CN106059951B - A kind of transfer control method for based on multilevel congestion feedback in DCN - Google Patents

A kind of transfer control method for based on multilevel congestion feedback in DCN Download PDF

Info

Publication number
CN106059951B
CN106059951B CN201610402552.4A CN201610402552A CN106059951B CN 106059951 B CN106059951 B CN 106059951B CN 201610402552 A CN201610402552 A CN 201610402552A CN 106059951 B CN106059951 B CN 106059951B
Authority
CN
China
Prior art keywords
congestion
queue
level
order
cwnd
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.)
Active
Application number
CN201610402552.4A
Other languages
Chinese (zh)
Other versions
CN106059951A (en
Inventor
黄家玮
徐文茜
王建新
郝晓强
张华健
有启迪
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.)
Central South University
Space Star Technology Co Ltd
Original Assignee
Central South University
Space Star Technology Co Ltd
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 Central South University, Space Star Technology Co Ltd filed Critical Central South University
Priority to CN201610402552.4A priority Critical patent/CN106059951B/en
Publication of CN106059951A publication Critical patent/CN106059951A/en
Application granted granted Critical
Publication of CN106059951B publication Critical patent/CN106059951B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic control in data switching networks
    • H04L47/10Flow control; Congestion control
    • H04L47/26Flow control; Congestion control using explicit feedback to the source, e.g. choke packets
    • H04L47/263Rate modification at the source after receiving feedback

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention discloses a kind of transfer control methods for based on multilevel congestion feedback in DCN, and interchanger length of buffer queue threshold value is arranged: marking thresholding Threshold1 and second level queue to mark thresholding Threshold2 including first order queue;Each exchange machine monitoring current queue size when multiple senders concomitantly send data, in transmission path;Compared with current queue size is marked thresholding Threshold1 and second level queue label thresholding Threshold2 with first order queue, current network congestion state is fed back;Alleviate the increasing window rate of slow turn-on according to first order congestion feedback;Slow start stage is exited in time according to second level congestion feedback and enters congestion avoidance phase.The present invention can effectively avoid DCN network concurrent transmit in due to slow start stage send window increase speed it is too fast caused by throughput collapse phenomenon, network throughput is substantially improved, promotes the overall performance of network application.

Description

A kind of transfer control method for based on multilevel congestion feedback in DCN
Technical field
The present invention designs a kind of for being based on multilevel congestion in DCN (data center network, Data CenterNetwork) The transfer control method of feedback.
Background technique
In recent years, more next with the rapid development of the network applications such as on-line search, social networks, e-commerce and universal More application on site systems are migrated in data center, are provided respectively using large-scale calculate with storage resource for user Kind network service.In face of huge user's amount of access, as Next Generation Internet calculating and storage core, data center Bearing capacity and service performance receive stern challenge.
Since data center network has the characteristic of ultra high bandwidth, ultralow time delay, this has very big with traditional wide area network Difference, so application effect of traditional TCP in data center network is very poor.High concurrent is same in data center network Step, which concurrently flows, easily causes certain paths instantaneously to become bottleneck, causes network congestion, data packetloss occurs and eventually lead to TCP flow Time-out.And transport stream is largely the low-volume traffic stream (90% is less than 100KB) of latency sensitive in data center network, and these Stream is likely to just complete to transmit in slow start stage.It is overtime caused by the transmission window rapid growth of TCP flow slow start stage Hangover, (90% is less than the time-out time (default minimum 200ms) of conventional TCP protocols with data center network round trip transmission delay Factor 1ms) etc. is mismatched, leads to throughput collapse phenomenon occur, TCP Incast phenomenon occurs, greatly reduce in data The communication capacity of heart network.
As Internet application becomes high diversity and complication, in order to solve the deficiency of Traditional control transport protocol, Alleviate the TCP Incast phenomenon in data center network, many scholars at home and abroad are improving TCP to network data congestion control Discovery and control ability in terms of done a large amount of improvement Journal of Sex Research, seek some stable methods allow Number dynamics change net Network user's energy justice shares limited network bandwidth resources.
ICTCP (Incast Congestion Control TCP) is primarily upon link throughput caused by avoiding high concurrent Rate collapses generated packet drop, and ICTCP only modifies recipient.Time is divided into the time with 2RTT (two-way time) for interval Slot carries out the assessment of available bandwidth in first RTT, calculates adjustment quota according to available bandwidth assessment result in second RTT To be adjusted to reception window.But the accuracy of RTT calculating directly affects the performance of ICTCP.
DCTCP (Data Center TCP) agreement is proposed aiming at the data center network of high bandwidth, the application of high fan-in Replacement conventional TCP protocols improvement transport protocol.It utilizes ecn (explicit congestion notification) (Explicit Congestion Notification, ECN) judge congestion state.DCTCP discusses tail portion delay comprehensively for the first time, at a low speed deposits being changed without Under the premise of commercial switches, high burst tolerance, the low latency, high-throughput of DCN (data center network) are realized.But When concurrency is smaller, interchanger caching still will appear part stream congestion packet loss and time-out occurs DCTCP agreement.And DCTCP Stream is under minimum transmission window, it is easier to the entire packet loss for sending window occur and cause RTO (Retransmission timeout time) overtime.
D2Transmission Control Protocol, using gamma correction function, improves the drop window mode of DCTCP on the basis of DCTCP, And the deferred constraint requirement of every data stream is considered to carry out reasonably dropping window during improved.D2TCP is not only increased The quantity for the data flow completed in deferred constraint, moreover it is possible to be coexisted well with conventional TCP, but due to the sensitive journey of deferred constraint Degree is insufficient, so there are also very big rooms for promotion for performance.
Therefore, data are allowed in order to avoid to increase window too fast and cause throughput crash issue for slow start stage in concurrent transmission, Central site network application can efficiently use available bandwidth, become a urgent problem to be solved.
Summary of the invention
Technical problem to be solved by the invention is to provide a kind of transmission for based on multilevel congestion feedback in DCN to control Method, data flow in data center network can be effectively relieved in this method there is hangover and surpasses in slow turn-on since increasing window is too fast When, to avoid the phenomenon that generating throughput collapse, network throughput is substantially improved, promotes the overall performance of network application.
The technical solution of invention is as follows:
A kind of transfer control method for based on multilevel congestion feedback in DCN,
Interchanger length of buffer queue threshold value is set: marking thresholding Threshold1 and second level team including first order queue Column label thresholding Threshold2;
Each exchange machine monitoring current queue size when multiple senders concomitantly send data, in transmission path;It will Compared with current queue size marks thresholding Threshold1 and second level queue label thresholding Threshold2 with first order queue, Feed back current network congestion state;
According to the comparison result of current queue size and first order queue label thresholding Threshold1, i.e. first order congestion Feedback alleviates the increasing window rate of slow turn-on;
According to the comparison result of current queue size and second level queue label thresholding Threshold2, i.e. second level congestion Feedback exits slow start stage in time and enters congestion avoidance phase.
The above-mentioned transfer control method for based on multilevel congestion feedback in DCN, specifically includes the following steps:
Step 1: initialization: it is long that setting sender sends the initial value of window cwnd size, setting machine monitoring current queue It spends threshold value: marking thresholding Threshold1 and second level queue label thresholding Threshold2, setting number including first order queue According to the initial value of packet first order congestion flag position ECT and second level congestion flag position CE, setting sender's slow turn-on flag bit ST Initial value be 0;
Step 2: each exchange machine monitoring current queue size in transmission path, when current queue size is more than the first order When queue marks thresholding Threshold1, the first order congestion flag position ECT of the data packet in current queue is set to 1 immediately; When current queue size is more than second level queue label thresholding Threshold2, then by second of the data packet in current queue Grade congestion flag position CE is set to 1;After data packet reaches recipient, recipient stamps the confirmation packet ACK of data packet corresponding Congestion flag, and will confirm that packet ACK is sent to sender;
After sender receives whole confirmation packet ACK in currently transmitted window, according to the first order in all confirmation packet ACK The quantity that congestion flag position ECT and second level congestion flag position CE is set to 1 calculates the first order and the second level Congestion Level SPCC α and β;
Step 3: sender is a flag bit of transmitting terminal according to its slow turn-on flag bit ST[ST] judge whether to continue In slow start stage, if ST is 0, keeps slow start stage and go to step four, otherwise exit slow start stage, into Enter congestion avoidance phase, and goes to step six;
Step 4: judge all confirmation packet ACK in currently transmitted window first order congestion flag position ECT whether be all 0, if it is, being updated to 2 × cwnd for window cwnd is sent, and it is transferred to step 8;Otherwise, five are gone to step;
Step 5: judge all confirmation packet ACK in currently transmitted window second level congestion flag position CE whether be all 0, if it is, being updated to cwnd+cwnd × (1-α) for window cwnd is sent3, it is transferred to step 8;Otherwise, slow turn-on mark is set Position ST is 1, is transferred to step 8;
Step 6: judge all confirmation packet ACK in currently transmitted window second level congestion flag position CE whether be all 0, if it is, being updated to cwnd+1 for window cwnd is sent, it is transferred to step 8;Otherwise, it is transferred to step 7;
Step 7: adjusting according to network congestion degree and send window size, will send window cwnd and is updated to cwnd × (1- β/2), it is transferred to step 8;
Step 8: transmitting terminal sends data by updated transmission window cwnd size, and judges whether data have sent Finish, return step two if not being sent, otherwise terminates data transmission.
The above-mentioned transfer control method for based on multilevel congestion feedback in DCN, the step 1 include: to send initial Window cwnd is dimensioned to 1, and first order queue label thresholding Threshold1 is set as 20, and second level queue marks thresholding Threshold2 is set as 65, and first order congestion flag position ECT and second level congestion flag position CE are both configured to 0;
The above-mentioned transfer control method for based on multilevel congestion feedback in DCN, in the step 2: sender, which receives, to be worked as After whole confirmation packet ACK in preceding transmission window, according to first order congestion flag position ECT and second in all confirmation packet ACK The quantity that grade congestion flag position CE is set to 1 calculates the first order and the second level Congestion Level SPCC α and β, calculation formula are as follows:
α=m/cwnd
β=n/cwnd
Wherein, m and n is that first order congestion flag position ECT and second in ACK is wrapped in all confirmations in currently transmitted window respectively Grade congestion CE is set to 1 quantity.
The utility model has the advantages that
The technical effects of the invention are that: when multiple senders concomitantly send data, the buffer queue of interchanger is long The successive thresholding Threshold1 and Threshold2 passed through twice is spent to feed back current network congestion state.Thresholding Threshold1 is used to alleviate the rapid growth that window is sent in slow start stage, prevents TCP flow overtime, so as to gentle transition To congestion avoidance phase, and thresholding Threshold2 then allows TCP flow to enter congestion avoidance phase, to reduce more numbers concurrently flowed According to the deadline of stream, network throughput is promoted.
Actual measurement shows that this method can effectively solve throughput collapse phenomenon obvious, detailed in Example.
Detailed description of the invention
Fig. 1 is flow chart of the invention.
Fig. 2 is data center TCP Incast schematic diagram of a scenario.
Fig. 3 is concurrent overtime test result comparison, and Fig. 3 (a) is the throughput of lower 6 stream of DCTCP, and Fig. 3 (b) is the present invention (GS-DCTCP) throughput of 6 stream when.
Fig. 4 is phase between DCTCP agreement and (GS-DCTCP) of the invention in the case that 25 working machines concurrently send data Reference index comparison is closed, wherein service request cell S RU size is 512Kbytes;Fig. 4 (a) window between different agreement becomes Change situation comparison;Fig. 4 (b) TCP packet serial number situation of change between different agreement compares;Fig. 4 (c) is exchanged between different agreement The comparison of machine queue length situation of change;Fig. 4 (d) total throughput situation of change comparison between different agreement.
Fig. 5 is that service request cell size is 64Kbytes, and different agreement is handled up as transmitting terminal is concurrently increased number of Rate comparison.
Fig. 6 is that service request unit total size is 2Mbytes, and different agreement is handled up as transmitting terminal is concurrently increased number of Rate comparison.
Specific embodiment
The present invention is described in further details below with reference to the drawings and specific embodiments:
Embodiment 1:
Referring to Fig. 1, Fig. 1 is flow chart of the invention: a kind of transmission controlling party for based on multilevel congestion feedback in DCN Method, process are as follows:
Initialization;The initial window cwnd that sends is dimensioned to 1, first order queue marks thresholding Threshold1 setting It is 20, second level queue label thresholding Threshold2 is set as 65, congestion flag position ECT and CE is both configured to 0, slow turn-on The mark ST in stage is set as 0;
After sender receives whole confirmation packet ACK in currently transmitted window, according to the first order in all confirmation packet ACK The quantity that congestion flag position ECT and second level congestion marking congestion flag position CE is set to 1 calculates the first order and second level congestion Degree α and β;
α=m/cwnd
β=n/cwnd
Wherein, m and n is that first order congestion flag position ECT and second in ACK is wrapped in all confirmations in currently transmitted window respectively Grade congestion CE is set to 1 quantity.
Transmitting terminal is that ST judges whether to keep slow start stage according to its slow turn-on mark:
When ST is 0, then slow start stage is kept, if the first order of ACK is wrapped in all confirmations in currently transmitted window Congestion flag position ECT is 0, then will send window cwnd and be updated to 2 × cwnd;And all confirmation packets in currently transmitted window When the first order congestion flag position ECT of ACK is not all 0, if the second level congestion mark of ACK is wrapped in all confirmations in currently transmitted window Will position CE is all 0, then will send window cwnd and be updated to cwnd+cwnd × (1-α)3;If all true in currently transmitted window Second level congestion flag position CE for recognizing packet ACK is not all 0, and setting slow turn-on flag bit ST is 1;
When ST is 1, slow start stage is exited, into congestion avoidance phase, if all confirmation packets in currently transmitted window Second level congestion flag position CE of ACK is all 0, then will send window cwnd and be updated to cwnd+1;Otherwise, according to network congestion Degree adjustment sends window size, will send window cwnd and is updated to cwnd × (1-β/2);
Transmitting terminal sends data by window cwnd size is sent after updating, and judges whether data flow is sent, if It is not sent, repeats the above process then data flow can be re-introduced into the step of updating network congestion degree, otherwise tie Beam data transmission.
To verify effect of the invention, performance test has been carried out on NS2.35 network simulation platform.
Fig. 2 illustrates the canonical topology that TCP Incast occurs.In data-center applications, in order to guarantee the high extension of service Property and reliability, be usually that cutting is stored in different servers using data, the data slot of each server storage is referred to as Server request unit (Server Request Unit, SRU).In general, being led in the following way inside data center Letter: convergence server sends request of data to all working server.After each workspace server receives request of data, it will transmit The data slot stored, i.e. SRU.After convergence server receives the SRU of all requests, data slot is merged or is handled, Then next round request of data is issued.
Fig. 3 is concurrent time-out test, in order to compare DCTCP and the method for the present invention (GS-DCTCP) smaller concurrent The case where lower appearance time-out phenomenon.Experimental situation is as follows: 7 working machines are linked into an interchanger via the link of 10Gbps, Wherein 6 working machines send data to the 7th working machine simultaneously.6 data stream sizes are 512KB, and interchanger caching is 100pkts, link delay are 100 μ s.For DCTCP agreement, the value of switch queue label thresholding K is set as 65.For this hair Bright, the value of interchanger multi-stage signature thresholding Threshold1 and Threshold2 are set to 20 and 65.
Find out from Fig. 3 (a) and (b), DCTCP just has already appeared showing for throughput collapse when workspace server quantity is 6 As the rapid growth due to sending window in slow start stage makes throughput quite high, but too swift and violent due to increasing, and causes There is the phenomenon of time-out not in time in Congestion Avoidance.And GS-DCTCP of the present invention, by level threshold tagging mechanism, alleviation is opened slowly The excessively rapid growth of window is sent in the dynamic stage, so that keeping avoiding overtime phenomenon in the higher situation of throughput, is prevented because of chain Road time delay mismatches with RTO re-transmission time and the case where application performance sharply declines occurs.
Fig. 4 is in the case that 25 working machines concurrently send data, and coherent reference index compares between different agreement, wherein Service request cell S RU size is 512Kbytes.
Fig. 4 (a) it can be seen that in DCTCP agreement, all lose by the entire all data packets sent in window, waits 200ms After retransmit, link utilization is low.GS-DCTCP agreement of the present invention then alleviates window excessively rapid growth in slow start stage in time, avoids Whole window Loss.Fig. 4 (b) display DCTCP agreement packet loss, which is crossed, often has time-out, and TCP packet serial number is interrupted.The present invention can then use up can It can be reduced packet loss, avoid time-out, effect is obvious.Fig. 4 (c) shows that interchanger is lined up situation, and wherein DCTCP agreement is led because of time-out Causing queue is sky, causes interchanger caching waste.The present invention then can effectively avoid cache overflow, guarantee link utilization.Fig. 4 It (d) is instantaneous total throughput situation, DCTCP agreement throughput collapse occurs since increasing window is too fast after taking bandwidth, time-out Re-transmission causes link bandwidth waste serious.And the present invention then slows down increasing window in due course, by slow start stage it is smooth be transitioned into congestion The stage is avoided, link bandwidth is made full use of, performance boost is obvious.
Fig. 5 is that service request cell size is 64Kbytes, and different agreement is handled up as transmitting terminal is concurrently increased number of The comparison of rate.Since overtime phenomenon occurs in DCTCP agreement, so throughput is very low, link utilization is low;And GS- of the present invention DCTCP is promoted obvious due to effectively avoiding hangover time-out, throughput.Fig. 6 is that service request unit total size is 2Mbytes, different agreement is as concurrently increased number of throughput compares transmitting terminal.It can be seen that the present invention is assisted with respect to DCTCP It is more obvious that the throughput of view promotes effect.

Claims (3)

1. a kind of transfer control method for based on multilevel congestion feedback in DCN, which is characterized in that
Interchanger length of buffer queue threshold value is set: marking thresholding Threshold1 and second level queue mark including first order queue Remember thresholding Threshold2;
Each exchange machine monitoring current queue size when multiple senders concomitantly send data, in transmission path;It will be current Compared with queue length marks thresholding Threshold1 and second level queue label thresholding Threshold2 with first order queue, feedback Current network congestion state;
According to the comparison result of current queue size and first order queue label thresholding Threshold1, i.e. first order congestion feedback Alleviate the increasing window rate of slow turn-on;
According to the comparison result of current queue size and second level queue label thresholding Threshold2, i.e. second level congestion feedback Slow start stage is exited in time and enters congestion avoidance phase;
Specifically includes the following steps:
Step 1: initialization: setting sender sends the initial value of window cwnd size, setting interchanger length of buffer queue threshold Value: thresholding Threshold1 and second level queue label thresholding Threshold2, setting data packet are marked including first order queue The first of sender's slow turn-on flag bit ST is arranged in the initial value of first order congestion flag position ECT and second level congestion flag position CE Initial value is 0;
Step 2: each exchange machine monitoring current queue size in transmission path, when current queue size is more than first order queue When marking thresholding Threshold1, the first order congestion flag position ECT of the data packet in current queue is set to 1 immediately;When working as When preceding queue length is more than second level queue label thresholding Threshold2, then the second level of the data packet in current queue is gathered around Plug flag bit CE is set to 1;After data packet reaches recipient, recipient stamps corresponding congestion to the confirmation packet ACK of data packet Mark, and will confirm that packet ACK is sent to sender;
After sender receives whole confirmation packet ACK in currently transmitted window, according to first order congestion in all confirmation packet ACK The quantity that flag bit ECT and second level congestion flag position CE is set to 1 calculates the first order and the second level Congestion Level SPCC α and β;
Step 3: sender judges whether to keep slow start stage according to its slow turn-on flag bit ST, if ST is 0, It keeps slow start stage and goes to step four, otherwise exit slow start stage, into congestion avoidance phase, and go to step six;
Step 4: judge the first order congestion flag position ECT of all confirmation packet ACK in currently transmitted window whether all for 0, such as Fruit is will then to send window cwnd and be updated to 2 × cwnd, and be transferred to step 8;Otherwise, five are gone to step;
Step 5: judge the second level congestion flag position CE of all confirmation packet ACK in currently transmitted window whether all for 0, such as Fruit is will then to send window cwnd and be updated to cwnd+cwnd × (1-α)3, it is transferred to step 8;Otherwise, slow turn-on flag bit ST is set It is 1, is transferred to step 8;
Step 6: judge the second level congestion flag position CE of all confirmation packet ACK in currently transmitted window whether all for 0, such as Fruit is will then to send window cwnd and be updated to cwnd+1, be transferred to step 8;Otherwise, it is transferred to step 7;
Step 7: according to network congestion degree adjust send window size, will send window cwnd be updated to cwnd × (1-β/ 2), it is transferred to step 8;
Step 8: transmitting terminal sends data by updated transmission window cwnd size, and judges whether data are sent, such as Fruit is not sent then return step two, otherwise terminates data transmission.
2. the transfer control method according to claim 1 for based on multilevel congestion feedback in DCN, which is characterized in that The step 1 includes: that the initial window cwnd that sends is dimensioned to 1, and first order queue marks thresholding Threshold1 setting It is 20, second level queue label thresholding Threshold2 is set as 65, by first order congestion flag position ECT and second level congestion mark Will position CE is both configured to 0.
3. the transfer control method according to claim 1 for based on multilevel congestion feedback in DCN, which is characterized in that In the step 2:
After sender receives whole confirmation packet ACK in currently transmitted window, according to first order congestion in all confirmation packet ACK The quantity that flag bit ECT and second level congestion flag position CE is set to 1 calculates the first order and the second level Congestion Level SPCC α and β, calculates Formula are as follows:
α=m/cwnd
β=n/cwnd
Wherein, m and n is that first order congestion flag position ECT and the second level are gathered around in all confirmation packet ACK in currently transmitted window respectively Plug CE is set to 1 quantity.
CN201610402552.4A 2016-06-08 2016-06-08 A kind of transfer control method for based on multilevel congestion feedback in DCN Active CN106059951B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610402552.4A CN106059951B (en) 2016-06-08 2016-06-08 A kind of transfer control method for based on multilevel congestion feedback in DCN

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610402552.4A CN106059951B (en) 2016-06-08 2016-06-08 A kind of transfer control method for based on multilevel congestion feedback in DCN

Publications (2)

Publication Number Publication Date
CN106059951A CN106059951A (en) 2016-10-26
CN106059951B true CN106059951B (en) 2019-03-01

Family

ID=57170634

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610402552.4A Active CN106059951B (en) 2016-06-08 2016-06-08 A kind of transfer control method for based on multilevel congestion feedback in DCN

Country Status (1)

Country Link
CN (1) CN106059951B (en)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107046507B (en) * 2016-12-13 2019-05-28 中南大学 A kind of jamming control method for multiplexed transport in DCN
CN108259383B (en) * 2016-12-29 2021-10-01 北京华为数字技术有限公司 Data transmission method and network equipment
CN107770082B (en) * 2017-10-19 2020-05-12 中南大学 Transmission control method based on task flow characteristics in data center network
CN108540395B (en) * 2018-04-18 2020-07-28 清华大学 Congestion judgment method in loss-free network and switch
CN110661723B (en) 2018-06-29 2023-08-22 华为技术有限公司 Data transmission method, computing device, network device and data transmission system
CN108965151B (en) * 2018-08-27 2020-08-28 华中科技大学 Explicit congestion control method based on queuing time delay
CN110098893B (en) * 2018-12-21 2021-09-10 中南大学 Differential explicit feedback transmission control method facing IPv6
CN112054965A (en) * 2019-06-05 2020-12-08 阿里巴巴集团控股有限公司 Congestion control method, equipment and computer readable medium
CN113746743B (en) * 2020-05-29 2023-08-08 华为技术有限公司 Data message transmission method and device
CN113194040A (en) * 2021-04-28 2021-07-30 王程 Intelligent control method for instantaneous high-concurrency server thread pool congestion
CN117240788B (en) * 2023-11-15 2024-03-08 南京邮电大学 SDN-based data center network congestion control method

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1324862C (en) * 2004-01-06 2007-07-04 北京邮电大学 Method for regulating congest window in communication network
CN101188473B (en) * 2007-12-20 2010-06-09 上海交通大学 Self-adapted adjustment method for transmission rate of transport layer in wireless self-organized network
US20130308458A1 (en) * 2012-05-18 2013-11-21 Alcatel-Lucent Usa Inc. System And Method For Implementing Active Queue Management Enhancements For Variable Bottleneck Rates
CN103457871B (en) * 2013-09-18 2016-03-30 中南大学 Based on the increasing window method of the congestion avoidance phase of deferred constraint in DCN
CN104009931B (en) * 2014-06-13 2017-07-14 中南大学 The jamming control method cooperateed with data center network based on advertised window and multithread

Also Published As

Publication number Publication date
CN106059951A (en) 2016-10-26

Similar Documents

Publication Publication Date Title
CN106059951B (en) A kind of transfer control method for based on multilevel congestion feedback in DCN
CN106533970B (en) Difference towards cloud computing data center network spreads transmission control method and device
CN103457871B (en) Based on the increasing window method of the congestion avoidance phase of deferred constraint in DCN
CN102546106B (en) Multipath parallel transmission method facing satellite network
CN108881008A (en) A kind of methods, devices and systems of data transmission
CN109412958A (en) The jamming control method and device of data center
CN106302228B (en) The transfer control method of task based access control perception in a kind of data center network
CN101860488B (en) Network congestion control method
CN104092625B (en) A kind of self adaptation being used in DCN asks dispatching method in batches
CN108270691A (en) A kind of alternative ecn (explicit congestion notification) labeling method for data center
CN103067301A (en) Fast and reliable congestion control improved algorithm based on user datagram protocol (UDP)
Zhang et al. Modeling and solving TCP incast problem in data center networks
CN107566275B (en) Multi-path transmission method based on the delay inequality opposite sex in data center network
CN110191065A (en) High-performance supported balanced System and method for based on software defined network
CN109873773B (en) Congestion control method for data center
CN107948103A (en) A kind of interchanger PFC control methods and control system based on prediction
CN101146032A (en) A media streaming transmission bandwidth self-adapted method
CN112995048B (en) Blocking control and scheduling fusion method of data center network and terminal equipment
CN104320809A (en) Wireless multi-hop network congestion control method and system based on RTT
Shukla et al. TCP PLATO: Packet labelling to alleviate time-out
CN103297346B (en) TCP (Transmission Control Protocol) friendly rate control method based on ECN (Explicit Congestion Notification) mechanism
CN102946361A (en) Method and system of flow control based on exchanger cache allocation
CN107154897A (en) Isomery stream partition method based on bag scattering in DCN
CN105207944B (en) Transfer control method based on FAST TCP
CN107046507B (en) A kind of jamming control method for multiplexed transport in DCN

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant