CN110138590A - A kind of general TCP performance diagnogtics method and device - Google Patents
A kind of general TCP performance diagnogtics method and device Download PDFInfo
- Publication number
- CN110138590A CN110138590A CN201910281276.4A CN201910281276A CN110138590A CN 110138590 A CN110138590 A CN 110138590A CN 201910281276 A CN201910281276 A CN 201910281276A CN 110138590 A CN110138590 A CN 110138590A
- Authority
- CN
- China
- Prior art keywords
- tcp
- propagation delay
- message
- window
- bottleneck bandwidth
- 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.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/12—Arrangements for detecting or preventing errors in the information received by using return channel
- H04L1/16—Arrangements for detecting or preventing errors in the information received by using return channel in which the return channel carries supervisory signals, e.g. repetition request signals
- H04L1/1607—Details of the supervisory signal
- H04L1/1657—Implicit acknowledgement of correct or incorrect reception, e.g. with a moving window
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0677—Localisation of faults
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
- H04L43/0852—Delays
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
- H04L43/0876—Network utilisation, e.g. volume of load or congestion level
- H04L43/0894—Packet rate
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/10—Active monitoring, e.g. heartbeat, ping or trace-route
- H04L43/106—Active monitoring, e.g. heartbeat, ping or trace-route using time related information in packets, e.g. by adding timestamps
Abstract
The present invention relates to a kind of general TCP performance diagnogtics method and devices, comprising: crawl TCP message simultaneously copies kernel buffers to；The message stored in kernel buffers is reassembled as User space by kernel state to restore to obtain each TCP flow；Propagation delay is calculated according to the timing information of TCP flow；TCP flow is subjected to granularity division, obtains granularity division result and corresponding filter length of window；Bottleneck bandwidth is calculated according to granularity division result and corresponding filter length of window；In conjunction with propagation delay and bottleneck bandwidth output performance diagnostic result.Compared with prior art, the present invention is suitable for various congestion avoidance algorithms, has the practicability of height and is easy to dispose, while can be with the multiple TCP connections of real-time diagnosis.
Description
Technical field
The present invention relates to network measures and network performance diagnostic field, more particularly, to a kind of general TCP performance diagnogtics
Method and device.
Background technique
With various new networks emerge in large numbers and the upgrading of network, the network service of enterprise can carry more and more
Number of users and flow.Once network failure occurs, enterprises must organize related personnel to carry out malfunction elimination and positioning,
And the process often consumes the excessive time, and enterprise and user is finally made to undertake unnecessary loss.One TCP connection by
Transmitting terminal, receiving end and nodes composition, three works relatively independently, and is all likely to become performance bottleneck,
Occur often to be difficult to determine when performance fault to be specifically which partially there is a problem.In addition, modern network application is often adopted
With the framework (multi-tier architecture) of layering, it is by user oriented front end (such as network agent and load
It is balanced) and IO or CPU intensive rear end (such as data base querying).This framework further increases the difficulty of fault location.
Where one automated network performance diagnogtics tool can point out performance bottleneck in time, thus operation maintenance personnel can be helped as early as possible
Repair failure.Existing tool takes two different schemes: the diagnosis of package level and the diagnosis for invading kernel.The former needs
Each TCP message is traversed, has very high difficulty in program realization；The latter needs to operate in the kernel protocol stack of server-side
In, the stability and performance of service can be reduced.
Summary of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide a kind of general TCP performances
Diagnostic tool.
The purpose of the present invention can be achieved through the following technical solutions:
A kind of general TCP performance diagnogtics method, comprising:
Crawl TCP message simultaneously copies kernel buffers to；
The message stored in kernel buffers is reassembled as User space by kernel state to restore to obtain each TCP flow；
Propagation delay is calculated according to the timing information of TCP flow；
TCP flow is subjected to granularity division, obtains granularity division result and corresponding filter length of window；
Bottleneck bandwidth is calculated according to granularity division result and corresponding filter length of window；
In conjunction with propagation delay and bottleneck bandwidth output performance diagnostic result.
It is described that propagation delay is calculated according to the timing information of TCP flow, comprising:
Timing information based on TCP flow obtains two-way time RTT；
Propagation delay is obtained according to two-way time RTT.
The propagation delay specifically:
Wherein, RTprop (T) is network propagation delay, WtFor we choose time window length, T be it is current when
It carves.
It is described that bottleneck bandwidth is calculated according to granularity division result and corresponding filter length of window, comprising:
The confirmation rate of transmission rate and receiving end based on transmitting terminal takes the two minimum value as message and submits rate；
Bottleneck bandwidth is obtained according to rate calculations are submitted.
The bottleneck bandwidth specifically:
Wherein, BtlBw (n) is the minimum value of the available bandwidth of all nodes in network link, WfFor selected by filter
Length of window, unit is flight, and n is the divided quantity of message.
A kind of general TCP performance diagnogtics device, comprising:
Packet capturing module, for grabbing TCP message and copying kernel buffers to；
TCP flow recombination module, for the message stored in kernel buffers to be reassembled as User space by kernel state to restore
Obtain each TCP flow；
Granularity division module obtains granularity division result and corresponding filter window for TCP flow to be carried out granularity division
Mouth length；
Network measure module, for propagation delay to be calculated according to the timing information of TCP flow, according to granularity division result
And corresponding filter length of window calculates bottleneck bandwidth；
Diagnosis algorithm module, for combining propagation delay and bottleneck bandwidth output performance diagnostic result.
The network measure module includes:
Two-way time acquiring unit obtains two-way time RTT for the timing information based on TCP flow；
Propagation delay acquiring unit, for obtaining propagation delay according to two-way time RTT.
The network measure module includes:
Rate acquiring unit is submitted, for comparing the transmission rate of transmitting terminal and the confirmation rate of receiving end, takes the two most
Small value submits rate as message；
Bottleneck bandwidth acquiring unit submits rate calculations for basis and obtains bottleneck bandwidth.
Compared with prior art, the invention has the following advantages:
1. the present invention has the versatility of height independently of the congestion avoidance algorithm of server-side and the realization details of application layer
And it is easy to dispose.
2. sharing the load for handling TCP message equally multiple CPU using the RSS network interface card with more queues as hardware support
In core, the message of the same TCP connection can be handled by the same core, improved the handling capacity of processing and reduced delay, together
When provide multithreading realize technical solution, being capable of the multiple TCP connections of real-time diagnosis.
3. the present invention can monitor the state change of TCP flow and the variation of network itself, real-time update diagnostic result makes user
It can check the latest development of network performance and the solution progress of failure.
4. the property of one TCP flow of positioning can be automated based on the network performance diagnosis algorithm of the adaptive granularity of flight
Energy bottleneck, shortens the time checked and positioned to failure, and operation maintenance personnel is helped quickly to repair failure.
Detailed description of the invention
Fig. 1 is whole design schematic diagram of the invention.
Fig. 2 is the design diagram of packet capturing module.
Specific embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.The present embodiment is with technical solution of the present invention
Premised on implemented, the detailed implementation method and specific operation process are given, but protection scope of the present invention is not limited to
Following embodiments.
As shown in Figure 1, the application is realized by computer system, executed by the processor of computer system, it is bottom-up
By five module groups such as packet capturing module, TCP flow recombination module, network measure module, granularity division module and diagnosis algorithm module
At.
Packet capturing module is located at the bottom of software systems, be responsible for efficiently in network interface card capture TCP message to User space for
Upper layer uses.Need to capture the timestamp of TCP header and capture, the precision of timestamp needs to be accurate to Millisecond or more.Such as
Shown in Fig. 2, packet capturing module uses the RSS network interface card with more queues as hardware support, is then come using PF_RING packet catcher
Realize the packet capturing of multithreading.RSS network interface card can share the load for handling TCP message on multiple core cpus equally, and same
The message of TCP connection is always handled by the same core, and this makes it possible to the performances for playing multi-core processor to improve processing
Handling capacity simultaneously reduces delay.PF_RING is responsible for copying TCP message in each queue into kernel buffers, and provides user
The access interface of state.
The major function of TCP flow recombination module is that the state of TCP connection is rebuild from bi-directional stream, and by the report of capture
Literary head is rearranged according to serial number, and stores necessary metamessage.Metamessage is mainly used to calculate RTT, transmission speed, connect
Speed and confirmation byte number etc. are received, bottleneck bandwidth and propagation delay are further measured.It is first in order to recombinate a TCP flow
Cryptographic Hash first is asked to the four-tuple of each message to be mapped to different TCP flows.Then from SYN message and SYN+ACK message
Obtain the value of the various options of information and TCP that connection is established；The information of connection closed is obtained from FIN message.In TCP flow
In rank, needs to safeguard and submitted byte number (delivered bytes) and the nearest time of delivery (delivered time).
" submitting " refers to the process of the data receiving end confirmation of transmitting terminal；" having submitted " indicates that transmitting terminal has received a confirmation report
Text, the message have submitted some data；" time of delivery " then refers to that transmitting terminal receives the time of effective confirmation message.Whenever receiving
One ACK message, has searched whether that message is identified inside the queue of data message first, while needing to consider the choosing of TCP
Confirmation (SACK) option is selected, which is able to confirm that discontinuous message.If there is message is identified, then adds up and submitted word
Joint number, the numerical value are equal to the payload size (payload size) of data message；It also needs to update the nearest time of delivery simultaneously, it should
Time value is equal to the timestamp of confirmation message.The timestamp of message is safeguarded by the kernel module of PF_RING, is not to deliver User space
When system time.
Granularity division module has used the adaptive dynamic patitioning algorithm based on flight, which can be according to network
The environment of itself and the state dynamic select time interval of TCP connection or message number carry out diagnosis performance bottleneck.Here
Flight is a unit for dividing message.Message is divided into multiple flight first by we, then according to certain matching
Rule finds the performance bottleneck of each flight.The size of flight itself be it is continually changing at any time, depend primarily on
Congestion avoidance algorithm and network environment itself.
Network measure module is responsible for measuring the bottleneck bandwidth and propagation delay of network link.Bottleneck bandwidth BtlBw
(bottleneck bandwidth) refers to that the minimum value of the available bandwidth of all nodes in network link, the value are a TCP
The upper limit of the connection institute's energy band to bandwidth.Measurement process designs three indexs:
(1) transmission rate (send rate).Transmission rate refers in the unit time, the byte number that transmitting terminal is sent, i.e.,
SendRate=Δ sendBytes/ Δ t.Transmission rate determines by the Transmission Control Protocol state (and application layer) of transmitting terminal, not
Directly by the constraint of bottleneck bandwidth, transmission rate can be greater than bottleneck bandwidth.
(2) confirm rate (ack rate).By the rate for the reliable transmission byte number that ACK message confirms.Whenever capturing
One ACK message, we can know byte number and the time of waiting of confirmation, ackRate can be obtained in conjunction with the two
=Δ ackedBytes/ Δ t.Unlike transmission rate, confirmation rate depends primarily on the state of network, and directly by
To the constraint of bottleneck bandwidth.
(3) rate (delivery rate) is submitted.Under normal circumstances, confirmation rate and submit rate be it is equal, still
The abnormal conditions that some confirmation velocity amplitude is much higher than available bandwidth are often encountered in actual network flow.In addition, it is contemplated that real
Confirm that rate is always less than transmission rate in the situation of border, in order to make numerical value more accurate, we take transmission rate and confirmation speed
The minimum value of both rates is used as and submits rate, i.e. deliveryRate=min (sendRate, ackRate).Rate is submitted to receive
The constraint of bottleneck bandwidth, it can therefore be concluded that deliveryRate≤BtlBw out.
Three above part it is the most key be the calculating for confirming rate, in conjunction with the existing information of TCP flow amount recombination module,
As soon as we calculate corresponding confirmation rate using following formula whenever receiving an ACK message:
Δ ackedBytes (ACK)=delivered-deliveredp
Δ t (ACK)=now-deliveredTimep
AckRate (ACK)=Δ ackedBytes (ACK)/Δ t (ACK)
Delivered is the byte number for flowing rank and submitting, delivered in above formulapFor datagram corresponding to ACK message
Text record submits byte number, and now is the time for capturing the ACK message, deliveredTimepFor corresponding to ACK message
The nearest time of delivery of data message record.We obtain a series of sampled values for submitting rate in this way.
Be then based on a TCP connection submits polydispersity index value to carry out the bottleneck of network link on time dimension
Estimation.Being arranged one has the maximum value filter (windowed maximum filter) of window to handle sampled value, and incites somebody to action
Estimated value of the output of filter as bottleneck bandwidth.Whenever receiving an ACK message, it is assumed that the message is divided into n-th
Flight is denoted as f (n), with following formula Calculation bottleneck bandwidth:
Wherein WfFor length of window selected by filter, unit is flight.
For network link corresponding to any one TCP connection, a network propagation delay RTprop is certainly existed
(round-trip propagation delay), the delay are the lower limits of network RTT (round-trip delay time),
That is RTprop≤RTT.Only when network is unimpeded does not have queueing delay, the RTprop=RTT of network.Based on time window
To estimate propagation delay:
W in formulatFor the length for the time window that we choose.As soon as we can calculate whenever capture ACK message
Corresponding RTT, and update propagation delay.Because RTT sampled value quantity is excessive, very waste memory source is all stored,
Therefore when realizing using simplified method, it may be assumed that as long as WtSmaller RTT sampled value is not observed in time, just with nearest
RTT sampled value substitute.
Finally based on the estimation to bottleneck bandwidth and propagation delay, and the dynamic of ACK message is divided, using voluntarily setting
The diagnosis algorithm of the heuristic TCP connection performance bottleneck of meter helps our positioning failures.The mode classification of the algorithm is one by one
Matching, and using the bottleneck of first successful match as the performance bottleneck of connection.Limiting factor is classified as by matching order: being opened slowly
It moves, receives window, network congestion, sends buffer area, application layer, congestion control, bottleneck bandwidth, Bufferbloat, such as 1 institute of table
Show.
Table 1
| Performance bottleneck | Diagnostic method |
| Slow turn-on | Continuous flight size is exponentially increased |
| Receive window | The smallest receiving window size is less than 3*MSS in one flight |
| Network congestion | It was found that retransmitting and there are RTT > α * RTprop, wherein α > 1 |
| Send buffer area | F < β * BtlBw*RTprop, wherein β ∈ (0,1) and continuous constant three times |
| Application layer | F < γ * BtlBw*RTprop, wherein (0,1) γ ∈, and have small messages |
| Congestion control | F < γ * BtlBw*RTprop, wherein (0,1) γ ∈ |
| Bottleneck bandwidth | At least half of delivery_rate > δ * BtlBw, wherein (0,1) δ ∈ |
| Bufferbloat | All RTT > ε * RTprop, wherein ε > > 1 |
Claims (10)
1. a kind of general TCP performance diagnogtics method characterized by comprising
Crawl TCP message simultaneously copies kernel buffers to；
The message stored in kernel buffers is reassembled as User space by kernel state to restore to obtain each TCP flow；
Propagation delay is calculated according to the timing information of TCP flow；
TCP flow is subjected to granularity division, obtains granularity division result and corresponding filter length of window；
Bottleneck bandwidth is calculated according to granularity division result and corresponding filter length of window；
In conjunction with propagation delay and bottleneck bandwidth output performance diagnostic result.
2. the general TCP performance diagnogtics method of one kind according to claim 1, which is characterized in that described according to TCP flow
Propagation delay is calculated in timing information, comprising:
Timing information based on TCP flow obtains two-way time RTT；
Propagation delay is obtained according to two-way time RTT.
3. the general TCP performance diagnogtics method of one kind according to claim 2, which is characterized in that the propagation delay tool
Body are as follows:
Wherein, RTprop (T) is network propagation delay, WtFor the length for the time window that we choose, T is current time.
4. the general TCP performance diagnogtics method of one kind according to claim 1, which is characterized in that described to be drawn according to granularity
Point result and corresponding filter length of window calculate bottleneck bandwidth, comprising:
The confirmation rate of transmission rate and receiving end based on transmitting terminal takes the two minimum value as message and submits rate；
Bottleneck bandwidth is obtained according to rate calculations are submitted.
5. the general TCP performance diagnogtics method of one kind according to claim 4, which is characterized in that the bottleneck bandwidth tool
Body are as follows:
Wherein, BtlBw (n) is the minimum value of the available bandwidth of all nodes in network link, WfFor window selected by filter
Length, unit are flight, and n is the divided quantity of message.
6. a kind of general TCP performance diagnogtics device characterized by comprising
Packet capturing module, for grabbing TCP message and copying kernel buffers to；
TCP flow recombination module, for the message stored in kernel buffers to be reassembled as User space by kernel state to restore to obtain
Each TCP flow；
Granularity division module obtains granularity division result and corresponding filter window is long for TCP flow to be carried out granularity division
Degree；
Network measure module, for propagation delay to be calculated according to the timing information of TCP flow, according to granularity division result and right
The filter length of window answered calculates bottleneck bandwidth；
Diagnosis algorithm module, for combining propagation delay and bottleneck bandwidth output performance diagnostic result.
7. the general TCP performance diagnogtics device of one kind according to claim 6, which is characterized in that the network measure mould
Block includes:
Two-way time acquiring unit obtains two-way time RTT for the timing information based on TCP flow；
Propagation delay acquiring unit, for obtaining propagation delay according to two-way time RTT.
8. the general TCP performance diagnogtics device of one kind according to claim 7, which is characterized in that the propagation delay tool
Body are as follows:
Wherein, RTprop (T) is network propagation delay, WtFor the length for the time window that we choose, T is current time.
9. the general TCP performance diagnogtics device of one kind according to claim 7, which is characterized in that the network measure mould
Block includes:
Rate acquiring unit is submitted, for comparing the transmission rate of transmitting terminal and the confirmation rate of receiving end, takes the two minimum value
Rate is submitted as message；
Bottleneck bandwidth acquiring unit submits rate calculations for basis and obtains bottleneck bandwidth.
10. the general TCP performance diagnogtics device of one kind according to claim 9, which is characterized in that the bottleneck bandwidth tool
Body are as follows:
Wherein, BtlBw (n) is the minimum value of the available bandwidth of all nodes in network link, WfFor window selected by filter
Length, unit are flight, and n is the divided quantity of message.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910281276.4A CN110138590A (en) | 2019-04-09 | 2019-04-09 | A kind of general TCP performance diagnogtics method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910281276.4A CN110138590A (en) | 2019-04-09 | 2019-04-09 | A kind of general TCP performance diagnogtics method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110138590A true CN110138590A (en) | 2019-08-16 |
Family
ID=67569439
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910281276.4A Pending CN110138590A (en) | 2019-04-09 | 2019-04-09 | A kind of general TCP performance diagnogtics method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110138590A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114598653A (en) * | 2022-05-09 | 2022-06-07 | 上海飞旗网络技术股份有限公司 | Data stream acceleration method based on time delay management model |
| CN115550217A (en) * | 2022-09-28 | 2022-12-30 | 浙江大学 | Network diagnosis method and device for seven-layer load balancing scene in cloud network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101051959A (en) * | 2007-05-11 | 2007-10-10 | 北京邮电大学 | Detecting method for network link band width based on hop-by-hop time stamp label |
| CN102014030A (en) * | 2010-12-31 | 2011-04-13 | 湖南神州祥网科技有限公司 | Network performance measurement diagnostic method and system based on transmission control protocol (TCP) |
| CN107734537A (en) * | 2016-08-10 | 2018-02-23 | 腾讯科技（深圳）有限公司 | Measuring method, transmitting terminal and the receiving terminal of network performance parameter |
| CN107800663A (en) * | 2016-08-31 | 2018-03-13 | 华为数字技术（苏州）有限公司 | The detection method and device of flow off-line files |
-
2019
- 2019-04-09 CN CN201910281276.4A patent/CN110138590A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101051959A (en) * | 2007-05-11 | 2007-10-10 | 北京邮电大学 | Detecting method for network link band width based on hop-by-hop time stamp label |
| CN102014030A (en) * | 2010-12-31 | 2011-04-13 | 湖南神州祥网科技有限公司 | Network performance measurement diagnostic method and system based on transmission control protocol (TCP) |
| CN107734537A (en) * | 2016-08-10 | 2018-02-23 | 腾讯科技（深圳）有限公司 | Measuring method, transmitting terminal and the receiving terminal of network performance parameter |
| CN107800663A (en) * | 2016-08-31 | 2018-03-13 | 华为数字技术（苏州）有限公司 | The detection method and device of flow off-line files |
Non-Patent Citations (2)
| Title |
|---|
| GUANG-QIAN PENG 等: ""Network Measurement and Performance Analysis at Server Side"", 《FUTURE INTERNET》 * |
| 蔡志平: "" 基于主动和被动测量的网络测量技术、模型和算法研究"", 《中国博士学位论文全文数据库信息科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114598653A (en) * | 2022-05-09 | 2022-06-07 | 上海飞旗网络技术股份有限公司 | Data stream acceleration method based on time delay management model |
| CN115550217A (en) * | 2022-09-28 | 2022-12-30 | 浙江大学 | Network diagnosis method and device for seven-layer load balancing scene in cloud network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mittal et al. | TIMELY: RTT-based congestion control for the datacenter | |
| DeLucia et al. | Multicast feedback suppression using representatives | |
| EP3661133B1 (en) | Delay-based transmission path control method, network controller and system | |
| US20080043716A1 (en) | Telemetry stream performance analysis and optimization | |
| CN103259696B (en) | Network bandwidth detection method, device and the network equipment | |
| US20050265235A1 (en) | Method, computer program product, and data processing system for improving transaction-oriented client-server application performance | |
| JP5673805B2 (en) | Network device, communication system, abnormal traffic detection method and program | |
| Chen et al. | Comprehensive understanding of TCP Incast problem | |
| CN115174432B (en) | RDMA network state monitoring method, device, equipment and readable storage medium | |
| CN110138590A (en) | A kind of general TCP performance diagnogtics method and device | |
| WO2018232917A1 (en) | Internet-of-things network congestion reduction method and router | |
| CN112311688A (en) | Rate update engine for reliable transport protocols | |
| CN103152278A (en) | Congestion determination method, congestion determination device and congestion determination network equipment | |
| CN109787861B (en) | Network data delay control method | |
| JP5039677B2 (en) | Edge node and bandwidth control method | |
| Feibish et al. | Flow-level loss detection with Δ-sketches | |
| Pereira et al. | Semantically reliable multicast protocols | |
| CN114095126A (en) | Method, system and device for adjusting data sending rate | |
| Zhang et al. | LearningCC: An online learning approach for congestion control | |
| JP2014112779A (en) | Data transmission controller, data transmission control method, and computer program | |
| US11863451B2 (en) | Hardware accelerated temporal congestion signals | |
| JP4766703B2 (en) | Edge node and bandwidth control method | |
| EP4266644A1 (en) | Network congestion control in sub-round trip time | |
| JP5537692B1 (en) | Quality degradation cause estimation device, quality degradation cause estimation method, quality degradation cause estimation program | |
| Hughes-Jones et al. | Performance measurements on gigabit ethernet nics and server quality motherboards |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 200030 Dongchuan Road, Minhang District, Minhang District, Shanghai Applicant after: Shanghai Jiaotong University Address before: 200030 Huashan Road, Shanghai, No. 1954, No. Applicant before: Shanghai Jiaotong University |
|
| CB02 | Change of applicant information | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190816 |
|
| RJ01 | Rejection of invention patent application after publication |