CN101051959A - Detecting method for network link band width based on hop-by-hop time stamp label - Google Patents

Detecting method for network link band width based on hop-by-hop time stamp label Download PDF

Info

Publication number
CN101051959A
CN101051959A CN 200710099115 CN200710099115A CN101051959A CN 101051959 A CN101051959 A CN 101051959A CN 200710099115 CN200710099115 CN 200710099115 CN 200710099115 A CN200710099115 A CN 200710099115A CN 101051959 A CN101051959 A CN 101051959A
Authority
CN
China
Prior art keywords
test packet
message
test
bandwidth
load
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN 200710099115
Other languages
Chinese (zh)
Other versions
CN100442730C (en
Inventor
崔毅东
金跃辉
焦利
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Posts and Telecommunications
Original Assignee
Beijing University of Posts and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing University of Posts and Telecommunications filed Critical Beijing University of Posts and Telecommunications
Priority to CNB2007100991150A priority Critical patent/CN100442730C/en
Publication of CN101051959A publication Critical patent/CN101051959A/en
Application granted granted Critical
Publication of CN100442730C publication Critical patent/CN100442730C/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The method comprises: at first, using a test message and a load message to form a test message queue; using the detection source end to send the test message queue to network by using back to back way; each router in the transmission link records the current time stamp of each router in each test message; so the detection destination end gets a sample set comprising multi test data samples; finally, using bandwidth algorithm to make filter for the test data sample set to get physical bandwidth of the network link.

Description

Method of measurement based on the network link bandwidth of hop-by-hop time stamp label
Technical field
The present invention relates to a kind of bandwidth measurement method of network link, exactly, relate to a kind of method of measurement of the network link bandwidth based on the hop-by-hop time stamp label, belong to the latency measurement technical field of communication network.
Background technology
Since communication network was born, the Measurement and analysis of network performance was important research direction always, it for the behavior of people's awareness network, improve network performance, improve service quality significance is all arranged.Simultaneously, concerning various underlying protocols and higher layer applications, the end to end performance parameter on the awareness network path helps to realize guaranteeing target such as service quality such as load balancing, congested control, QoS routes.Yet along with the increase of network complexity, accurate network performance parameter is measured and often is difficult to realize.The network bandwidth is to characterize estimating of network information transfer ability, also is the important parameter of reaction network service quality, plays an important role in traffic engineering.Using as mass data such as FTP, P2P, VoIP exchanges all needs abundance, stable network bandwidth with guaranteed qos, and the bandwidth measurement technology can be QoS and detects and provide support.
In view of bandwidth measurement significant aspect the evaluating network QoS, correlation technique obtains extensively, research in depth.According to the different criteria for classifying, can carry out following classification to bandwidth measurement: (1) according to whether injecting detection packet to network internal can be divided into passive measurement and initiatively measure.Most of bandwidth measurement technology adopt initiatively metering system.(2) whether need the cooperation of node router according to measuring process, the link bandwidth that the bandwidth measurement technology can be divided into hop-by-hop measure and end to end path bandwidth measure.(3) by different estimating, bandwidth measurement can be divided into link bandwidth (Link Capacity), path bandwidth (Path Capacity) and three kinds of measuring techniques of available bandwidth (Available Bandwidth).(4) say from method of measurement, have based on the method for measurement that becomes packet sequence VPS (Variable Packet Size), the method of measurement of disperseing PTD (Packet TrainDispersion) based on bag, based on the method from duty cycle stream SLoPS (Self-Loading Periodic Stream), the method for measurement of discharge model Network Based.
Introduce the characteristics and the defective of three kinds of prior aries related to the present invention below respectively:
First kind of prior art is to connect by simulation TCP to obtain path throughput BTC (Bulk TransferCapacity).This method is to study from the angle of transport layer, and basic thought is to set up a TCP session connection for measurement, and the session connection by this simulation obtains some parameters, utilizes these parameters to obtain the bandwidth utilized of path again.Its implicit thought be this that set up for measurement connect must and original background traffic 100% occupy the whole bandwidth of link, when measuring, also require to adopt the congestion avoidance algorithm that is similar to the TCP connection simultaneously; And, in order to reach the balancing speed (balancing speed is the speed in the congestion avoidance procedure) in the congested control procedure of TCP, Session Time must reach more than 10 seconds end to end (needs the regular hour because set up stabilization process), measures the parameter that obtains like this and could reflect the size that can utilize bandwidth truly.But in fact,, and adopted congestion avoidance algorithm, thereby the measurement result of its acquisition reflects can only be the TCP throughput, but not can utilize bandwidth because the measurement of carrying out based on this method is closely related with the agreement of TCP layer.Survey tool based on this thought has two kinds: Treno and Cap.In fact, Treno is the simple realization version that TCP is used to measure, and its independent proposition is in order to solve the different compatibility issue of TCP version in measurement.It can provide two kinds of measurement results simultaneously: the Mean Speed in Mean Speed-whole data transmission procedure, the speed of balancing speed-Treno in congestion avoidance procedure, or perhaps a kind of ideal rate, because it does not comprise that retransmission timer overflows the overtime and slow start-up course of RTO (retransmissiontimer overflow).Cap and Treno are similar, have some differences only in realization.One of its feature is to require to cooperatively interact at transmitting-receiving two-end, and this has increased the actual difficulty that uses undoubtedly again.
The shortcoming of this prior art is: connect the method for measurement that obtains BTC by simulation TCP and must consume a large amount of bandwidth, if improper use also may cause network over loading, thereby have influence on normally carrying out of customer service; Therefore this method is called as " bandwidth killer ".Though some actual measurements have been verified the accuracy of these instruments than higher,, this method of measurement causes it to be difficult to obtain real application for the huge loss of bandwidth, and this is its disadvantage place just also.
Second kind of prior art is based on the bandwidth measurement method of elongated packet sequence model.In order to measure some parameters of hop-by-hop, its measurement result should be the link physical bandwidth to this model at first.But the instrument based on this model also can be used for measuring the path bottleneck bandwidth.Its assumed condition has: transmission delay and bag size are linear, background traffic only can increase packet delay, in the observed value for a plurality of bags of specific size, can be similar to and think that minimum the wrapping in of round-trip delay RTT (Round Trip Time) value do not experience queueing delay in the link, bag forwarding time can ignore.Method of measurement is: the source end sends n measurement bag of the identical size of row earlier to destination, this process repeats k time, all gets different big or small measurement bags at every turn.After destination receives, return corresponding ack and confirm bag, after these confirm that bag returns the source end, the RTT value of onesize n the bag that calculating is received, select wherein minimum value, the bag of every kind of size is not always the case, and by calculating the difference and corresponding RTT difference of k bag size, can obtain bandwidth again.Because calculating the numerical value of being got all is difference, and assumed condition is arranged, thereby can eliminate the influence of background traffic.Can obtain the result by graphing method in the practical application, be longitudinal and transverse coordinate described point with packet delay and bag size respectively, makes straight line with linear regression method again, and the inverse of this straight slope is the link physical bandwidth.Use the main tool of this method that pathchar is arranged, Clink, pchar, netimer etc.
Because what this method adopted when calculating is the difference of bag size and the difference of minimum RTT, thereby the result who measures can not be subjected to the influence of background traffic, thereby just determined survey tool based on this principle can only be used for the measure link band width in physical yet but not can utilize bandwidth.This method has following limitation:
1. Measuring Time is long, and the bandwidth of consumption is bigger.As being 10ms, 10 Ethernets of jumping for an average round-trip delay, instrument pathchar measures once needs 144 seconds, and needs to send the measurement data of 10Mb.
2. the survey tool based on VPS uses ICMP Request bag or UDP bag to measure mostly, requires to obtain feedback information timely from each node.But a lot of routers and main frame may give this bag of two types lower processing rank in the network, even disregard, and this will inevitably cause the inaccurate of measurement result.
3. store-and-forward node such as two layers and following equipment such as switch, bridge, repeater in the network, because it does not generally have the IP address, they are sightless " hiding " nodes for the employing instrument that IP life span TTL (Time To Live) value measures that successively decreases, but in fact, can take certain processing time when wrapping through them, this also can cause the deviation of measurement result.
4. that adopt in this model is two-way delay RTT, but not one-way latency OWD (One-Way Delay), because two-way link is generally asymmetric, but in this model, normally two-way link is considered as that symmetric links handles, this also can cause the deviation of measurement result.
5. there is the measurement upper limit in this method, the time granularity that path MTU MTU (Maximum TransportUnit) and end system can provide has determined the bandwidth upper limit that it can record, limit by present software and hardware technology, can not be used for the measurement of express network based on the instrument of this measuring principle.
The third prior art is based on the method for bag to model.1988, V.Jacobson at first proposed to wrap right notion.This model comprise bag to bag row two kinds of submodels.Because the assumed condition of two kinds of submodels is different with requirement, the parameter that they can be measured is also inequality, and bag is to being used to measure the path bottleneck bandwidth, and the bag row are used to measure path and can utilize bandwidth.The measuring principle of wherein wrapping model is: bag need be based on to draw a conclusion to model, if promptly two wrap in the bottleneck link place be before and after followed by, these two wrap in all later links even the time interval on the reverse link all is constant so, and this time interval is that the bandwidth by bag size and bottleneck link place is determined.By the size of this time interval and bag, can obtain the bandwidth of path.Main tool: b/cprobe, pathrate, TOPP etc.The measuring principle of bag row model is: need known path bottleneck bandwidth, on this basis, the source end sends row via path to be measured to destination and measures bag, requires to measure the speed height of bag to being enough to make itself and background traffic use the path bottleneck bandwidth in 100% ground within a certain period of time.The source end is noted the time of advent of measuring Bao Liezhong first and last affirmation bag that returns, adopt certain processing method after, can draw the background traffic at bottleneck link place, obtain the value that can utilize bandwidth more indirectly.The measurement path that proposes based on this principle can utilize the bandwidth instrument that Cprob is arranged.
Bag mainly contains following shortcoming to method: 1 queuing failure: measure bag and perhaps do not line up at the bottleneck link place.2. background traffic: can have influence on the interval between bag centering two bags.3. packet loss: the part bag may directly be lost, and other bags can produce router overflows after to be lost, thereby obtains the result of mistake.4. downstream links is congested: wrap on the bottleneck link link afterwards as if two and line up once more, can cause that the time interval of two parlors changes, thereby make the result depart from actual value.5. the measuring accuracy problem of multichannel (Multi-channel), form by a plurality of channels such as a path, because the bag in the one-shot measurement is to can only be via a Channel Transmission, thereby its measurement result is single-channel bottleneck bandwidth, but not the path bottleneck bandwidth.
Summary of the invention
In view of this, the method of measurement that the purpose of this invention is to provide a kind of network link bandwidth based on the hop-by-hop time tag, this method can overcome the various defectives of prior art, improve the precision of bandwidth measurement, eliminate of the influence of the asymmetry of measurement path to bandwidth measurement, reduce the mass data message that produces in the measuring process simultaneously, and strengthen the real-time of bandwidth measurement.
In order to achieve the above object, the invention provides a kind of method of measurement of the network link bandwidth based on the hop-by-hop time stamp label, it is characterized in that: adopt test packet and load message to form a kind of test packet formation earlier, send this test packet formation in mode back-to-back to network by the detection source end again, each router in the transmission link stabs this message queue and is recorded in respectively in each test packet of this formation through the current time of each router, thereby obtain a plurality of test data sample sets at the detection destination, by the bandwidth statistic algorithm test data sample set is carried out Filtering Processing at last, obtain the band width in physical of network link.
Described method comprises the following steps:
(1), forms the test packet formation with test packet that has the timestamp option and load message at the test source end;
(2) send this test packet formation in mode back-to-back to destination from the test source end;
(3) each hop router of test packet formation process all will stab in each probe messages that records this formation the current time;
(4) destination receives the test packet formation, extracts the time stamp data in each test packet, obtains the measuring data sample set;
(5) adopt the bandwidth statistic algorithm that measuring data sample is gathered and carry out Filtering Processing, keep satisfactory data, and abandon undesirable data, obtain the band width in physical of this network link at last.
Described mode back-to-back is meant that two or more IP messages adopt successively, the mode in the no time interval sends in the network via network interface unit.
Described timestamp option is the one or more flag bits in the IP message, and when the router on the IP message process network that has this flag bit, router can write the current time in this IP message; Organize No. 781 document RFC781 of the request note RFC standard of issue to be an example of timestamp option at the Internet engineering duty group IETF.
The formation of described method test packet is set to high priority, so that reduce the influence between probe messages formation and the background traffic and reduce the influence of the extra load of probe messages to network as far as possible, improves efficiency of measurement simultaneously.
Described test packet formation comprises two parts: lay respectively at message queue head and afterbody, be used to obtain message queue through the test packet of link router time be positioned in the middle of the message queue, be used to reflect the load message of network traffics characteristic, it is formed structure and is: P Test={ M H1, M H2... M HN, LOAD, M T1, M T2..., M TN, in the formula, natural number N is the sequence number of test packet, M H1, M H2..., M HNBe the head test packet sequence that N test packet formed, M T1, M T2..., M TNBe the afterbody test packet sequence that N test packet formed, each test packet all is provided with the timestamp option, for router this message is write down wherein through the current time of this router; LOAD is the load message, and the total byte length of load message queue is 10 times of total byte length of test packet formation at least, and total message number of load message queue should be more than 100; The load newspaper Wen Liezhong of team, the ratio of the shared load message total of the message of all lengths should meet the feature of current internet network flow; Packet such as 40 bytes, 552 bytes (or 576 bytes) and 1500 bytes in the load message row accounts for 50%, 35% and 15% of sum respectively.
Test packet is the internet control message protocol request echo ICMP Echo Request message that has the transmission control protocol TCP/ user datagram protocol UDP message of timestamp option or have the timestamp option.
The computational methods of described bandwidth statistic algorithm are as follows:
(51) obtain k head test packet Head in the probe messages formation respectively kThe timestamp sequence T Hk = { t 1 Hk , t 2 Hk , . . . , t M Hk } With k afterbody test packet Tail kThe timestamp sequence T Tk = { t 1 Tk , t 2 Tk , . . . , t M Tk } , wherein, natural number k and M are respectively the number of links of the sequence number and the tested path of test packet;
(52) calculating the byte total length is L Totol=Len Load+ N * Len MearMessage queue through every jumping spent in the test path time G M k: G M k = T Tk - T Hk , in the formula, L TotolBe the total byte length of all test packets and load message, Len LoadBe the total byte length of load message, Len MearBe the byte length of single probe messages, ignore because of the caused error of test packet in this test process, because the shared byte length of logging timestamp is much smaller than load message queue size in the test packet;
(53) calculate the band width in physical b of every jumping link in the path that k head test packet and afterbody test packet detected M k: b M k = L totol G M k = len load + N × Len mear G M k ;
(54) according to the method described above, respectively N test packet carried out computing one by one, obtain the test sample book data acquisition system of the band width in physical value of every jumping link, again Filtering Processing is carried out in this sample data set, finally obtain the band width in physical of every jumping link.
The described filtering technique that this sample data set is handled has following four kinds:
A, directly average: because of internet traffic has randomness, the estimation error of this method is big;
B, exponential filter: B k=B K-1* α+(1-α) * b k, in the formula, B kAnd B K-1Be respectively filtered bandwidth, b kBe the bandwidth sampled value, α be in this computing formula before and after two to gained result after the filtering influence the weights factor, the α span is 0.3～0.8; This method realizes simple, but works as b kThe high time error of change frequency is big;
C, based on the filter of kernel function algorithm for estimating: measure bandwidth for satisfying formula max{ α [d (x)/d (x) Max]+β p (x)+γ r (x) } x; In the formula, α, β and γ are the filter function coefficient, satisfy normalizing condition alpha+beta+γ=1; Density function is d ( x ) = 1 n Σ i = 1 n K ( x - x 1 c · x ) , K ( t ) = 1 + t , t ≤ 0 1 - t , t > 0 Be kernel function; This method takes into full account a variety of causes that causes error, filtering interfering bandwidth preferably, but amount of calculation is big, and convergence rate is slow, is not easy to real-time measurement;
D discretization low pass filter: B k = α k B k - 1 + ( 1 + α k ) b k + b k - 1 2 , in the formula, B kAnd B K-1Be respectively filtered bandwidth, b kAnd b K-1Be respectively the bandwidth sampled value, α k=(2 τ-Δ κ)/(2 τ+Δ κ), τ is the sampling period of filter, Δ k is the time difference; This method is simply quick, the effective interference flow of filtering burst and effectively utilize the correlation of adjacent measurements, and fast convergence rate, amount of calculation is little.
The present invention is a kind of method of using bag formation Measurement Network bandwidth, i.e. PTTSM (Packet TrainTime Stamp Method) method.A kind of new bag formation of design earlier of this method, this bag formation then obtains by the timestamp of the test packet of portion end to end in this message queue through the disengaging time of router, rather than obtains by traditional ACK message.Like this, this PTTS method sends a probe messages formation, just can obtain a plurality of data samples, makes that the probe messages number in the test process significantly reduces, for the data processing filtering in later stage is provided convenience.In a word, beneficial effect of the present invention comprises following content: have a plurality of test packets in each probe messages formation of this method, so just but one-shot measurement just obtains a plurality of detection samples, has reduced the transmission number of times of probe messages formation, has reduced the influence of detection flows to background service effectively.Bandwidth measurement of the present invention is to adopt the processing time of timestamp record message through link, than the additive method precision height of prior art.In addition, the method for this employing hop-by-hop record message timestamp makes a probe messages formation just can add up a data sample of the every jumping bandwidth of acquisition, for the batch process of data good help is arranged.At last, the return time that obtains the ACK message that is not that employing is traditional of detection time of the present invention so just can effectively reduce the influence of reverse path to test result.
Description of drawings
Fig. 1 is the method for measurement flow diagram that the present invention is based on the network link bandwidth of hop-by-hop time stamp label.
Fig. 2 is that the structure of test packet formation of the present invention is formed schematic diagram.
Fig. 3 is the schematic diagram of the network test system of one embodiment of the invention-employing TCP/UDP packet.
Fig. 4 is the schematic diagram of the network test system of another embodiment of the present invention-employing ICMPcho message.。
Embodiment
For making the purpose, technical solutions and advantages of the present invention clearer, the present invention is described in further detail below in conjunction with accompanying drawing.
The present invention is a kind of method of measurement of the network link bandwidth based on the hop-by-hop time stamp label, it is to adopt test packet and load message to form a kind of test packet formation earlier, send this test packet formation in mode back-to-back to network by the detection source end again, part message in this message queue is provided with " timestamp option ", each router stabs this message queue in each test packet that is recorded in this formation respectively (referring to the regulation of RFC 781) through the current time of each router, thereby obtain a plurality of test data sample sets at the detection destination, by the bandwidth statistic algorithm test data sample set is carried out Filtering Processing at last, obtain the band width in physical of network link.In order further to reduce the influence between probe messages formation and the background traffic, probe messages formation of the present invention is set to high priority, can reduce the influence of the extra load of probe messages to network like this, has improved efficiency of measurement simultaneously.
Referring to Fig. 1, introduce the operating procedure of this method:
(1), forms the test packet formation with test packet that has the timestamp option and load message at the test source end;
(2) send this test packet formation in mode back-to-back to destination from the test source end;
(3) each hop router of test packet formation process all will stab in each probe messages that records this formation the current time;
(4) destination receives the test packet formation, extracts the time stamp data in each test packet, obtains the measuring data sample set;
(5) adopt the bandwidth statistic algorithm that measuring data sample is gathered and carry out Filtering Processing, keep satisfactory data, and abandon undesirable data, obtain the band width in physical of this network link at last.
For the truth of better reflection network, not the probe messages that sends equal length simply in the inventive method, but send the probe messages formation.Its structure is formed as shown in Figure 2.This examination message queue is made up of test packet and load message two parts, and wherein test packet lays respectively at message queue head and afterbody, is used to obtain the time of message queue through the link router.The load message is positioned in the middle of the message queue, be used to reflect the network traffics characteristic.Wherein the total byte length of load message row must be far longer than the total byte length (being 10 times or more times of total byte length of test packet row such as the total byte length of load message row) of test packet row.Total message number of load message row should abundant (such as more than 100).Research statistics according to correlative study mechanism, in present IP network flow, the message composition of all lengths roughly is: the packet that the packet that the packet of 40 bytes accounts for 50%, 552 byte (or 576 bytes) accounts for 35%, 1500 byte accounts for 15%.The sequence arrangement of message queue is: P Test={ M H1, M H2... M HN, LOAD, M T1, M T2..., M TN, in the formula, head test packet number and afterbody test packet number are N, M H1, M H2..., M HNBe the head test packet sequence that N test packet formed, M T1, M T2..., M TNAfterbody test packet sequence for N test packet composition, each test packet all is provided with one or more flag bits as the timestamp option, for router this message is write down wherein (referring to RFC 781 " regulation of IP agreement timestamp option " " ASPECIFICATION OF THE INTERNET PROTOCOL (IP) TIMESTAMPOPTION ") through the current time of this router.
The concrete grammar of the bandwidth statistic algorithm in the aforesaid operations step (5) is as follows:
(51) obtain k head test packet Head in the probe messages formation respectively kThe timestamp sequence T Hk = { t 1 Hk , t 2 Hk , . . . , t M Hk } With k afterbody test packet Tail kThe timestamp sequence T Tk = { t 1 Tk , t 2 Tk , . . . , t M Tk } , wherein, natural number k and M are respectively the number of links of the sequence number and the tested path of test packet;
(52) calculating the byte total length is L Totol=Len Load+ N * Len MearMessage queue through every jumping spent in the test path time G M k: G M k = T Tk - T Hk , in the formula, L TotolBe the total byte length of all test packets and load message, Len LoadBe the total byte length of load message, Len MearBe the byte length of single probe messages, because the shared byte length of logging timestamp is much smaller than load message queue size, so ignore because of the caused error of test packet in the test packet in this test process.
(53) calculate the available bandwidth b of every jumping link in the path that k head test packet and afterbody test packet detected M k: b M k = L totol G M k = len load + N × Len mear G M k ;
(54) according to the method described above, respectively N test packet carried out as above computing one by one, obtain the test sample book data acquisition system of the band width in physical value of every jumping link, again Filtering Processing is carried out in this sample data set, finally obtain the band width in physical of link.
This moment, adoptable filtering technique was as follows:
A, directly average: because internet traffic has randomness, the estimation error of this method is big;
B, exponential filter: B k=B K-1* α+(1-α) * b k, in the formula, B kAnd B K-1Be respectively filtered bandwidth, b kBe the bandwidth sampled value, α is two weighing factors to gained result after the filtering in front and back in this computing formula, and the α span is 0.3～0.8; This exponential filter method realizes simple, but can not filter out the high bandwidth value of frequency well;
C, based on the filter of kernel function algorithm for estimating: measure bandwidth for satisfying formula max{ α [d (x)/d (x) Max]+β p (x)+γ r (x) } x; In the formula, α, β and γ are the filter function coefficient, satisfy normalizing condition alpha+beta+γ=1; Density function is d ( x ) = 1 n Σ i = 1 n K ( x - x i c · x ) , K ( t ) = 1 + t , t ≤ 0 1 - t , t > 0 Be kernel function; This method takes into full account a variety of causes that causes error, filtering interfering bandwidth preferably, but amount of calculation is big, and convergence rate is slow, is not easy to real-time measurement;
D, discretization low pass filter: B k = α k B k - 1 + ( 1 + α k ) b k + b k - 1 2 , in the formula, B kAnd B K-1Be respectively filtered bandwidth, b kAnd b K-1Be respectively the bandwidth sampled value, α k=(2 τ-Δ κ)/(2 τ+Δ κ), τ is the sampling period of filter, Δ k is the time difference; This method is simply quick, and can effectively leach the interference flow of burst and effectively utilize the correlation of adjacent measurements, fast convergence rate, amount of calculation is little.
Comprehensive more above-mentioned various algorithm filters have good filter effect based on the filter and the discretization low pass filter of Density Estimator algorithm.Because the inventive method is mainly used in real-time Measurement Network bandwidth, for the operand and the convergence rate of algorithm specific (special) requirements is arranged, so suggestion adopts the discretization low pass filter to eliminate the error of measurement and the influence of disturbing flow.
The present invention has carried out implementing test, and the result of test is successful, has realized goal of the invention.The situation of following brief description embodiment experiment.
The test macro of first embodiment as shown in Figure 3, the detection source end has sent probe messages formation of the present invention to surveying destination, wherein test packet is for having the prolate-headed TCP/UDP message of timestamp, each router of process all will record the data field in the timestamp option of test packet the current time midway, receive these probe messages at the detection destination again, and extraction time stab data and carry out the processing of bandwidth statistic algorithm, obtain each link Router1～Router2, link Router2～Router3, the link physical bandwidth of link RouterN-1～RouterN.
Referring to Fig. 4, introduce second embodiment test macro of the present invention and test situation thereof.The detection destination of current embodiment can be a various network nodes of supporting the IP agreement, comprises router, main frame etc.The detection source end sends probe messages formation of the present invention to surveying destination, wherein test packet is the ICMP Echo Request message that has the timestamp option, and each router of process all writes down timestamp data field in the timestamp of each test packet midway.After test packet arrives and surveys destination, survey destination node and create ICMP Echo message, former message option partial content is copied in the ICMP Echo message, then this message is sent it back the detection source end, surveying destination to the path of detection source end, each router all similarly records timestamp in each test packet in regular turn.After the detection source end is received test packet, extraction time is stabbed the data in the extension header, carry out the processing of bandwidth statistic algorithm, obtain each link Router1～Router2, link Router2～Router3 ..., link RouterN-1～RouterN, link RouterN～RouterN-1 ..., link Router2～Router1 the link physical bandwidth.The benefit of the metering system of this embodiment is to realize single-ended measurement, promptly only need send probe messages at the detection source end, and receives probe messages.

Claims (9)

1, a kind of method of measurement of the network link bandwidth based on the hop-by-hop time stamp label, it is characterized in that: adopt test packet and load message to form a kind of test packet formation earlier, send this test packet formation in mode back-to-back to network by the detection source end again, each router in the transmission link stabs this message queue and is recorded in respectively in each test packet of this formation through the current time of each router, thereby obtain a plurality of test data sample sets at the detection destination, by the bandwidth statistic algorithm test data sample set is carried out Filtering Processing at last, obtain the band width in physical of network link.
2, method of measurement according to claim 1 is characterized in that, described method comprises the following steps:
(1), forms the test packet formation with test packet that has the timestamp option and load message at the test source end;
(2) send this test packet formation in mode back-to-back to destination from the test source end;
(3) each hop router of test packet formation process all will stab in each probe messages that records this formation the current time;
(4) destination receives the test packet formation, extracts the time stamp data in each test packet, obtains the measuring data sample set;
(5) adopt the bandwidth statistic algorithm that measuring data sample is gathered and carry out Filtering Processing, keep satisfactory data, and abandon undesirable data, obtain the band width in physical of this network link at last.
3, method of measurement according to claim 1 and 2 is characterized in that: described mode back-to-back is meant that two or more IP messages adopt successively, the mode in the no time interval sends in the network via network interface unit.
4, method of measurement according to claim 1, it is characterized in that: described timestamp option is the one or more flag bits in the IP message, when the router on the IP message process network that has this flag bit, router can write the current time in this IP message; Organize No. 781 document RFC781 of the request note RFC standard of issue to be an example of timestamp option at the Internet engineering duty group IETF.
5, method of measurement according to claim 1, it is characterized in that: the formation of described method test packet is set to high priority, so that reduce the influence between probe messages formation and the background traffic and reduce of the influence of the extra load of probe messages, improve efficiency of measurement simultaneously as far as possible to network.
6, method of measurement according to claim 1, it is characterized in that: described test packet formation comprises two parts: lay respectively at message queue head and afterbody, be used to obtain message queue through the test packet of link router time be positioned in the middle of the message queue, be used to reflect the load message of network traffics characteristic, it is formed structure and is: P Test={ M H1, M H2... M HN, LOAD, M T1, M T2..., M TN), in the formula, natural number N is the sequence number of test packet, M H1, M H2..., M HNBe the head test packet sequence that N test packet formed, M T1, M T2..., M TNBe the afterbody test packet sequence that N test packet formed, each test packet all is provided with the timestamp option, for router this message is write down wherein through the current time of this router; LOAD is the load message, and the total byte length of load message queue is 10 times of total byte length of test packet formation at least, and total message number of load message queue should be more than 100; The load newspaper Wen Liezhong of team, the ratio of the shared load message total of the message of all lengths should meet the feature of current internet network flow.
7, according to claim 1 or 2 or 6 described method of measurement, it is characterized in that: test packet is the internet control message protocol request echo ICMP Echo Request message that has the transmission control protocol TCP/ user datagram protocol UDP message of timestamp option or have the timestamp option.
8, method of measurement according to claim 2 is characterized in that: the computational methods of described bandwidth statistic algorithm are as follows:
(51) obtain k head test packet Head in the probe messages formation respectively kThe timestamp sequence T Hk = { t 1 Hk , t 2 Hk , . . . , t M Hk } With k afterbody test packet Tail kThe timestamp sequence T Tk = { t 1 Tk , t 2 Tk , . . . , t M Tk } , Wherein, natural number k and M are respectively the number of links of the sequence number and the tested path of test packet;
(52) calculating the byte total length is L Totol=Len Load+ N * Len MearMessage queue through every jumping spent in the test path time G M k: G M k = T Tk - T Hk , In the formula, L TotolBe the total byte length of all test packets and load message, Len LoadBe the total byte length of load message, Len MearBe the byte length of single probe messages, ignore because of the caused error of test packet in this test process, because the shared byte length of logging timestamp is much smaller than load message queue size in the test packet;
(53) calculate the band width in physical b of every jumping link in the path that k head test packet and afterbody test packet detected M k: b M k = L totol G M k = len load + N × Len mear G M k ;
(54) according to the method described above, respectively N test packet carried out computing one by one, obtain the test sample book data acquisition system of the band width in physical value of every jumping link, again Filtering Processing is carried out in this sample data set, finally obtain the band width in physical of every jumping link.
9, method of measurement according to claim 1 is characterized in that: the described filtering technique that this sample data set is handled has following four kinds:
A, directly average: the estimation error of this method is big;
B, exponential filter: B k=B K-1* α+(1-α) * b k, in the formula, B kAnd B K-1Be respectively filtered bandwidth, b kBe the bandwidth sampled value, α be in this computing formula before and after two to gained result after the filtering influence the weights factor, the α span is 0.3～0.8; Work as b kWhen change frequency was high, this methodical error was big;
C, based on the filter of kernel function algorithm for estimating: measure bandwidth for satisfying formula max{ α [d (x)/d (x) Max]+β p (x)+γ r (s) } x; In the formula, α, β and γ are the filter function coefficient, satisfy normalizing condition alpha+beta+γ=1; Density function is d ( x ) = 1 n Σ i = 1 n K ( x - x i c · x ) , K ( t ) = 1 + t , t ≤ 0 1 - t , t > 0 Be kernel function; This method amount of calculation is big, and convergence rate is slow, is not easy to real-time measurement;
D discretization low pass filter: B k = α k B k - 1 ( 1 + α k ) b k + b k - 1 2 , In the formula, B kAnd B K-1Be respectively filtered bandwidth, b kAnd b K-1Be respectively the bandwidth sampled value, α k=(2 τ-Δ k)/(2 τ+Δ k), τ is the sampling period of filter, Δ k is the time difference; This method effectively filtering burst the interference flow and effectively utilize the correlation of adjacent measurements, fast convergence rate, amount of calculation is little.
CNB2007100991150A 2007-05-11 2007-05-11 Detecting method for network link band width based on hop-by-hop time stamp label Expired - Fee Related CN100442730C (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CNB2007100991150A CN100442730C (en) 2007-05-11 2007-05-11 Detecting method for network link band width based on hop-by-hop time stamp label

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CNB2007100991150A CN100442730C (en) 2007-05-11 2007-05-11 Detecting method for network link band width based on hop-by-hop time stamp label

Publications (2)

Publication Number Publication Date
CN101051959A true CN101051959A (en) 2007-10-10
CN100442730C CN100442730C (en) 2008-12-10

Family

ID=38783165

Family Applications (1)

Application Number Title Priority Date Filing Date
CNB2007100991150A Expired - Fee Related CN100442730C (en) 2007-05-11 2007-05-11 Detecting method for network link band width based on hop-by-hop time stamp label

Country Status (1)

Country Link
CN (1) CN100442730C (en)

Cited By (24)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101917339A (en) * 2010-08-17 2010-12-15 建汉科技股份有限公司 Network bandwidth upper limit counting and adjusting method of network equipment
CN101217497B (en) * 2008-01-18 2011-04-13 北京邮电大学 A path selecting method of wireless mesh network
CN102053217A (en) * 2009-11-05 2011-05-11 上海华虹Nec电子有限公司 Method for quickly processing wafer after sort interruption under working condition
CN102347933A (en) * 2010-07-27 2012-02-08 财团法人交大思源基金会 Network traffic capturing, recovering and replaying method
CN101729306B (en) * 2008-10-28 2012-05-23 中兴通讯股份有限公司 Method and system for testing downlink broadband of access line of broadband network user
CN102710464A (en) * 2012-06-06 2012-10-03 清华大学 Fine-grained single-flow delay detection method based on reference delay insertion structure
CN102946335A (en) * 2012-12-11 2013-02-27 广州中国科学院软件应用技术研究所 Network condition detection method and network condition detection system
WO2013091180A1 (en) * 2011-12-21 2013-06-27 华为技术有限公司 Method, device and system for determining initial file transmission speed
CN103391253A (en) * 2013-08-05 2013-11-13 四川启程科技发展有限公司 Method, device and system for controlling network congestion
CN105024875A (en) * 2014-04-30 2015-11-04 中国电信股份有限公司 Method and system for measuring network flow rate of broadband classified users
CN105591835A (en) * 2015-08-12 2016-05-18 杭州华三通信技术有限公司 Link quality detection method and device
CN106685565A (en) * 2016-12-30 2017-05-17 国网浙江省电力公司绍兴供电公司 Method and system for achieving transparent clock transmission of communication network
CN106850294A (en) * 2017-01-24 2017-06-13 钟炎培 The methods, devices and systems of bandwidth estimation
CN108512726A (en) * 2018-03-29 2018-09-07 上海连尚网络科技有限公司 A kind of method and apparatus of data monitoring
CN109639684A (en) * 2018-12-17 2019-04-16 北京微吼时代科技有限公司 The method and system of dynamic application bandwidth
CN110138590A (en) * 2019-04-09 2019-08-16 上海交通大学 A kind of general TCP performance diagnogtics method and device
CN111130923A (en) * 2019-11-29 2020-05-08 北京达佳互联信息技术有限公司 Network bandwidth determining method and device, electronic equipment and storage medium
CN111162964A (en) * 2019-12-17 2020-05-15 国网智能科技股份有限公司 Intelligent station message integrity analysis method and system
CN112565100A (en) * 2020-11-30 2021-03-26 中国人民解放军国防科技大学 Network congestion control method combining active and passive modes
CN112602294A (en) * 2018-08-29 2021-04-02 华为技术有限公司 Method and device for detecting bandwidth
CN112615764A (en) * 2021-01-04 2021-04-06 重庆邮电大学 Air-ground broadband link network bandwidth measuring method based on active and passive combination
CN113032413A (en) * 2021-03-10 2021-06-25 北京嘀嘀无限科技发展有限公司 Data sampling method, device, electronic equipment, storage medium and program product
CN113852496A (en) * 2021-09-13 2021-12-28 电子科技大学 High-precision network bandwidth measuring system capable of diagnosing position of tight link
CN114172830A (en) * 2021-10-28 2022-03-11 阿里巴巴（中国）有限公司 Network delay detection method, device and system

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100407662C (en) * 2003-01-16 2008-07-30 华为技术有限公司 Network bandwidth detecting method for implementing stream medium service
FR2852762B1 (en) * 2003-03-19 2005-06-17 Acterna Ipms METHOD FOR EVALUATING THE BANDWIDTH OF A DIGITAL LINK
CN100391195C (en) * 2003-12-15 2008-05-28 上海贝尔阿尔卡特股份有限公司 Method for dynamically adjusting data packet length based on network service quality

Cited By (37)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101217497B (en) * 2008-01-18 2011-04-13 北京邮电大学 A path selecting method of wireless mesh network
CN101729306B (en) * 2008-10-28 2012-05-23 中兴通讯股份有限公司 Method and system for testing downlink broadband of access line of broadband network user
CN102053217A (en) * 2009-11-05 2011-05-11 上海华虹Nec电子有限公司 Method for quickly processing wafer after sort interruption under working condition
CN102347933A (en) * 2010-07-27 2012-02-08 财团法人交大思源基金会 Network traffic capturing, recovering and replaying method
CN101917339A (en) * 2010-08-17 2010-12-15 建汉科技股份有限公司 Network bandwidth upper limit counting and adjusting method of network equipment
WO2013091180A1 (en) * 2011-12-21 2013-06-27 华为技术有限公司 Method, device and system for determining initial file transmission speed
CN103339909A (en) * 2011-12-21 2013-10-02 华为技术有限公司 Method, device and system for determining initial file transmission speed
CN102710464B (en) * 2012-06-06 2015-04-15 清华大学 Fine-grained single-flow delay detection method based on reference delay insertion structure
CN102710464A (en) * 2012-06-06 2012-10-03 清华大学 Fine-grained single-flow delay detection method based on reference delay insertion structure
CN102946335A (en) * 2012-12-11 2013-02-27 广州中国科学院软件应用技术研究所 Network condition detection method and network condition detection system
CN102946335B (en) * 2012-12-11 2015-12-23 广州中国科学院软件应用技术研究所 A kind of network condition detection method and system
CN103391253A (en) * 2013-08-05 2013-11-13 四川启程科技发展有限公司 Method, device and system for controlling network congestion
CN105024875A (en) * 2014-04-30 2015-11-04 中国电信股份有限公司 Method and system for measuring network flow rate of broadband classified users
CN105024875B (en) * 2014-04-30 2018-06-19 中国电信股份有限公司 The network flow-speed measurement method and system of broadband sorted users
CN105591835B (en) * 2015-08-12 2019-04-09 新华三技术有限公司 A kind of detection method and device of link-quality
CN105591835A (en) * 2015-08-12 2016-05-18 杭州华三通信技术有限公司 Link quality detection method and device
CN106685565A (en) * 2016-12-30 2017-05-17 国网浙江省电力公司绍兴供电公司 Method and system for achieving transparent clock transmission of communication network
CN106850294A (en) * 2017-01-24 2017-06-13 钟炎培 The methods, devices and systems of bandwidth estimation
CN106850294B (en) * 2017-01-24 2019-07-09 西安万像电子科技有限公司 The methods, devices and systems of bandwidth estimation
CN108512726B (en) * 2018-03-29 2020-06-02 上海连尚网络科技有限公司 Data monitoring method and equipment
CN108512726A (en) * 2018-03-29 2018-09-07 上海连尚网络科技有限公司 A kind of method and apparatus of data monitoring
CN112602294B (en) * 2018-08-29 2021-10-15 华为技术有限公司 Method and device for detecting bandwidth
CN112602294A (en) * 2018-08-29 2021-04-02 华为技术有限公司 Method and device for detecting bandwidth
US11388090B2 (en) 2018-08-29 2022-07-12 Huawei Technologies Co., Ltd. Bandwidth measurement method and measurement device
CN109639684A (en) * 2018-12-17 2019-04-16 北京微吼时代科技有限公司 The method and system of dynamic application bandwidth
CN110138590A (en) * 2019-04-09 2019-08-16 上海交通大学 A kind of general TCP performance diagnogtics method and device
CN111130923A (en) * 2019-11-29 2020-05-08 北京达佳互联信息技术有限公司 Network bandwidth determining method and device, electronic equipment and storage medium
CN111162964A (en) * 2019-12-17 2020-05-15 国网智能科技股份有限公司 Intelligent station message integrity analysis method and system
CN111162964B (en) * 2019-12-17 2021-10-12 山东鲁软数字科技有限公司智慧能源分公司 Intelligent station message integrity analysis method and system
CN112565100B (en) * 2020-11-30 2022-05-20 中国人民解放军国防科技大学 Network congestion control method combining active and passive modes
CN112565100A (en) * 2020-11-30 2021-03-26 中国人民解放军国防科技大学 Network congestion control method combining active and passive modes
CN112615764A (en) * 2021-01-04 2021-04-06 重庆邮电大学 Air-ground broadband link network bandwidth measuring method based on active and passive combination
CN113032413A (en) * 2021-03-10 2021-06-25 北京嘀嘀无限科技发展有限公司 Data sampling method, device, electronic equipment, storage medium and program product
CN113852496A (en) * 2021-09-13 2021-12-28 电子科技大学 High-precision network bandwidth measuring system capable of diagnosing position of tight link
CN113852496B (en) * 2021-09-13 2023-07-25 电子科技大学 High-precision network bandwidth measurement system capable of diagnosing tight link position
CN114172830A (en) * 2021-10-28 2022-03-11 阿里巴巴（中国）有限公司 Network delay detection method, device and system
CN114172830B (en) * 2021-10-28 2023-12-12 阿里巴巴（中国）有限公司 Network delay detection method, device and system

Also Published As

Publication number Publication date
CN100442730C (en) 2008-12-10

Similar Documents

Publication Publication Date Title
CN100442730C (en) Detecting method for network link band width based on hop-by-hop time stamp label
US9246822B2 (en) Methods and apparatus for using a layered gear to analyze and manage real-time network quality of service transmission for mobile devices on public networks
CN101272290B (en) Measuring method, measuring apparatus for route congestion state in IP network
US20070217448A1 (en) Estimating Available Bandwidth With Multiple Overloading Streams
US20020169880A1 (en) Method and device for robust real-time estimation of the bottleneck bandwidth in the internet
CN1710888A (en) Method of generating a monitoring datagram
RU2695093C2 (en) Method for performing communication bandwidth testing from first network station to second network station in communication network, corresponding devices for executing steps of method and corresponding computer programs
CN1812350A (en) One-way delay measuring method
CN103152285A (en) Local terminal network element, opposite terminal network element and sending window adjustment method
Gál et al. Performance evaluation of massively parallel and high speed connectionless vs. connection oriented communication sessions
Siekkinen et al. A root cause analysis toolkit for TCP
KR20110034895A (en) Network available bandwidth high-speed measurement method and transmission delay trend measurement apparatus
CN108494629B (en) Bag sending device and method for OWAMP and TWAMP
KR20090082773A (en) The per link available bandwidth measurement method using the total length field in IP packet header and the available bandwidth information of a link management method
EP2719104B1 (en) Methods and apparatus for using a layered gear to analyze and manage real-time network quality of service transmission for mobile devices on public networks
JP2004312725A5 (en)
Jakimoski et al. Measurements of available bandwidth in computer networks
CN113923133B (en) Quality of experience index monitoring method for encrypted webpage traffic based on QUIC
WO2010063104A1 (en) Method and apparatus for measuring ip network performance characteristics
JP4305383B2 (en) End-to-end quality estimation apparatus and method
Besson Performance of TCP in a wide-area network: Influence of successive bottlenecks and exogenous traffic
Piratla et al. A comparative analysis of packet reordering metrics
Qian et al. Some observations on individual TCP flows behavior in network traffic traces
Ma et al. vNetRadar: Lightweight and Network-Wide Traffic Measurement in Virtual Networks
Chen et al. Accurate and Low-cost Scheme for Network Path Delay Measurement.

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
C17 Cessation of patent right
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20081210

Termination date: 20120511