CN114553783B - Load balancing method for self-adaptive regulation of cell granularity of data center network - Google Patents

Load balancing method for self-adaptive regulation of cell granularity of data center network Download PDF

Info

Publication number
CN114553783B
CN114553783B CN202210169362.8A CN202210169362A CN114553783B CN 114553783 B CN114553783 B CN 114553783B CN 202210169362 A CN202210169362 A CN 202210169362A CN 114553783 B CN114553783 B CN 114553783B
Authority
CN
China
Prior art keywords
path
congestion
granularity
paths
cells
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
CN202210169362.8A
Other languages
Chinese (zh)
Other versions
CN114553783A (en
Inventor
高为民
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.)
Hunan Institute of Technology
Original Assignee
Hunan Institute of Technology
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 Hunan Institute of Technology filed Critical Hunan Institute of Technology
Priority to CN202210169362.8A priority Critical patent/CN114553783B/en
Publication of CN114553783A publication Critical patent/CN114553783A/en
Application granted granted Critical
Publication of CN114553783B publication Critical patent/CN114553783B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00—Traffic control in data switching networks
    • H04L47/10—Flow control; Congestion control
    • H04L47/12—Avoiding congestion; Recovering from congestion
    • H04L47/125—Avoiding congestion; Recovering from congestion by balancing the load, e.g. traffic engineering
    • 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
    • H04L43/0864—Round trip delays
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00—Traffic control in data switching networks
    • H04L47/10—Flow control; Congestion control
    • H04L47/19—Flow control; Congestion control at layers above the network layer
    • H04L47/193—Flow control; Congestion control at layers above the network layer at the transport layer, e.g. TCP related
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00—Traffic control in data switching networks
    • H04L47/10—Flow control; Congestion control
    • H04L47/28—Flow control; Congestion control in relation to timing considerations
    • H04L47/283—Flow control; Congestion control in relation to timing considerations in response to processing delays, e.g. caused by jitter or round trip time [RTT]
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/50—Reducing energy consumption in communication networks in wire-line communication networks, e.g. low power modes or reduced link rate

Landscapes

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

Abstract

The invention discloses a load balancing method for self-adaptively adjusting flow cell granularity of a data center network, which is applied to load balancing of the data center network, wherein RTT measurement is carried out on paths at a transmitting end, detection packets with timestamp options are periodically transmitted, RTT sample information of each link in the network is obtained, then path asymmetry estimation is carried out on the obtained RTT sample information, congestion paths and non-congestion paths are distinguished, the optimal size of the flow cell granularity is calculated according to path conditions, and the optimal size of the flow cell granularity is utilized to balance the disorder rate and the link utilization rate. The invention adopts a TCP congestion control mechanism, and is easier to be deployed in a large-scale network. The invention adaptively adjusts the size of the flow unit to realize low data packet disorder and high link utilization rate, and reduces the influence of uncertainty under high path asymmetry.

Description

Load balancing method for self-adaptive regulation of cell granularity of data center network
Technical Field
The invention relates to the field of data center load balancing, in particular to a load balancing method for self-adaptive regulation of cell granularity of a data center network.
Background
Modern data center network topologies typically take the form of multiple trees, through rich parallel paths, to provide high bandwidth. However, in data center networks, particularly production data center networks, there is a wide range of path diversity caused by dynamic traffic, link failures, and heterogeneous switch devices, which makes multipath load balancing methods in data centers robust to path diversity. The existing solutions are mainly fine-grained schemes and coarse-grained schemes, wherein the fine-grained schemes such as RPS (random packet diffusion) and Presto make full use of available paths, but they are prone to data packet disorder problems under asymmetric topologies; coarse-grained schemes such as ECMP (equal cost multi-path routing strategy) and Letflow effectively avoid packet misordering, but tend to result in lower utilization of the multi-path.
Disclosure of Invention
The invention aims to provide a load balancing method for adaptively adjusting flow cell granularity of a data center network, which can adaptively adjust the flow cell granularity according to path diversification so as to reduce packet disorder under a larger asymmetric topology and reduce the flow cell granularity at the same time so as to obtain higher link utilization rate under a smaller asymmetric topology.
The technical scheme adopted for solving the technical problems is as follows: a load balancing method for self-adaptively adjusting granularity of cells in a data center network comprises the following steps:
step 1, RTT (round trip time) measurement is carried out on a path at a transmitting end, a detection packet with a time stamp option is periodically transmitted, and RTT sample information of each link in a network is obtained;
step 2, carrying out path asymmetry estimation on the RTT sample information obtained in the step 1 to obtain a real-time congestion state, and dividing the path into a congestion path and a non-congestion path;
step 3, calculating the optimal size of the granularity of the flow cells according to the path state separated in the step 2, so as to adjust the granularity of the flow cells;
and 4, after the optimal size of the flow cell granularity calculated in the step 3, the flow cells are transmitted on all paths through polling so as to balance the disorder rate and the link utilization rate.
Further, the path asymmetry estimation described in step 2 adopts a TCP (transmission control protocol) congestion control mechanism, and when the sender receives ACK (acknowledgement character) packets, the bad path probability calculation is equal to the ratio of the number of ACK packets having a large transmission delay received at the sender to the total number of received ACK packets, based on the path transmission delay corresponding to each ACK.
Further, the calculating the optimal size of the flow cell granularity in the step 3 specifically includes:
step 3-1, determining the value range of the granularity of the flow cells: f (F) size And the number of stored packets in which the gran represents the granularity of the TCP stream and the stream cells, respectively, the minimum granularity of the stream cells is 1500B (1 packet) and the maximum granularity is 64KB (44 packets), so the value of the gran ranges from 1 to 44, and then the stream is cut into pieces according to the granularity of the gran
Figure GDA0004232724590000021
Individual cells;
step 3-2, calculating the transmission delay ratio X of the bad path transmission delay and the good path transmission delay: when the flow cells can select parallel paths, the paths are formed by N b Individual congestion paths and N g The non-congestion paths are composed of propagation delays D b And D g . The ratio of the number of bad paths to the number of good paths is R, R is equal to
Figure GDA0004232724590000022
Then X is equal to->
Figure GDA0004232724590000023
Step 3-3, calculating the disorder probability P of the flow cells: number of congestion paths N b ，P g And P b Respectively representing the probability of selecting a non-congestion path and a congestion path in the transmission process of the flow cell, the probability of selecting the congestion path can be calculated
Figure GDA0004232724590000024
Probability of non-congested paths
Figure GDA0004232724590000025
When a sequence of data packets is transmitted on a congested path and subsequently transmitted flow cells are transmitted on a non-congested path, an out-of-order event occurs, where the out-of-order probability P of the congested path is calculated as:
Figure GDA0004232724590000031
the out-of-order probability P for a non-congested path is calculated as:
Figure GDA0004232724590000032
in the above formula, n is the number of flow cells, and m is the number of parallel paths;
step 3-4, calculating an average congestion window
Figure GDA0004232724590000033
When detecting out-of-order data packets, a TCP sender reduces the congestion window by half, and in data transmission, the out-of-order rate of the congestion window and a flow cell meets the following relation:
Figure GDA0004232724590000034
Figure GDA0004232724590000035
in the above, W 0 Representing an initial value of a network congestion window; maxW represents the maximum window for each round of transmission on the path; i represents the number of rounds in data transmission, n i Representing the number of the ith alternate cells; w (W) i Representing an ith round of congestion window;
thereafter, calculate each round W i Is a congestion window of (a):
Figure GDA0004232724590000036
according to W i Is calculated to a value of the total congestion window number W S The method comprises the following steps:
Figure GDA0004232724590000037
when WS is greater than or equal to F for the first time size When calculating the average congestion window
Figure GDA0004232724590000038
Figure GDA0004232724590000039
In the above formula, r represents the number of data packet transmission rounds;
step 3-5, calculating the optimal flow cell granularity gram: by using more paths through small-granularity flow cells, the total used bandwidth is further increased, and since each flow cell randomly selects its transmission path and sets the link bandwidth C of each path, the total bandwidth n of the flow cells is equal to C, and the average congestion is known
Figure GDA0004232724590000041
Average end-to-end round trip time +.>
Figure GDA0004232724590000042
The method comprises the following steps:
Figure GDA0004232724590000043
obtained according to the above formula (10)
Figure GDA0004232724590000044
The completion time for the entire data stream (FCT) is then calculated as:
Figure GDA0004232724590000045
finally, a cell granularity gram is obtained according to the formula (11):
Figure GDA0004232724590000046
compared with the prior art, the invention has the following beneficial effects:
1. the invention realizes the periodic collection of path delay with limited expenditure through the architecture modes of the path asymmetry estimation module and the flow cell granularity adjustment module, thereby rapidly distinguishing the congested path from the non-congested path and improving the utilization rate of multipath.
2. The invention adopts a TCP congestion control mechanism to realize compatibility with the existing transport layer protocol, thereby being easier to be deployed in a large-scale network.
3. The invention adaptively adjusts the size of the flow unit to realize low data packet disorder and high link utilization rate, lightens the influence of uncertainty under high path asymmetry, and reduces the size of the flow cell to avoid high tail delay and data packet disorder, and simultaneously uses all paths to realize high utilization rate and network throughput, only needs to be deployed at a transmitting end, does not need to make any modification on a TCP/IP protocol stack of an opposite-end host or a switch, and has good performance and convenient use.
Drawings
FIG. 1 is a schematic diagram of a load balancing method for adaptively adjusting granularity of cells in a data center network according to an embodiment of the present invention;
FIG. 2 is an algorithm diagram of a load balancing method for adaptively adjusting the granularity of cells in the data center network of FIG. 1;
FIG. 3 is a graph comparing performance under an asymmetrical topology network search workload according to an embodiment of the present invention;
FIG. 4 is a graph comparing performance under an asymmetric topology Web service workload according to an embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the drawings and examples.
Referring to fig. 1, the present embodiment provides a load balancing method for adaptively adjusting granularity of cells in a data center network, which includes the following steps;
step 1, RTT (round trip time) measurement is carried out on a path at a transmitting end, a detection packet with a time stamp option is periodically transmitted, and RTT sample information of each link in a network is obtained;
step 2, carrying out path asymmetry estimation on the RTT sample information obtained in the step 1 to obtain a real-time congestion state, and dividing the path into a congestion path and a non-congestion path;
step 3, calculating the optimal size of the granularity of the flow cells according to the path state separated in the step 2, so as to adjust the granularity of the flow cells;
and 4, after the optimal size of the flow cell granularity calculated in the step 3, the flow cells are transmitted on all paths through polling so as to balance the disorder rate and the link utilization rate.
The path asymmetry estimation in step 2 adopts a TCP (transmission control protocol) congestion control mechanism, and when the sender receives ACK (acknowledgement character) packets, the bad path probability calculation is equal to the ratio of the number of ACK packets having a large transmission delay received at the sender to the total number of received ACK packets, based on the path transmission delay corresponding to each ACK. For example, the transmitting end transmits 10 flow cells, the number of the flow cells should be generally consistent with the number of paths, and since the switch uses polling scattering to route the flow cells, the transmitting end calculates the probability of a bad path according to the path delay, when the transmission time of two flow cells is obviously longer than that of the other 8 flow cells, the two flow cells may take a bad path, and then the probability of the bad path is 0.2. In order to reduce overhead and enhance scalability, the present embodiment periodically uses a small amount of data and ACK packets to carry delay information in the option field of the TCP header.
Under different network topology asymmetries, the size of the flow cells affects the probability of TCP disorder and the network utilization, so that optimization calculation of the size of the flow cell granularity is needed.
Referring to fig. 2, the calculating the optimal size of the granularity of the flow cell in step 3 specifically includes:
step 3-1, determining the value range of the granularity of the flow cells: f (F) size And the number of stored packets in which the gran represents the granularity of the TCP stream and the stream cells, respectively, the minimum granularity of the stream cells is 1500B (1 packet) and the maximum granularity is 64KB (44 packets), so the value of the gran ranges from 1 to 44, and then the stream is cut into pieces according to the granularity of the gran
Figure GDA0004232724590000061
Individual cells;
step 3-2, calculating the transmission delay ratio X of the bad path transmission delay and the good path transmission delay: when the cells transmit on multiple paths, the cells are out of order only when at least one of the transmitted data cells arrives at the destination earlier than the previously transmitted cells, and when the cells can select parallel paths, the paths are defined by N b Individual congestion paths and N g The non-congestion paths are composed of propagation delays D b And D g . The ratio of the number of bad paths to the number of good paths is R, R is equal to
Figure GDA0004232724590000062
Then X is equal to
Figure GDA0004232724590000063
Step 3-3, calculating the disorder probability P of the flow cells: due toThe uncertainty of network traffic, the number of congested and uncongested paths will vary with the diversity of the paths, according to the number of previous congested paths N in order to avoid synchronization effects b Setting P g And P b Respectively representing the probability of selecting a non-congestion path and a congestion path in the transmission process of the flow cell, the probability of selecting the congestion path can be calculated as
Figure GDA0004232724590000064
Probability of non-congested paths
Figure GDA0004232724590000065
When a sequence of data packets is transmitted on a congested path and subsequently transmitted flow cells are transmitted on a non-congested path, an out-of-order event occurs, where the out-of-order probability P of the congested path is calculated as:
Figure GDA0004232724590000071
the out-of-order probability P for a non-congested path is calculated as:
Figure GDA0004232724590000072
in the above formula, n is the number of flow cells, and m is the number of parallel paths;
step 3-4, calculating an average congestion window
Figure GDA0004232724590000073
When detecting out-of-order data packets, a TCP sender reduces the congestion window by half, and in data transmission, the out-of-order rate of the congestion window and a flow cell meets the following relation:
Figure GDA0004232724590000074
Figure GDA0004232724590000075
in the above, W 0 Representing an initial value of a network congestion window; maxW represents the maximum window for each round of transmission on the path; i represents the number of rounds in data transmission, n i Representing the number of the ith alternate cells; w (W) i Representing an ith round of congestion window;
thereafter, calculate each round W i Is a congestion window of (a):
Figure GDA0004232724590000076
according to W i Is calculated to a value of the total congestion window number W S The method comprises the following steps:
Figure GDA0004232724590000077
when W is S For the first time greater than or equal to F size When calculating the average congestion window
Figure GDA0004232724590000078
Figure GDA0004232724590000079
In the above formula, r represents the number of data packet transmission rounds;
step 3-5, calculating the optimal flow cell granularity gram: by using more paths through small-granularity flow cells, the total used bandwidth is further increased, and since each flow cell randomly selects its transmission path and sets the link bandwidth C of each path, the total bandwidth n of the flow cells is equal to C, and the average congestion is known
Figure GDA0004232724590000081
Average end-to-end round trip time +.>
Figure GDA0004232724590000082
The method comprises the following steps:
Figure GDA0004232724590000083
obtained according to the above formula (10)
Figure GDA0004232724590000084
The completion time for the entire data stream (FCT) is then calculated as:
Figure GDA0004232724590000085
finally, a cell granularity gram is obtained according to the formula (11):
Figure GDA0004232724590000086
in this embodiment, a large-scale leaf ridge topology structure is constructed to evaluate the performance of the method, that is, 256 hosts are set, 8 leaf switches and 8 trunk switches are connected with 256 hosts through a plurality of 1Gbps links to form a layer of transverse network structure parallel to the longitudinal network structure of the trunk, the size of a switch buffer is set to 250kb,8 equivalent paths are formed between the 8 leaf switches and the 8 trunk switches, the delay between any pair of hosts in the 8 equivalent paths is set to be less than 100 μs, then one path is randomly selected and the round-trip propagation delay is set to be 300 μs, so that the path asymmetry is generated, and then the two paths are used as the workload of the data center by using a Web server and a network search; the Web server has at most short data streams with a duty ratio of 86% and the number of long data streams with a duty ratio of only 14%, and the data size of the long data streams of the Web server exceeds 1MB; the number of long data streams of the network search is 38%, and the number of short data streams is 62%; in both workloads, the number of short data streams is higher than that of long data streams, and the experiment is carried out according to the Poisson process, wherein the number is randomly generated and the load change is from 0.1 to 0.8, the performance of a load balancing method (PDLB) for adaptively adjusting the granularity of a flow cell of a data center network is evaluated by comparing with RPS, letFlow, ECMP (equivalent routing), the experimental result is shown in fig. 3 and fig. 4, and compared with the existing most advanced load balancing scheme, the experiment result shows that the leveling completion time is reduced by 11-53%.

Claims (2)

1. The load balancing method for the self-adaptive regulation of the granularity of the cells of the data center network is characterized by comprising the following steps of:
step 1, RTT (round trip time) measurement is carried out on a path at a transmitting end, a detection packet with a time stamp option is periodically transmitted, and RTT sample information of each link in a network is obtained;
step 2, carrying out path asymmetry estimation on the RTT sample information obtained in the step 1 to obtain a real-time congestion state, and dividing the path into a congestion path and a non-congestion path;
step 3, calculating the optimal size of the granularity of the flow cells according to the path state separated in the step 2, so as to adjust the granularity of the flow cells;
step 4, after the optimal size of the flow cell granularity calculated in the step 3, the flow cells are transmitted on all paths through polling so as to balance the disorder rate and the link utilization rate;
wherein, in step 3, the calculating the optimal size of the flow cell granularity specifically includes:
step 3-1, determining the value range of the granularity of the flow cells: f (F) size And the number of stored packets in which the gran represents the granularity of the TCP stream and the stream cells, respectively, the minimum granularity of the stream cells is 1500B (1 packet) and the maximum granularity is 64KB (44 packets), so the value of the gran ranges from 1 to 44, and then the stream is cut into pieces according to the granularity of the gran
Figure QLYQS_1
Individual cells;
step 3-2, calculating the transmission delay ratio X of the bad path transmission delay and the good path transmission delay: when the flow cells can select parallel paths, the paths are formed by N b Individual congested pathsN g The non-congestion paths are composed of propagation delays D b And D g The method comprises the steps of carrying out a first treatment on the surface of the The ratio of the number of bad paths to the number of good paths is R, R is equal to
Figure QLYQS_2
X is equal to->
Figure QLYQS_3
；
Step 3-3, calculating the disorder probability P of the flow cells: number of congestion paths N b ，P g And P b Respectively representing the probability of selecting a non-congestion path and a congestion path in the transmission process of the flow cell, the probability of selecting the congestion path can be calculated
Figure QLYQS_4
（1）
Probability of non-congested paths
Figure QLYQS_5
；
When a sequence of data packets is transmitted on a congested path and subsequently transmitted flow cells are transmitted on a non-congested path, an out-of-order event occurs, where the out-of-order probability P of the congested path is calculated as:
Figure QLYQS_6
（3）
the out-of-order probability P for a non-congested path is calculated as:
Figure QLYQS_7
（4）
in the above formula, n is the number of flow cells, and m is the number of parallel paths;
step 3-4, calculating an average congestion window
Figure QLYQS_8
: when detecting out-of-order data packets, the TCP sender reduces its congestion window by half, during data transmissionThe congestion window and the disorder rate of the flow cells satisfy the following relationship:
Figure QLYQS_9
（5）
Figure QLYQS_10
（6）
in the above, W 0 Representing an initial value of a network congestion window; maxW represents the maximum window for each round of transmission on the path; i represents the number of rounds in data transmission, n i Representing the number of the ith alternate cells; w (W) i Representing an ith round of congestion window;
thereafter, calculate each round W i Is a congestion window of (a):
Figure QLYQS_11
（7）
according to W i Is calculated to a value of the total congestion window number W S The method comprises the following steps:
Figure QLYQS_12
（8）
when WS is greater than or equal to F for the first time size When calculating the average congestion window
Figure QLYQS_13
：
Figure QLYQS_14
（9）
In the above formula, r represents the number of data packet transmission rounds;
step 3-5, calculating the optimal flow cell granularity gram: by using more paths through small-granularity flow cells, the total used bandwidth is further increased, and since each flow cell randomly selects its transmission path and sets the link bandwidth C of each path, the total bandwidth n of the flow cells is equal to C, and the average congestion is known
Figure QLYQS_15
Average end-to-end round trip time +.>
Figure QLYQS_16
The method comprises the following steps:
Figure QLYQS_17
（10）
obtained according to the above formula (10)
Figure QLYQS_18
The completion time for the entire data stream (FCT) is then calculated as:
Figure QLYQS_19
（11）
finally, a cell granularity gram is obtained according to the formula (11):
Figure QLYQS_20
（12）。
2. the load balancing method according to claim 1, wherein in step 2, the path asymmetry estimation uses a TCP (transmission control protocol) congestion control mechanism, and when the sender receives ACK (acknowledgement character) packets, the bad path probability calculation is equal to a ratio of the number of ACK packets having a large transmission delay received at the sender to the total number of received ACK packets, based on the path transmission delay corresponding to each ACK.
CN202210169362.8A 2022-02-23 2022-02-23 Load balancing method for self-adaptive regulation of cell granularity of data center network Active CN114553783B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210169362.8A CN114553783B (en) 2022-02-23 2022-02-23 Load balancing method for self-adaptive regulation of cell granularity of data center network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210169362.8A CN114553783B (en) 2022-02-23 2022-02-23 Load balancing method for self-adaptive regulation of cell granularity of data center network

Publications (2)

Publication Number Publication Date
CN114553783A CN114553783A (en) 2022-05-27
CN114553783B true CN114553783B (en) 2023-06-16

Family

ID=81676832

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210169362.8A Active CN114553783B (en) 2022-02-23 2022-02-23 Load balancing method for self-adaptive regulation of cell granularity of data center network

Country Status (1)

Country Link
CN (1) CN114553783B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102474455A (en) * 2009-06-29 2012-05-23 香港中文大学 Network traffic accelerator
CN110351187A (en) * 2019-08-02 2019-10-18 中南大学 Data center network Road diameter switches the adaptive load-balancing method of granularity
CN111542089A (en) * 2020-05-14 2020-08-14 海南大学 Multipath congestion control method and device based on time delay and window jitter

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
AU2017383561B2 (en) * 2016-12-21 2022-11-10 Dejero Labs Inc. Packet transmission system and method
CN107395325B (en) * 2017-07-10 2020-04-17 中国电子科技集团公司第二十八研究所 Adaptive estimation transmission method serving tactical communication network
CN107634911B (en) * 2017-10-31 2020-03-10 河南科技大学 Adaptive congestion control method based on deep learning in information center network
CN109981693A (en) * 2017-12-27 2019-07-05 上海文广互动电视有限公司 The method of rate control and system of 3D content distribution platform

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102474455A (en) * 2009-06-29 2012-05-23 香港中文大学 Network traffic accelerator
CN110351187A (en) * 2019-08-02 2019-10-18 中南大学 Data center network Road diameter switches the adaptive load-balancing method of granularity
CN111542089A (en) * 2020-05-14 2020-08-14 海南大学 Multipath congestion control method and device based on time delay and window jitter

Also Published As

Publication number Publication date
CN114553783A (en) 2022-05-27

Similar Documents

Publication Publication Date Title
CN101404622B (en) Wireless internet congestion control method based on multi-path load balancing and controller thereof
CN102006230B (en) Method for controlling congestion control by fusing three kinds of information in wired/wireless hybrid network
CN112804125B (en) Named data network congestion control method based on fuzzy comprehensive evaluation algorithm
Ren et al. An explicit congestion control algorithm for named data networking
Aboobaker et al. Streaming media congestion control using bandwidth estimation
JP4726978B2 (en) Congestion control device and congestion control method
Zhang et al. A survey of TCP congestion control algorithm
CN109039941B (en) Adaptive packet scattering method based on path classification in data center network
Lee et al. Reviving delay-based TCP for data centers
CN114553783B (en) Load balancing method for self-adaptive regulation of cell granularity of data center network
Hamadani et al. TCP contention control: a cross layer approach to improve TCP performance in multihop ad hoc networks
CN110324255B (en) Data center network coding oriented switch/router cache queue management method
Lien et al. A new TCP congestion control mechanism over wireless ad hoc networks by router-assisted approach
CN113162863B (en) Fair multipath routing method based on wheel disc algorithm
CN110267322A (en) One kind being based on the improved unmanned plane ad hoc network fast reaction routing algorithm of DSR agreement
CN114828081A (en) Cooperative hybrid congestion control method based on path recovery
Pu et al. Enhancements on router-assisted congestion control for wireless networks
Asmaa et al. EC-elastic an explicit congestion control mechanism for named data networking
CN115134307A (en) Load balancing method based on packet loss rate coding in cloud computing
CN109257302B (en) Packet scattering method based on packet queuing time
CN111416777A (en) Load balancing method and system based on path delay detection
CN113766561B (en) Unmanned cluster network congestion control method based on cross-layer optimization
Li et al. Achieving Low Latency for Multipath Transmission in RDMA Based Data Center Network
CN116760777B (en) Multipath congestion control method based on ABEA3C
Brachman et al. On TCP NCR improvement using ECN-capable routers

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
GR01 Patent grant
GR01 Patent grant