CN108631947A - A kind of RDMA network data transmission methods based on correcting and eleting codes - Google Patents
A kind of RDMA network data transmission methods based on correcting and eleting codes Download PDFInfo
- Publication number
- CN108631947A CN108631947A CN201810487054.3A CN201810487054A CN108631947A CN 108631947 A CN108631947 A CN 108631947A CN 201810487054 A CN201810487054 A CN 201810487054A CN 108631947 A CN108631947 A CN 108631947A
- Authority
- CN
- China
- Prior art keywords
- data
- frame
- block
- receiving terminal
- correcting
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000005540 biological transmission Effects 0.000 title claims abstract description 49
- 238000000034 method Methods 0.000 title claims abstract description 20
- 230000003139 buffering effect Effects 0.000 claims description 15
- 230000000903 blocking effect Effects 0.000 claims description 2
- 239000011159 matrix material Substances 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000009897 systematic effect Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000013467 fragmentation Methods 0.000 description 1
- 238000006062 fragmentation reaction Methods 0.000 description 1
- 238000003306 harvesting Methods 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 238000010025 steaming Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
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
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0056—Systems characterized by the type of code used
-
- 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
- H04L1/0001—Systems modifying transmission characteristics according to link quality, e.g. power backoff
- H04L1/0006—Systems modifying transmission characteristics according to link quality, e.g. power backoff by adapting the transmission format
-
- 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
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0056—Systems characterized by the type of code used
- H04L1/0061—Error detection codes
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Quality & Reliability (AREA)
- Communication Control (AREA)
Abstract
The invention discloses a kind of RDMA network data transmission methods based on correcting and eleting codes, the described method comprises the following steps：(1) transmitting terminal obtains the data of application, and data are cut into the data frame of fixed size, then by each frame data cutting is the identical data block of k size, is then sent to receiving terminal；(2) if at the appointed time there is data block not reach receiving terminal in threshold value, transmitting terminal generates m check block according to block coding to this k number, check block is sent to receiving terminal；(3) as soon as receiving terminal, which receives this k+m arbitrary k in the block, can recover frame data, a data transmission is completed when receiving all data frames.The present invention generates redundant data by correcting and eleting codes allows receiving terminal without waiting for the data packet on delay path, to reduce transmission delay；It is unloaded by correcting and eleting codes hardware, does not increase the computing cost of CPU.
Description
Technical field
The invention belongs to network technique fields, and in particular to a kind of RDMA network data transmission methods based on correcting and eleting codes.
Background technology
In face of the application of the high concurrents low delays such as cloud computing, big data, artificial intelligence calculating, traditional TCP/IP software and hardwares
Framework cannot be satisfied demand.Because of the multiple memory copying of TCP/IP software and hardware architectures, interrupt processing, context switching, complexity
ICP/IP protocol processing, store and forward message mode and packet loss result in excessively high propagation delay time and additional CPU computing costs.
It is mutual as one of TCP/IP that remote direct data accesses (Remote Direct Memory Access, RDMA)
Benefit technology provides the messenger service for directly accessing the virtual memory on remote machine.Since RDMA network interface cards can be with direct copying number
According to, it is possible to complicated network protocol stack is got around, and minimizes the participation of operating system.This design realizes data biography
It is defeated postpone, the high expense for handling up and reducing CPU.
Modern data center is usually made of hundreds of server, and inside needs the two-way tape of TB ranks wide to send out
Send data.In order to meet such network demand and reduce expense, most of data center uses the net of Clos topological structures
Network (usual three layers even more), expand horizontally network capacity.These Web vector graphic interchangers are replaced with a large amount of networks
The router of port.Fig. 1 is the topology of the data center network of a typical Clos structure.
Data center server runs a large amount of cloud service.These services produce diversified flow rate mode.This
A little different data flows have network performance different requirements, wherein the data volume of big steaming transfer is big, to bandwidth requirement height；Small stream
The data volume of transmission is small but generally all carries control information, more demanding to transmission delay.Big stream and small stream mix in a network
When closing transmission, small stream is probably due to being dispatched to transmission after big stream and generating very high delay.If using fine-grained
Transmission, the small stream that big stream is cut into fixed size are transmitted, and can may be used also to avoid the above problem, and greatly after stream cutting
With by the multipath parallel transmission of RDMA networks in data center, to reduce the propagation delay time of data.But at the same time,
This also brings problem：The deadline of one data flow is determined by the data block finally reached, will produce long tail effect；And
Recipient needs to realize a complicated sort algorithm to reset out initial data.
Invention content
Goal of the invention：Based on the deficiencies of the prior art, the present invention proposes that a kind of RDMA network datas based on correcting and eleting codes pass
Transmission method reduces propagation delay time by fine granularity transmission and multipath parallel transmission, and eliminates network transmission using correcting and eleting codes
Long tail effect and the operation of reordering for simplifying receiving terminal, while correcting and eleting codes are calculated using the unloading of the correcting and eleting codes hardware of RDMA network interface cards,
Reduce CPU computing costs.
Technical solution：A kind of RDMA network data transmission methods based on correcting and eleting codes of the present invention, including following step
Suddenly：
S1, transmitting terminal obtain the data of application, data are cut into the data frame of fixed size, then each frame data are cut
It is divided into the identical data block of k size, is then sent to receiving terminal；
If also having data block not reach receiving terminal in S2, at the appointed time threshold value, transmitting terminal compiles this k number according to block
Code generates m check block, and check block is sent to receiving terminal；
S3, receiving terminal receive this k+m data it is arbitrary k in the block when, a frame data are recovered, when receiving terminal receives
When all data frames, a data transmission is just completed.
Further, the step S1 includes：Transmitting terminal copies the data in application memory according to frame sign to one by one
In buffering area, and a frame number is distributed for this frame data, one can all be added by often sending a frame data frame number.By a frame data cutting
It is k number according to block, the size of each data block is block_size byte.Then use k QP concurrently by these data
Block is sent to receiving terminal.Then this data frame is added in the sequence of events of timer.
The step S2 includes：Transmitting terminal is that a timer is arranged in each frame data, is entangled to data frame calculating after time-out
Code is deleted, check block is sent to receiving terminal.Timer module checks whether the data frame that each sends is overtime, if after time-out
K number also has in does not successfully arrive at receiving terminal, then delay has been likely to occur, at this moment by correcting and eleting codes module to this frame data
It is encoded, generates m check block, then data transmission module is called to send it to receiving terminal with other m QP.
The step S3 includes：Receiving terminal decodes data frame after receiving data block.Receiving terminal receives a data block
After, identify which data frame it belongs to by the immediate of the Write with immediate operation of RDMA, according to connecing
The QP of the data block of receipts knows it is which of this data frame data block or check block, when being collected into data frame
When data block adds the sum of check block to reach k, the original contents of this frame data are obtained by the decoding of correcting and eleting codes, most
This frame data is added in the queue of data accepted afterwards, application program is waited for read.
Further, all data frames received are ranked into initial data by receiving terminal, complete to receive.Receiving terminal according to
Frame number reads the data frame having been received one by one, and receiving terminal records the data next to be read with a variable read_pos
The frame number of frame reads data and is divided into block type and non-block type, and application program is always waited for until having received under block type
Occurs the data frame that frame number is read_pos in data queue, under non-blocking mode, if in the data queue received
There is no the data frame that frame number is read_pos just directly to return, application program has read a frame data from the reception buffering area of system
After, by read_pos plus one, next frame data are then read, until the buffering area of user is full or encounters a data transmission
Last frame data.
Advantageous effect：Compared with the prior art, the present invention has the following advantages：
1, the present invention using correcting and eleting codes by data cutting be the identical data frame of size, then by each frame data cutting be k
The identical data block of a size, receiving terminal is sent to by data block, is calculated correcting and eleting codes according to block to this k number, is generated m and verify
Block, then k number is sent to receiving terminal according to block.Receiving terminal, which only needs to receive this k+m k in the block, can recover number
According to without waiting for all data block, transmission delay is reduced by data redundancy.
2, in order to reduce the CPU computing costs of coding, the present invention uses the correcting and eleting codes hardware offloading functions of RDMA network interface cards, makes
Correcting and eleting codes are calculated with hardware, do not increase the computing cost of CPU when using correcting and eleting codes.
3, the present invention uses system code scheme, includes primary data information (pdi) in systematic code, when k number is all first according to block
When check block reaches receiving terminal, initial data can be directly obtained, calculating need not be decoded, reduces decoding expense.
4, present system is using the strategy of delay coding, the delay of data and to lose opposite be less generation in network
, by setting a time threshold, only when also having data block not successfully arrive at receiving terminal after time-out, then
It calculates check block and sends, further decrease encoding overhead.
5, the present invention proposes a kind of algorithm of the rearrangement based on correcting and eleting codes.In transmitting terminal by data fragmentation, then
Correcting and eleting codes are calculated, when by data block and check block by multipath parallel transmission, because being not in loss of data or nothing
The case where time limit waits for, so receiving terminal can utilize this feature to simplify the algorithm resequenced.
Description of the drawings
Fig. 1 is the network topological diagram using the data center of Clos frameworks；
Fig. 2 is Transmission system structure chart according to the ... of the embodiment of the present invention；
Fig. 3 is the data frame structure figure of network transmission according to the ... of the embodiment of the present invention；
Fig. 4 is the asynchronous model schematic diagram of RDMA correcting and eleting codes computing hardwares unloading.
Specific implementation mode
Technical scheme of the present invention is described further below in conjunction with the accompanying drawings.
Fig. 1 is the topology of the data center network of a typical Clos structure.Interchanger (ToR) in rack passes through d0
Server in the link connection rack (Rack) of a 10G, then passes through the link connection of d1 40G to aggregation switch
(Agg) on.Each aggregation switch connects interchanger (ToR) and d3 host exchange (Spine) in d2 rack.Institute
To have the same path of d1 items between the host in same convergence region (pod), across between the different regions of host exchange
There is the same path of d1*d3 items.For example, when d1=d3=8, such path has 64.Make full use of these parallel
Path, by data flow cutting be fine-grained data block carry out parallel transmission, can effectively accelerate the data between node to pass
It is defeated, reduce the network delay of data transmission.
The data transmission method of the present invention mainly includes the following steps that：(1) transmitting terminal obtains the data of application, and data are cut
It is divided into the data frame of fixed size, then by each frame data cutting is the identical data block of k size, is then sent to reception
End；(2) if at the appointed time also having data block not reach receiving terminal in threshold value, transmitting terminal gives birth to this k number according to block coding
At m check block, check block is sent to receiving terminal；(3) receiving terminal receives this k+m arbitrary k in the block and can restore
As soon as going out frame data, a data transmission is completed when receiving all data frames.
In one embodiment, data transmission system structure is as shown in Figure 2.The system is the solution party of a host computer side
Case, and be a data transmission middleware for being located at client layer, therefore it is independently of transport layer and the net to host computer side
Network stack is transparent.In transmitting terminal, using the memory address and length for providing the data to be sent, the network address of receiving terminal, system
The data that application can be intercepted and captured, are copied in the buffering area of system administration, then by its cutting be fixed size (such as
Data block 64KB) simultaneously encodes data block.Data block and check block are finally sent to receiving terminal, are responsible for by receiving terminal
Decode and submit to the application of receiving terminal.In receiving terminal, system receives data block, while decoding initial data, resequences
It afterwards will be in the memory field of original data block copy to application according to original order.Because each frame data can be by decoding
It arrives, so the data block inside frame need not sort, because the deadline of each frame is controlled in a threshold value, only
Serial number need to simply be pressed waits for data to reach, and also need not particularly sort.System includes five main modules：It is slow
Rush area, data transmission module, correcting and eleting codes module, timer module, event processing module.Their function is respectively：
(1) buffering area includes to send buffering area and receive buffering area, for the number kept in data to be sent He received
According to.Why it is because of one interim region of memory of needs when calculating correcting and eleting codes, and the biography of data with buffering area
It is defeated to need buffering area to do synchronization.When transmission data, first copies data to buffering area by the frame of fixed size, then send out
Receiving terminal is given, if there is delay, then data is encoded, check block is then sent to receiving terminal.Receiving terminal receives
A frame data will be decoded after to enough data, and application program is waited for read.Both sides need the position being written when synchronous send
It sets, because the data that receiving terminal is read not yet cannot be covered.
(2) data transmission touches the transmitting-receiving that block is responsible for data.Data transmission procedure：Data to be sent are obtained from application,
From the beginning every time one frame data of copy (frame data are equal to the number of size data block when being multiplied by coding of data block) to being
In the buffering area of system.Then it is k number by this frame data cutting to operate the number after cutting according to block, then with the Write of RDMA
It is write in the caching of receiving terminal by different QP (Queue Pairs) according to block, finally gives the information of this frame to timer
Module waits to be encoded.This process is repeated until all data are sent completely.DRP data reception process：Foot will have been received
The data of the frame of enough data blocks, which decode, to be come, and is then copied in the memory of application, when application buffering area is full or arrival one
It is returned to when the tail portion of data.
(3) correcting and eleting codes module is responsible for the number for calling the correcting and eleting codes hardware of network interface card to unload to encode or according to having been received
Initial data is decoded according to block.
(4) timer module counts the arrival information of the data block sent in each frame data, from these data blocks
Start to calculate when transmission, when more than the time threshold of setting, timer can call coding/decoding module to encode data, so
The check block that coding obtains is sent to receiving terminal afterwards.
(5) event processing module mainly handles two class events：First, data send successful event, second is that receiving new number
According to event.When receiving a data transmission successful event, event processing module can update the every of timer preservation
The successful statistical information of data block in one frame, timer module can decide whether to be encoded according to this information, then allow number
Check block is sent to receiving terminal according to transmission module.When receiving the event for receiving new data, if it is a data frame
In first data block/check block, a new decoding context will be created, otherwise can update the decoding of corresponding data frame
Contextual information.When receiving enough data block/check blocks, note that this frame data of data transmission module have connect
Harvest work(.
It is group that transmission data of the present invention, which is by data frame, is unit transmission by the data block in data frame, data frame
Structure is as shown in Figure 3.Each connection forms (i.e. QP) by k+m virtual link, and each frame data have k number from the above
According to block and m check block, altogether by k+m block.I-th piece of each data frame is sent to by i-th of QP in receiving terminal
It deposits on i-th piece of position of corresponding frame, does not have to thus directly judge data block from the QP of reception into line label
Which block is data be.Because the data for being discontented with a frame when coding for size need zero padding, in a frame not necessarily
All it is valid data, so low 24 of the beginning data of one 32bit of each frame preserve the significant figure preserved in this frame
According to size, its value range is 0~(block_size*k-4), and most-significant byte is used for preserving the type of frame, and (frame is original at one
Position in data), it can be with value：
(1)FULL：This frame is exactly a complete data；
(2)BEGIN：This is the first frame of a data；
(3)MID：This is the frame being located among a data；
(4)END：This is the frame of an end of data.
The present invention reduces the expense of correcting and eleting codes calculating using the hardware offloading functions of the correcting and eleting codes of RDMA network interface cards calculating.It entangles
It is as shown in Figure 4 to delete yard encoding model of computing hardware unloading.Reed-Solomon (RS) is used to encode.RS codings are to be based on
A kind of encryption algorithm of finite field uses GF (2 in RS codingsw), wherein 2w>(wherein w is the length of symbol to=k+m, and k is number
According to the number of block, m is the number of check block).RS codings are coding and decoding unit with symbol, and big data block is split to word length
For the symbol of w (value is generally 8 or 16), encoding and decoding then are carried out to symbol.Because of data delay and loss of data phase
It is less generation for, in the case where all data blocks all normally reach, it is desirable to need not decode.The present invention uses
A kind of coding mode being referred to as systematic code includes primary data information (pdi) after coding, in data block.In this coding
Under, data frame is split as k data blocks, passes through coding and generates m check block.This k check block has as initial data
Content, when they reach receiving terminals when, data directly can be submitted into upper layer application without decoding operate.When going out
When existing data-bag lost or network delay, data can be recovered to decode using check block.The net that the present invention selects
Card correcting and eleting codes hardware Unloading Model is asynchronous computing model, this model it is more efficient because calculating correcting and eleting codes in network interface card
When, process is completed without waiting for calculating, but can go to calculate or execute other tasks, its workflow is：
(1) encod_async (data, code ...) interface is called, data is data block, and code is check block；
(2) data block is sent to receiving terminal, withouts waiting for coding and completes；
(3) completion to be encoded such as asynchronous, CPU can handle other tasks during this；
(4) coding is completed, and check block is sent to receiving terminal node.
RDMA network interface cards need extraneous offer encoder matrix and decoding matrix when calculating correcting and eleting codes, this is needed by third party
Correcting and eleting codes calculate library and provide.Jerasure be one support it is cross-platform open soft correcting and eleting codes calculating instrument library, support model moral
Cover the RS codings and cauchy RS codings of moral matrix.The present invention calculates the matrix of encoding and decoding using Jerasure.
Claims (6)
1. a kind of RDMA network data transmission methods based on correcting and eleting codes, which is characterized in that the described method comprises the following steps：
S1, transmitting terminal obtain the data of application, data are cut into the data frame of fixed size, then be by each frame data cutting
The identical data block of k size, is then sent to receiving terminal；
If also having data block not reach receiving terminal in S2, at the appointed time threshold value, transmitting terminal gives birth to this k number according to block coding
At m check block, check block is sent to receiving terminal；
S3, receiving terminal receive this k+m data it is arbitrary k in the block when, recover a frame data, when receiving terminal receive it is all
Data frame when, just complete a data transmission.
2. a kind of RDMA network data transmission methods based on correcting and eleting codes according to claim 1, which is characterized in that described
Step S1 includes：Transmitting terminal copies the data in application memory in buffering area according to frame sign to one by one, and is this frame number
According to one frame number of distribution, one can all be added by often sending a frame data frame number；Then it is k number according to block by a frame data cutting, with k
These data blocks are concurrently sent to receiving terminal by QP.
3. a kind of RDMA network data transmission methods based on correcting and eleting codes according to claim 1, which is characterized in that described
Step S2 includes：Transmitting terminal is that a timer is arranged in each frame data, correcting and eleting codes is calculated to data frame after time-out, by check block
It is sent to receiving terminal.
4. a kind of RDMA network data transmission methods based on correcting and eleting codes according to claim 3, which is characterized in that described
Step S2 is specially：Timer module checks whether the data frame that each sends is overtime, if k number also has in after time-out
Receiving terminal is not successfully arrived at, then this frame data is encoded by correcting and eleting codes module, m check block is generated, then calls
Data transmission module sends it to receiving terminal with other m QP.
5. a kind of RDMA network data transmission methods based on correcting and eleting codes according to claim 1, which is characterized in that described
Step S3 includes：After receiving terminal receives a data block, operated immediately by the Write with immediate of RDMA
Number identifies which data frame it belongs to, and knows it is which of this data frame data block according to the QP of the data block of reception
Or check block passes through correcting and eleting codes when the data block for being collected into a data frame adds the sum of check block to reach k
Decoding obtains the original contents of this frame data, finally this frame data is added in the queue of data accepted, waiting is answered
It is read with program.
6. a kind of RDMA network data transmission methods based on correcting and eleting codes according to claim 1, which is characterized in that also wrap
It includes：Receiving terminal reads the data frame having been received according to frame number one by one, and receiving terminal is connect down with a variable read_pos record
Carry out the frame number for the data frame to be read, read data and be divided into block type and non-block type, application program is always etc. under block type
The data frame for being read_pos until occurring frame number in the data queue received is waited for, under non-blocking mode, if having connect
There is no the data frame that frame number is read_pos just directly to return in the data queue of receipts, reception buffering area of the application program from system
After having read a frame data, by read_pos plus one, next frame data are then read, until the buffering area of user is full or meets
To the last frame data of a data transmission.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810487054.3A CN108631947B (en) | 2018-05-21 | 2018-05-21 | RDMA (remote direct memory Access) network data transmission method based on erasure codes |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810487054.3A CN108631947B (en) | 2018-05-21 | 2018-05-21 | RDMA (remote direct memory Access) network data transmission method based on erasure codes |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108631947A true CN108631947A (en) | 2018-10-09 |
| CN108631947B CN108631947B (en) | 2021-06-25 |
Family
ID=63693962
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810487054.3A Active CN108631947B (en) | 2018-05-21 | 2018-05-21 | RDMA (remote direct memory Access) network data transmission method based on erasure codes |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108631947B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109412755A (en) * | 2018-11-05 | 2019-03-01 | 东方网力科技股份有限公司 | A kind of multimedia data processing method, device and storage medium |
| CN109861967A (en) * | 2018-12-10 | 2019-06-07 | 中国人民银行清算总中心 | Remote direct memory based on Spark Shuffle accesses system |
| CN110113425A (en) * | 2019-05-16 | 2019-08-09 | 南京大学 | A kind of SiteServer LBS and equalization methods based on the unloading of RDMA network interface card correcting and eleting codes |
| CN111782609A (en) * | 2020-05-22 | 2020-10-16 | 北京和瑞精准医学检验实验室有限公司 | Method for rapidly and uniformly fragmenting fastq file |
| CN113055434A (en) * | 2021-02-02 | 2021-06-29 | 浙江大华技术股份有限公司 | Data transmission method, electronic equipment and computer storage medium |
| WO2023018779A1 (en) * | 2021-08-13 | 2023-02-16 | Intel Corporation | Remote direct memory access (rdma) support in cellular networks |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101282300A (en) * | 2008-03-03 | 2008-10-08 | 北京航空航天大学 | Method for processing HTTP packet based on non-blockage mechanism |
| US9495324B2 (en) * | 2012-03-30 | 2016-11-15 | Intel Corporation | Efficient distribution of subnet administration data over an RDMA network |
| CN106227617A (en) * | 2016-07-15 | 2016-12-14 | 乐视控股（北京）有限公司 | Self-repair method and storage system based on correcting and eleting codes algorithm |
| CN107070923A (en) * | 2017-04-18 | 2017-08-18 | 上海云熵网络科技有限公司 | Reduce P2P live broadcast systems and method that coding fragment is repeated |
| CN107276722A (en) * | 2017-06-21 | 2017-10-20 | 北京奇艺世纪科技有限公司 | A kind of data transmission method and system based on UDP |
| CN107623646A (en) * | 2017-09-06 | 2018-01-23 | 华为技术有限公司 | Data flow transmission method, send equipment and receiving device |
-
2018
- 2018-05-21 CN CN201810487054.3A patent/CN108631947B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101282300A (en) * | 2008-03-03 | 2008-10-08 | 北京航空航天大学 | Method for processing HTTP packet based on non-blockage mechanism |
| US9495324B2 (en) * | 2012-03-30 | 2016-11-15 | Intel Corporation | Efficient distribution of subnet administration data over an RDMA network |
| CN106227617A (en) * | 2016-07-15 | 2016-12-14 | 乐视控股（北京）有限公司 | Self-repair method and storage system based on correcting and eleting codes algorithm |
| CN107070923A (en) * | 2017-04-18 | 2017-08-18 | 上海云熵网络科技有限公司 | Reduce P2P live broadcast systems and method that coding fragment is repeated |
| CN107276722A (en) * | 2017-06-21 | 2017-10-20 | 北京奇艺世纪科技有限公司 | A kind of data transmission method and system based on UDP |
| CN107623646A (en) * | 2017-09-06 | 2018-01-23 | 华为技术有限公司 | Data flow transmission method, send equipment and receiving device |
Non-Patent Citations (1)
| Title |
|---|
| ZHUZHONG QIAN, BAOLIU YE, SANGLU LU等: "RDMAvisor: Toward Deploying Scalable and Simple RDMA as a Service in", 《ARXIV:1802.01870V1》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109412755A (en) * | 2018-11-05 | 2019-03-01 | 东方网力科技股份有限公司 | A kind of multimedia data processing method, device and storage medium |
| CN109412755B (en) * | 2018-11-05 | 2021-11-23 | 东方网力科技股份有限公司 | Multimedia data processing method, device and storage medium |
| CN109861967A (en) * | 2018-12-10 | 2019-06-07 | 中国人民银行清算总中心 | Remote direct memory based on Spark Shuffle accesses system |
| CN110113425A (en) * | 2019-05-16 | 2019-08-09 | 南京大学 | A kind of SiteServer LBS and equalization methods based on the unloading of RDMA network interface card correcting and eleting codes |
| CN111782609A (en) * | 2020-05-22 | 2020-10-16 | 北京和瑞精准医学检验实验室有限公司 | Method for rapidly and uniformly fragmenting fastq file |
| CN111782609B (en) * | 2020-05-22 | 2023-10-13 | 北京和瑞精湛医学检验实验室有限公司 | Method for rapidly and uniformly slicing fastq file |
| CN113055434A (en) * | 2021-02-02 | 2021-06-29 | 浙江大华技术股份有限公司 | Data transmission method, electronic equipment and computer storage medium |
| CN113055434B (en) * | 2021-02-02 | 2022-07-15 | 浙江大华技术股份有限公司 | Data transmission method, electronic equipment and computer storage medium |
| WO2023018779A1 (en) * | 2021-08-13 | 2023-02-16 | Intel Corporation | Remote direct memory access (rdma) support in cellular networks |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108631947B (en) | 2021-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108631947A (en) | A kind of RDMA network data transmission methods based on correcting and eleting codes | |
| Biersack | Performance evaluation of forward error correction in an ATM environment | |
| JP5828966B2 (en) | Method, apparatus, system, and storage medium for realizing packet transmission in a PCIE switching network | |
| US9049218B2 (en) | Stateless fibre channel sequence acceleration for fibre channel traffic over Ethernet | |
| US5583859A (en) | Data labeling technique for high performance protocol processing | |
| US6675346B2 (en) | Code transmission scheme for communication system using error correcting codes | |
| JP4712039B2 (en) | System and method for higher throughput over a transport network | |
| CN107276722A (en) | A kind of data transmission method and system based on UDP | |
| CN107948094A (en) | A kind of high speed data frame Lothrus apterus is joined the team the device and method of processing | |
| TW200814672A (en) | Method and system for a user space TCP offload engine (TOE) | |
| WO1995010150A1 (en) | Method and apparatus for connecting a node to a wireless network using a standard protocol | |
| JPH06223011A (en) | Computer and method data transfer | |
| CN1985492B (en) | Method and system for supporting iSCSI read operations and iSCSI chimney | |
| KR20050065630A (en) | Reducing packet drop in ieee 802.11 handoff by packet forwarding using imaged driver queue | |
| US6122275A (en) | Real-time processing for virtual circuits in packet switching | |
| CN111522656A (en) | Edge calculation data scheduling and distributing method | |
| CA2187833C (en) | System and method for a scalable and reliable transmission of electronic software distribution | |
| US7839798B2 (en) | Seamlessly switching overlay network relay trees | |
| Feldmeier | A survey of high performance protocol implementation techniques | |
| CN114401208B (en) | Data transmission method and device, electronic equipment and storage medium | |
| CN113179228B (en) | Method, device, equipment and medium for improving switch stacking reliability | |
| CN107707546A (en) | A kind of packet copy method and the new network adapter for supporting message copy function | |
| WO2019015487A1 (en) | Data retransmission method, rlc entity and mac entity | |
| CN112954405A (en) | Data packet recovery method, device and system | |
| CN107196860A (en) | The method that network code carries out multicast transmission in SDN |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20190722 Address after: 210093 Nanjing, Gulou District, Jiangsu, No. 22 Hankou Road Applicant after: Nanjing University Applicant after: Zhejiang Electric Power Co., Ltd. Applicant after: NARI Group Co. Ltd. Address before: 210093 Nanjing, Gulou District, Jiangsu, No. 22 Hankou Road Applicant before: Nanjing University |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |