)	(802.1ag?CFM)
Version (5 bit)	0 (identical) with the CFM version
				MA rank (3 bit)	Determine by source MEP
OpCode (8 bit)	ETH-PM
				MA_ID (a kind of (VPNID of the integer of character string, 2 bytes, RFC 2,685 7 bytes), every type 1 TLV in three types); At least 11 bytes	Determine by source MEP
Source MEP (9 byte TLV)	Determine by source MEP	Be (2)
				Time stamp (64 bit)	Obtain and be provided with from clock at the last moment by transmitter, before reading any other information, read by receiver
IID (32 bit)	Interval id (determining) by source MEP	Be
				SeqNum (sequence number) (32 bit)	Frame id at interval	Be (3)
Mistake (16 bit)	In proportion and multiplication, the value of I representative is 0.23ns (2 -32S), it is proposed to be similar to OWAMP
				Fill: length and value have possible security consideration (IZP)			Be

The note of table 1:

(1) DE/DP/UP (meeting the condition of abandoning, loss priority, User Priority) is used as based on the index among the MIB at interval, rather than the index among the independent frame MIB.

(2) MEP index in source is the serial connection of MA_ID and source MEP.

(3) SeqNum is used as the index among the independent frame MIB, rather than based on the index among the MIB at interval.

(4) be provided with based on NMS.In order to solve the selection of DE/DP/UP field in S-mark and C-mark, the particular value of randomization or one group of band NMS order is fine.

The most of fields that provide above are shared for the Ethernet maintenance frame.Whether Ethertype field designated frame is the Frame of maintenance frame.The OpCode field is appointed as PM frame (and if there are a plurality of PM frame types, then this field can be specified the type of PM frame) with frame.

Time stamp, IID, SeqNum, mistake and filling field are that the PM frame is distinctive.In given realization, can omit some field, the dissimilar PM frame 20 with different fields perhaps can be arranged.

The time stamp field comprises one or more time stamps that the transmitter 34 by PM module 20 provides.Being right after sending between the PM frame 20 by transmitter 34 provides time stamp, and is right after after receiving PM frame 20 by receiver 36 and reads local clock, to have as far as possible little overhead delay.

IID and SeqNum comprise id field 28.IID represents a spacer segment, i.e. the stream of Xiang Guan PM frame 20.SeqNum is each frame 20 distinctive sequence number in an interval.In the PM frame, provide IID and the SeqNum can be so that need before sending the PM frame, not set up session.For example, if destination MEP 50 begins to receive the PM frame that has new IID, then it can suppose calling new performance measurement circulation, and begins this stream is measured.

For example, the delay distortion measurement module need be selected some frame in a stream, and selection algorithm can depend on frame ID (although the most probable use is Min-to-Max (minimum to maximum) choice function, the Min-Max choice function is selected minimum delay Min and maximum delay Max in an interval, wherein the difference between Min and the Max is the delay distortion that is calculated).Another example is to lose measurement module, and it must identify the measurement frame at interval that belongs to agreement: based on frame ID limit the mistake that may take place when having avoided at interval limiting at interval with time value (mistakenly will near beginning or the frame that ends up count or omission the frame of close beginning or ending).For the metric parameter that reorders (performance monitoring to real-time video transmission is useful) in possible future, sequence number also can be essential.As for the length of ID, should use 32 bits at least, so that can be more than distinguishing between 4,000,000,000 frames in a stream.In order to prove that ID need be more than 16 bits (it can provide less than 60,000 6 thousand unique ID), should consider, because miss-threshold is a few minutes (being assumed to 200 seconds) usually, given line interface speed be 40Gb/s, minimum ethernet frame be 72 bytes (comprise preamble and verification and, cyclic redundancy check (CRC) (or " CRC ") for example, or Frame Check Sequence (or " FCS ")) under the situation, maximum frame number will be near 14,000,000,000, and wherein having only sub-fraction can be OAM-PM (owing to the demand to insignificant business).Therefore, can prove that the 16-bit is too limited.

Fill field and have variable size and content.The advantage that size provided that changes the PM frame in the interval is to measure more accurately, because actual user data frame may have any size, the movable frame of measuring of OAM that therefore has variable-size provides actual measurement.Change content, promptly the sequence of " 0 " and " 1 " provides equally more accurately and has measured, reason be some hardware optimization of for example data compression and so under the situation that the OAM frame that all has the same characteristic features data is measured, may make and measure the generation deviation.Therefore, preferred embodiment is used payload, and this payload is represented actual Frame, and more is not vulnerable to the influence of measured deviation.

Fig. 5 shows the state diagram that is used for transmitter 34.Transmitter 34 has two states: state of activation 62 or halted state 60.In halted state 60, do not send PM grouping 20.Afterwards, transmitter 34 just is transformed into state of activation 62 receiving initiation command (usually from NMS 42).In state of activation, transmitter 34 sends the stream of one or more PM groupings 20 to the node that is associated with destination MEP 50d.If measuring is bidirectional measurement, then MEP 50d in destination will beam back the PM grouping to source MEP 50s.Fig. 5 shows the false code (pseudo-code) of the operation that is used for state of activation 62.At first, determine time started and dwell time at interval, and distribute IID at interval to this.Replacement SeqNum.When arriving the time started, transmitter 34 begins to send PM frame 20.Preferably with at random interval, each frame to sending calculates predetermined transmitting time.For this frame is prepared header, IID, SeqNum etc., therefore except time stamp, this frame is got ready for sending.Be right after before sending PM frame 20 and insert time stamp.In case sent the PM frame, SeqNum just increases progressively, and if have next PM frame, then by calculating predetermined transmitting time and preparing the PM frame and prepare next PM frame.

The state diagram of Fig. 5 is at source MEP 50s.If middle MIP 52 will add time stamp, MIP 52 will only need add the time stamp (this creates new frame from the information of the PM frame 20 that is received and realize by using usually) of the time stamp that receives and transmission to PM frame 20 then, and new frame is sent to next MIP 52 or destination MEP 50d.

Transmitter can be activated or be forbidden by NMS 42.When transmitter 34 was activation, it was determined being arranged in the measurement in the future duration at interval one at a time.Transmitter 34 is safeguarded unique IID (ID at interval) in the scope of transmitter device.Source MEP, IID} have defined an interval (in one of IID sequence number is taken turns) uniquely, and as the master index in the storage list.If a plurality of examples operate among the MEP, then these examples will be shared same IID distributor, make when an example begins new interval, guarantee the IID of each unique MEP.When having a plurality of example, at interval can be overlapping.According to the memory capacity of transmission rate and receiver, should be set to not oversize, not too short " reasonably " value the duration.Duration can be randomized, but and nonessential like this.

When measurement begins at interval, adopt suitable algorithm computation to be used to send the next scheduled time of next PM frame.In IETF " draft-ietf-ippm-owdp-14.txt, AOne-way Active Measurement Protocol (OWAMP) " (the unidirectional active measurement agreement) in December, 2004, a kind of such algorithm has been described.Repeat this algorithm as follows:

Algorithm S: given arithmetic number ' mu ', generating average is the exponential random variable of ' mu '.

Computational constant at first, in advance

Q[k]＝(ln2)/(1！)+(ln2)^2/(2！)+...+(ln2)^k/(k！)，1＜＝k＜＝11

In next part, provide the exact value that must be realized use by all.Must guarantee that all realize all having generated exactly identical pseudo random sequence.

S1.[obtains U, and carries out displacement] generate the equally distributed random binary mark (fraction) of 32 bits

U=(.b0 b1 b2...b31) [noting binary point (binary point)]

If U＜-(.b_{j+1}...b31) situation under, the first zero bit b_j are positioned, and j+1 bit before removing

Annotate:, stipulate that this algorithm returns (mu*32*ln2) can not find under the zero a few cases.

Does S2.[accept immediately? if] U＜ln2, then establish X＜-mu* (j*ln2+U), and stop this algorithm.(annotate: Q[1]=ln2.)

S3.[minimizes.] find minimum k＞=2 to make U＜Q[k].Generate k new all

The even random binary mark U1 that distributes ..., Uk, and establish V＜-min (U1 ..., Uk).

S4.[transmits answer.] establish X＜-mu* (j+V) * ln2.

For the sake of clarity, can indicate when k infinitely increases Q[k] the limit be 1, and binary fraction symbol (.b0 b1..b31) means $\underset{i=0}{\overset{31}{\mathrm{\&Sigma;}}}\frac{b_i}{2^{i+1}}{\mathrm{<figure-callout\; id="0"\; label="b"\; filenames="A20061009226100251.png"\; state="\{\{state\}\}">b</figure-callout>}}_0\&CenterDot;2^{-1}+b_1\&CenterDot;2^{-2}+...+{\mathrm{<figure-callout\; id="31"\; label="b"\; filenames="A20061009226100241.png"\; state="\{\{state\}\}">b</figure-callout>}}_{31}\&CenterDot;2^{-32}.$ Q[k] 11 values following (drawing) from above referenced draft-ietf-ippm-owdp-14.txt:

#1 0xB17217F8，

#2 0xEEF193F7，

#3 0xFD271862，

#4 0xFF9D6DD0，

#5 0xFFF4CFD0，

#6 0xFFFEE819，

#7 0xFFFFE7FF，

#8 0xFFFFFE2B，

#9 0xFFFFFFE0，

#10 0xFFFFFFFE，

#11 0xFFFFFFFF

For example, Q[1]=ln2 is in fact approximate by 0xB17217F8/ (2**32)=0.693147180601954; For j＞11, Q[j] be 0xFFFFFFFF.

This algorithm returns time duration X, and X is used to calculate moment T in the future, and frame will be sent out at this moment T.Before this predetermined transmitting time T arrives, should in memory, be ready to randomized frame, and described frame is got ready for sending except that lacking the time stamp, can read time stamp at the last moment, perhaps uses T enough accurately to estimate this time stamp.In case in frame, be provided with time stamp, can also calculate verification and.Other field should be filled according to table 1.Frame length also is randomized (evenly distributing between minimum (min) and maximum (max) length), and filling bit also is randomized (evenly distributing between 0 and 1).And any and security-related frame construction process should be realized at this moment.This realization should be considered the fact that T may very fast arrival, and can have the frame that is ready to send at any time by constructing some frames in advance.

When measuring when finishing at interval, transmitter writes down the number of each frame that sends at interval, and NMS recovers to be used for after a while, and restarts circulation.

Fig. 6 shows send 575 and 857 frames respectively two examples of IID=27 and IID=28 at interval.As shown in the figure, the time that sends between the PM frame 20 is not equally distributed, and the length of PM frame also is different.

Fig. 7 shows the state diagram of receiver 36.Receiver 36 can be in halted state 64 or state of activation 66.When receiver 36 was disabled, it was ignored and abandons the PM frame of receiving.When receiver 36 was activated, it at first collected its local system clock, arrives time stamp the most accurately so that have.Realize to optimize and can comprise the clock module that utilizes in the framer (framer), in case frame access arrangement, this clock module will obtain time value, and stores with certain effective and efficient manner (inner header, metadata etc.).If there is not available this module, then frame will experience more delays (port queuing, port are handled, bridge control centre lines up), and just can handle in the part of the PM of receiver specific (PM-specific) subsequently.In the case, the alignment error (to the estimation of additional delay) of system should be included in the delay estimation, and it is reported to NMS 42.

Then, calculate delay, and will postpone to compare with miss-threshold.If postpone to have surpassed miss-threshold, think that then frame loses, and ignore this frame.Otherwise, in MIB38, adding new clauses and subclauses, { seq_num} carries out index to these clauses and subclauses for source MEP, IID by tlv triple.At present, in 802.1ag, this source MEP (SourceMEP) sign is defined as the serial connection of the Local MEP ID of MD name, MA name and 13 bits.Store 4 values: describe priority 3 fields (DE, DP UP) and the delay that is used for this frame,, can not define according to having defined which mark in the frame (DE, DP, UP).

Also safeguard and lose counting (being actually the number of received frame), this lose counting by 5 tuples srcMEP, IID, DE, DP, UP} carries out index.The user may wish to see the details of this degree, perhaps only sees every { srcMEP, IID} mean value on all priority.In order to carry out correct index, the tabulation of the value of DE, DP, UP comprises " undefined " (undefined): the value of DE be 0,1, undef}, and the value of DP and UP be 0,1,2,3,4,5,6,7, undef}.In addition, index (undef) (undef) (undef) is invalid, and DE and DP are undefined, perhaps all define.

Table 2 shows the example of the receiver in " MEP3 " arbitrarily, only shows corresponding to from " MEP6 " and the clauses and subclauses of the PM frame in 27 at interval thereof.In this example, receive all frames except that frame #3, in table, do not have the seq_num of frame #3.In the example below, for the sake of simplicity, the DE/DP/UP index is not shown.These examples can be corresponding to the measurement situation with all identical priority.

Example-frame delay of table 2 receiver MIB

srcMEP	IID	seq_num	Length of delay
				MEP6
	27	0	?25.3μs
				MEP6
	27	1	?29.2μs
				MEP6
	27	2	?21.5μs
				MEP6
	27	4	33.2μs
				MEP6
	27	5	25.4μs
				...	...	...	...
MEP6	27	574	38.7μs

Table 3 shows and is used for the counting that Loss Rate is estimated, does not have { DE, DP, UP} priority index equally.

Table 3 receives successful example

?srcMEP	?IID	?rcvd_OK
			?MEP6	?27	?574
?MEP6	?28	?857

From reducing the load on the NMS 42 and the angle of data volume to be exported, except that with NMS 42 communicates by letter, NMS interface module 40 can also realize " post-processed " (post-processing) function, for example DV (delay distortion) calculates, at MIB when saturated or after certain predetermined delay, carry out MIB and remove, perhaps more Premium Features.

Figure 8 illustrates DV calculates.DV computing function operation choice function, normally Min-to-Max (search minimum delay and maximum delay in sampling).This choice function be applicable to limit by source MEP, may { measurement on the basis of UP} at interval for DE, DP every.Choice function may need or may not need to guarantee to finish at interval.If this function need be guaranteed to finish at interval, then a kind of method of detection resources MEP interval N end is to observe the frame that arrives: arrive if belong to the frame at follow-up interval, then the time stamp of this frame can not have to be used under the situation of signaling positively to determine a moment, no longer can receive frame from interval N at this moment destination MEP.In other words, if N is to be verified at interval, the frame that then needs only from interval N '＞N arrives, and just can guarantee that T=frame.Timestamp+loss_threshold (time stamp+miss-threshold of T=frame) is after the ending of interval N constantly.The simpler selection function of for example " preceding two " (selecting preceding two clauses and subclauses of sampling) and so on does not need to finish at interval just can begin to calculate, and reason is at least two clauses and subclauses of its needs.

Remove function vacating space in MIB: it removes all clauses and subclauses from an interval, and replaces them with summary (minimum delay, average retardation postpone intermediate value, maximum delay).A kind of possible method that realizes is that the data that will remove and the frame count that is received and DV (delay distortion) combine, and is as shown in table 4:

The MIB example based on the interval of table 4 receiver

srcMEP	?IID	Minimum delay (minDelay)	Average retardation (avgDelay)	Maximum delay (maxDelay)	DV (F：Min-to-Max)	Counting
							MEP6	?27	?21.5μs	?26.6μs	?38.7μs	?17.2μs	?574
MEP6	?28	?20.2μs	?25.9μs	?105.0μs	?84.8μs	?857

NMS will carry out delay and the information of delay distortion and the collection of correlation, thereby calculate Loss Rate:

The MEP6-sender report has sent 575 frames in interval 27

The report of MEP3-receiver has received 574 frames corresponding to the interval 27 of MEP6 from MEP6

NMS can calculate the estimated value of Loss Rate to the interval 27 that MEP6 limits:

(1-574/575)＝0.00174＝0.174％

The MEP6-sender report has sent 857 frames in interval 28

MEP3-receiver report has received 857 frames corresponding to the interval 28 of MEP6 from MEP6

NMS can calculate the estimated value of Loss Rate to the interval 28 that MEP6 limits:

(1-857/857)＝0.00000＝0.000％

Loss Rate estimates to have the precision identical with sample size: thumb principle (rule ofthumb) is to observe about 20 or 30 incidents at least, to draw statistical conclusions.Therefore, size can not detect Loss Rate less than 20/X for the sampling of X.This is not a major issue, because purpose is that these problems may cause SLA inconsistent, disconnect even connect to the potential problem of NMS 42 circulars.

The invention provides the significant advantage that is better than prior art.At first, the present invention can easily realize in existing OAM structure, and the present invention can share the hardware that has been used for OAM(Operation Administration Maintenance).This makes operator and provider be easy to manage the present invention.The second, this solution can adopt the statistical significance of randomization to guarantee that the better second layer is measured.The operation of second layer driver might be than the 3rd layer of processing module in operation (if the second layer drive is not to run directly on the network interface card) under the much closer situation of network interface card, thus obtain to actual transmitting time better estimation (just before sending first bit) and to the better estimation (after just receiving last bit) of actual reception time.Although these errors can compensate by calibration, had better minimize these errors or fully avoid these errors.

The 3rd, the present invention need not inform the signaling protocol of the node among the MA in order to the parameter during will measuring; The equivalents that the IID source that makes can initiation session at interval, and do not need to set up actual session, so performance monitoring need be such as the session protocol OWAMP or the TWAMP (two-way active measurement agreement).By the parameter of source of configuration, ID provides the ability of regulating sample frequency and sampled packet size at interval, has therefore improved sampling quality so that obtain better statistical significance.The 4th, the receiver function need not change based on the particularity of measuring session; No matter test is (being initiated by operator) that initiated automatically by transmitter or as required, and no matter what (for example fills and change) the randomization selection is, the function of receiver all remains unchanged.Performance monitoring not only works in the environment of point-to-point, and works in the environment of multiple spot.In other words, identical receiver function can be used for a plurality of metric parameter, wherein for example a source to a plurality of destinations broadcast/multi broadcast OAM frame.The 5th, because the present invention is integrated among the OAM, so do not need the superimpose device of adding.The 6th, balance of the present invention the advantage of second layer monitoring: time stamp and deterministic path more accurately.

Invention has been described in conjunction with the Ethernet protocol that is used for the second layer, yet should note and can also use the present invention in conjunction with other second layer agreement.

Although detailed description of the present invention is at specific exemplary embodiment, those of ordinary skill in the art can obtain about the various modifications of these embodiment and the enlightenment of alternate embodiment.Present invention resides in modification arbitrarily and alternate embodiment in the claim scope.

Claims (10)

1. method that is used for carrying out at network performance monitoring comprises step:

Generate the performance monitoring frame of the second layer by first node, it is one or more type fields of performance monitoring frame that wherein said performance monitoring frame has frame identification, specify the initiation time stamp field of the time that sends described frame, and one or more features of wherein said performance monitoring frame change;

In predefined interval, send described performance monitoring frame by network;

Receive described performance monitoring frame by the Section Point in the described network, and the arrival time stamp that is associated with the reception of described performance monitoring frame of storage;

Described initiation time stamp and described arrival time stamp according to described performance monitoring frame are determined performance index.

2. method according to claim 1, wherein said performance monitoring frame are to send with the time interval that changes.

3. method according to claim 1, the length of wherein said performance monitoring frame changes.

4. method according to claim 1, the content of wherein said performance monitoring frame changes.

5. method according to claim 1, wherein said initiation time stamp was right after before described forwarding step and generates.

6. method according to claim 1, wherein each performance monitoring frame also comprises one or more id fields, its sign specific performance properties is measured at interval.

7. method according to claim 6, wherein said id field has also shown sequence number, it identifies each the performance monitoring frame in the described specific performance properties measurement at interval uniquely.

8. method according to claim 6, wherein said Section Point detect the beginning of performance monitoring session by the described id field with reference to the performance measurement grouping that is received.

9. method according to claim 1, wherein the one or more nodes between described first node and described Section Point also generate the time stamp that is stored in the described performance monitoring frame.

10. circuit that is used for carrying out at network performance monitoring comprises:

Be used for generating the circuit of the performance monitoring frame of the second layer by first node, it is one or more type fields of performance monitoring frame that wherein said performance monitoring frame has frame identification, specify the initiation time stamp field of the time that sends described frame, and one or more features of wherein said performance monitoring frame change;

Be used for sending by network the circuit of described performance monitoring frame at predefined interval;

Be used for the circuit that Section Point by described network receives the arrival time stamp that described performance monitoring frame and storage be associated with the reception of described performance monitoring frame;

Be used for the circuit of determining performance index according to the described initiation time stamp and the described arrival time stamp of described performance monitoring frame.

Images