CN107870813A - A kind of method and device of distributed algorithm processing data - Google Patents
A kind of method and device of distributed algorithm processing data Download PDFInfo
- Publication number
- CN107870813A CN107870813A CN201610843366.4A CN201610843366A CN107870813A CN 107870813 A CN107870813 A CN 107870813A CN 201610843366 A CN201610843366 A CN 201610843366A CN 107870813 A CN107870813 A CN 107870813A
- Authority
- CN
- China
- Prior art keywords
- algorithm
- data
- distributed
- sub
- distributed node
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5083—Techniques for rebalancing the load in a distributed system
Abstract
According to an aspect of the present invention, there is provided a kind of method based on distributed algorithm processing data, this method include the algorithm pre-set being divided into independent sub-step according to the execution sequence of the algorithm process data；By pre-defined rule the sub-step is distributed to each distributed node；Data to be processed will be needed to be handled according to the execution sequence in each distributed node, wherein, the characteristic value after the sub-step processing is transmitted between each distributed node.The present invention to algorithm steps by standardizing, the data processing section of algorithm is placed individually on distributed node and run, only transmission data processing step needs characteristic value rather than total data to be processed between each distributed node, flow is greatly decreased in algorithm motion time cloth system, so as to greatly improve network transmission efficiency and bandwidth utilization rate, systematic function and real-time are improved, strengthens Consumer's Experience.
Description
Technical field
The present invention relates to computer realm, more particularly to a kind of method and device based on distributed algorithm processing data.
Background technology
With the large-scale application of the development of internet artificial intelligence, particularly speech recognition and image recognition, internet
The pressure of data transfer is increasing, by taking picture analyzing in image recognition as an example, the recognition of face in picture analyzing, and imperfect picture
Analysis, the quick-fried probably current application of picture recognition is more and more wider, but the processing of mass picture, particularly to the picture in the case of multitask
Processing, such as a pictures simultaneously to determine whether there is face and whether the demand of imperfect picture, not only needed picture processing requirement but also
It is performance and real-time to need consideration.Picture Algorithm Analysis is computing resource intensive task, although passing through cloud computing at present
Distributed computing technology solves the problems, such as to need to dispose substantial amounts of server.But network is brought to provide after computing resource bottleneck is released
The problem that source can't bear the heavy load, particularly under some real-time scenes, picture and algorithm are distributed on distributed node and run,
When multi-task parallel is handled, it is necessary to by picture distribution to each node, in the case of mass picture, network bandwidth is seriously taken,
Influence the real-time performance of whole system.How cloud computing efficiency of bandwidth use, existing skill improved in cloud computing distributed system
Art does not have solution method also.
The content of the invention
The invention provides a kind of method and device based on distributed algorithm processing data, by by the calculation of data processing
Method is standardized step decomposition, and in data processing, only pass-algorithm needs characteristic value to be processed in a network, and non-primary
Data, efficiency of bandwidth use is greatly improved, to solve the problems, such as that current cloud computing distributed system bandwidth availability ratio is low.
According to an aspect of the present invention, there is provided a kind of method based on distributed algorithm processing data, this method include
The algorithm pre-set is divided into independent sub-step according to the execution sequence of the algorithm process data；Given by pre-defined rule
Each distributed node distributes the sub-step；Data to be processed will be needed according to the execution sequence in each distributed node
Processing, wherein, the characteristic value after the sub-step processing is transmitted between each distributed node.
Further, this method also includes distributing globally unique mark to the algorithm pre-set.
Further, include by pre-defined rule to each distributed node distribution sub-step：The calculating according to needed for the sub-step
Resource allocation distributed node, wherein each distributed node comprises at least a sub-steps
Further, it would be desirable to which the data of processing include according to the execution sequence in the processing of each distributed node：For needing
Data distribution task number to be processed；Execution sequence number and distributed node sign are distributed according to the execution sequence；This is counted
Indicated distributed node processing data is indicated in distributed node according to according to execution sequence number.
Further, for needing to include before data distribution task number to be processed：According to ad hoc rules to the data point
With unique identity；Do not reallocated if the identity repeats with other data identity being stored in database
Task number, terminate handling process.
Further, distributing the sub-step to each distributed node by pre-defined rule is included according to load-balancing algorithm to each
Individual distributed node distributes sub-step
Further, it would be desirable to which the data of processing will most terminate according to the execution sequence after the processing of each distributed node
Fruit is stored in database.
Further, distributing the sub-step to each distributed node by pre-defined rule includes, if distributed node occurs
Failure, the sub-step for distributing to the malfunctioning node is reassigned to other nodes.
According to another aspect of the present invention, a kind of device of distributed algorithm processing data is additionally provided, the device includes
Module is decoupled, for the algorithm pre-set to be divided into independent sub-step according to the execution sequence of the algorithm process data
Suddenly；Distribute module, for distributing the sub-step to each distributed node by pre-defined rule；Scheduler module, for will need to locate
The data of reason are handled according to the execution sequence in each distributed node, wherein, transmitted between each distributed node
Characteristic value after the sub-step processing..
Further, the device also includes database module, for storing algorithm sign, data identity and data warp
The result crossed after algorithm process.
Universal performance of the invention based on data processing algorithm, by being standardized to algorithm steps, at the data of algorithm
Reason part, which is placed individually on distributed node, to be run, and only transmission data processing step needs spy to be processed between each distributed node
Value indicative rather than total data, flow is greatly decreased in algorithm motion time cloth system, so as to greatly improve network transmission effect
Rate and bandwidth utilization rate, systematic function and real-time are improved, strengthen Consumer's Experience.
Brief description of the drawings
Fig. 1 is the method flow diagram of distributed algorithm processing data according to embodiments of the present invention；
Fig. 2 is a kind of device logical construction block diagram based on distributed algorithm processing data of the embodiment of the present invention；
Fig. 3 is a kind of device logic mechanism frame of distributed imperfect picture recognizer processing data of the embodiment of the present invention
Figure；
Fig. 4 is the distributed task management module's logic structure block diagram of the embodiment of the present invention；
Fig. 5 is the imperfect picture process chart of the embodiment of the present invention；
Embodiment
Describe the present invention in detail below with reference to accompanying drawing and in conjunction with the embodiments.It should be noted that do not conflicting
In the case of, the feature in embodiment and embodiment in the application can be mutually combined.
It should be noted that term " first " in description and claims of this specification and above-mentioned accompanying drawing, "
Two " it is etc. for distinguishing similar object, without for describing specific order or precedence.
According to an aspect of the present invention, there is provided a kind of method based on distributed algorithm processing data, Fig. 1 are according to this
The method flow diagram of the distributed algorithm processing data of inventive embodiments, as shown in figure 1, this method includes
S101：The algorithm pre-set is divided into independent sub-step according to the execution sequence of the algorithm process data
Suddenly；
S102：By pre-defined rule the sub-step is distributed to each distributed node；
S103：Data to be processed will be needed to be handled according to the execution sequence in each distributed node, wherein, it is described
The characteristic value after the sub-step processing is transmitted between each distributed node.
By the way that the execution sequence of algorithm process data is divided into each sub-step, by the scattered operation of each sub-step and each point
Cloth node, the characteristic value between node required for a pass-algorithm sub-step processing data, can so be saved to greatest extent
Network traffics about in system, bandwidth availability ratio is improved, improve system real-time response efficiency.
Further, because the device can run many algorithms simultaneously, it is therefore necessary to the algorithm each pre-set
Globally unique mark is distributed, in order to avoid repeated during processing.
Further, utilized to improve the utilization rate of device and balance the computing resource of each distributed node, each sub-step
Suddenly can dynamic according to needed for the sub-step computational resource allocation distributed node, wherein each distributed node comprises at least
One sub-steps；If distributed node breaks down, the sub-step for distributing to the malfunctioning node is reassigned to other
Node.
Further, the needs that device is handled for adaptation multi-task parallel, it would be desirable to which the data of processing are according to the execution
Order includes in the processing of each distributed node：Performed for needing data distribution task to be processed to be distributed according to the execution sequence
Serial number and distributed node sign；The data are indicated into meaning according to the execution sequence number in the distributed node
The distributed node processing data shown.
Further, to avoid the reprocessing of data, the data distribution unique identity is given according to ad hoc rules；
Do not reallocated if the identity and other data identity for being stored in database repeat task number, at end
Manage flow.
Further, final result is stored in database after each distributed node processing, facilitates follow-up utilization.
The embodiment of the present invention additionally provides a kind of device based on distributed algorithm processing data, and Fig. 2 is implementation of the present invention
A kind of device logical construction block diagram based on distributed algorithm processing data of example, as shown in Fig. 2 the device includes partition module
21, for the algorithm pre-set to be divided into independent sub-step according to the execution sequence of the algorithm process data；Distribution
Module 22, for distributing the sub-step to each distributed node by pre-defined rule；Scheduler module 23, it is to be processed for that will need
Data are handled according to the execution sequence in each distributed node, wherein, between each distributed node described in transmission
Characteristic value after sub-step processing.
Sub-step is separately positioned at the distributed node of cluster, subsequent treatment data by the device by being decomposed to algorithm
When in a distributed system a transfer characteristic value to next processing node, network traffics in saving system, improve real-time.
Further, in order to which the data storage after algorithm process, the device are also included into database module, calculated for storing
Method sign, the result of data identity and data after algorithm process.
Picture processing as existing frequently-used Distributed Calculation application, the preferred embodiments of the present invention with picture processing not
Plan deliberately piece is identified as this method and is described further.
Fig. 3 is a kind of device logic mechanism frame of distributed imperfect picture recognizer processing data of the embodiment of the present invention
Figure, the present apparatus as shown in Figure 3 include following module, and picture receiving module, algorithm management module, picture sentence molality block, distributed
Task management module, database module and user's display module.Picture input module 31 is used to be responsible for receiving image data, and deposits
Enter in distributed file system, distribution of notifications formula system has picture processing task to handle.Algorithm management module is used for various
The algorithm groupware of task is managed, and the normal work to do increase, delete, change, looked into, Uniform provisions input/output interface, is additionally operable to figure
The management of piece task, increase or deletion task, and layout is carried out, it is specified that figure according to the order of processing to the algorithm involved by task
The processing sequence of piece, to whether there is algorithm characteristics extraction unit to conduct a compulsory examination.
Fig. 4 is the distributed task management module's logic structure block diagram of the embodiment of the present invention, as shown in figure 4, mainly including
Scheduling unit and scheduling broker unit, for the management of distributed node and task, according to the execution sequence of the execution of algorithm, really
Protect the accurate delivery of output message；The load of calculate node simultaneously, to carry out the distribution of algorithm task, node such as few to load
Increase algorithm task, improve system throughput, while be responsible for the reliability management of system, if node can not service, other sections
Point can rapid taking over tasks.Database module is nucleus module, for storing algorithms library, the related feature of the task of picture and
Task result.User's display module is used for the picture upload and the displaying of result for being responsible for user.
Fig. 5 is the imperfect picture process chart of the embodiment of the present invention, as shown in figure 5, comprising the steps of：Picture is filled
Reception is put, and distribution of notifications formula system is handled；Detect whether to repeat with the picture in system, repeat in then direct returned data storehouse
Result before the picture, otherwise picture can be handled by the feature extraction node of corresponding task.The feature meeting of extraction
It first is put onto data Kuku；The characteristic processing node of next step is then communicated to, carries out the characteristic processing of corresponding task, is handled
As a result it is stored in database.
Imperfect picture Processing Algorithm involved by present specification embodiment, the algorithm of imperfect picture can be with according to execution sequence
It is decomposed into sub-step：1. dimension of picture normalizes the color histogram of 2. pictures；Figure feature extraction and HOG (Histogram of
Oriented Gradient, histograms of oriented gradients) feature extraction 3.KNN improvement searching algorithms；Wherein dimension of picture normalizes
Belong to the sub-step 1 of sub-step, color histogram feature extraction and the HOG feature extractions of picture belong to the sub-step that task associates
Suddenly, although two sub-steps, because the degree of association is high, and need to do particular value extraction process to initial data i.e. picture,
Cross-node can then produce network traffics repetition, and merging into a sub-steps 2, to be encapsulated in operational efficiency on same node higher；And
KNN improves searching algorithm and then belongs to sub-step 3,；The task number for such as assuming imperfect picture retrieval is 1, algorithm numbering as described above
Respectively：Dimension of picture normalizes (0, FALSE, 0, picture, bytes), feature extraction algorithm：(1, TRUE, 1,
Picture, bytes), and the algorithm mark category of characteristic key algorithm (2, TRUE, 1, bytes, bytes) characteristic key algorithms
The implication of property is：Algorithm number is 2, related to task, and association is No. 1 task, and input is byte stream (condition code), and output is word
Symbol string is used to describe whether have imperfect picture and confidence level parameter value.
To imperfect picture retrieval tasks, i.e., the algorithm process order of No. 1 task for the normalization of 1. dimension of pictures (0, FALSE,
0, picture, bytes)；2. feature extraction algorithm (1, TRUE, 1, picture, bytes)；3. and characteristic key algorithm (2,
TRUE, 1, bytes, bytes)；By algorithm orchestration module, imperfect picture is that three algorithm layouts of No. 1 task are as follows, 1. figures
Chip size normalizes ((0, FALSE, 0, picture, bytes), 0,5, FALSE)；2. feature extraction algorithm ((1, TRUE, 1,
Picture, bytes), 1,6, TRUE)；3. characteristic key algorithm ((2, TRUE, 1, bytes, bytes), 2,8, TRUE).Its
In, the task of dimension of picture normalization ((0, FALSE, 0, picture, bytes), 0,5, FALSE) is that the algorithm is No. 1 and appointed
1st execution sub-step of business, performs serial number 0, performing example has 5, and the data after execution are unable to cross-node, that is, after performing
When output data is transmitted to the feature extraction algorithm that Perform sequence is 1, a feature extraction example of the machine can only be transmitted to, it is impossible to across
Node, otherwise flow can be caused to increase；
The specific implementation procedure of dispatching management module is as follows：What the management module of management and running received is task number and algorithm
Perform sequence, be No. 1 task：((0, FALSE, 0, picture, bytes), 0,5, FALSE)-＞ ((1, TRUE, 1,
Picture, bytes), 1,6, TRUE)-＞ ((2, TRUE, 1, bytes, bytes), 2,8, TRUE) assume share 4 nodes,
Then a possible allocative decision is：At least one 0 sequence algorithm thread of each node, and supporting at least one 1 sequence algorithm
Thread, because the algorithm of 0 and 1 sequence must exist in same node simultaneously, ensure that data are unable to cross-node transmission, 0 sequence
The remaining 1 algorithm thread of row and the remaining 2 algorithm threads of 1 sequence can be evenly distributed on lower 4 nodes.And 2 sequence algorithms, not by
The basic principle 1 of scheduling influences, can mean allocation 22 sequences of each node algorithm thread；
After algorithm thread is assigned, in task process is performed, the scheduling agent module of scheduler module, it is necessary to ensure and calculate
The accurate delivery of the output message of method thread；Output such as the algorithm thread of No. 0 sequence on No. 1 node must be delivered to No. 1
The algorithm thread of No. 1 sequence on node gets on, and can be delivered if multiple according to certain regular (such as polling), it is impossible to deliver
Gone in No. 1 sequence on to other nodes；And the output message of No. 1 sequence algorithm on No. 1 node can be delivered to any section
No. 2 sequence algorithm threads on point get on, as long as whole system load balancing；
Before processing is carried out in image data, is first into picture receiving module, usually http uploads to the server；
After the server receives, a message informing can be triggered, the picture of system sentences molality block and (is deployed in the same section of picture receiving module
Point), it can judge whether the picture has existed in systems, using photodna algorithms, if existed in system, can return
The global picture number of the picture is returned, the processing to the picture, then directly returns to the corresponding information stored of the picture number；If the figure
Piece is not present, and can distribute a globally unique picture number.And by picture storage into the distributed file system of system, triggering
The processing of this number picture starts message.
The picture can be handled by the algorithm thread sequence 0 of the imperfect picture of any node in system, and sequence 0 has been handled
Into the output of sequence 0 be sent to the algorithm thread process of the sequence 1 of same node by the scheduling agent module of the node by meeting；
The algorithm thread process of sequence 1 is completed, and output message is delivered in system that other nodes (may also by the proxy module that can be scheduled
This node) sequence 2 algorithm thread process.When scheduling broker delivers message, task number and picture number two can be carried all the time
Individual field, therefore algorithm thread, can by the information handled with picture number for index field, be deposited into database；
Obviously, those skilled in the art should be understood that above-mentioned each module of the invention or each step can be with general
Computing device realize that they can be concentrated on single computing device, or be distributed in multiple computing devices and formed
Network on, alternatively, they can be realized with the program code that computing device can perform, it is thus possible to they are stored
Performed in the storage device by computing device, and in some cases, can be with different from shown in order execution herein
The step of going out or describing, they are either fabricated to each integrated circuit modules respectively or by multiple modules in them or
Step is fabricated to single integrated circuit module to realize.So, the present invention is not restricted to any specific hardware and software combination.
The preferred embodiments of the present invention are the foregoing is only, are not intended to limit the invention, for the skill of this area
For art personnel, the present invention can have various modifications and variations.Within the spirit and principles of the invention, that is made any repaiies
Change, equivalent substitution, improvement etc., should be included in the scope of the protection.
Claims (10)
- A kind of 1. method of distributed algorithm processing data, it is characterised in that including：The algorithm pre-set is divided into independent sub-step according to the execution sequence of the algorithm process data；By pre-defined rule the sub-step is distributed to each distributed node；Data to be processed will be needed to be handled according to the execution sequence in each distributed node, wherein, each distribution The characteristic value after the sub-step processing is transmitted between node.
- 2. according to the method for claim 1, it is characterised in that including：Globally unique mark is distributed to the algorithm pre-set Know.
- 3. according to the method for claim 1, it is characterised in that distribute the sub-step to each distributed node by pre-defined rule Suddenly include：According to computational resource allocation distributed node needed for the sub-step, wherein each distributed node comprises at least a son Step.
- 4. according to the method for claim 3, it is characterised in that will need data to be processed according to the execution sequence each Distributed node processing includes：For needing data distribution task number to be processed；Execution sequence number and distributed node sign are distributed according to the execution sequence；The data are indicated into indicated distributed node in the distributed node according to the execution sequence number and handle number According to.
- 5. according to the method for claim 4, it is characterised in that for needing to include before data distribution task number to be processed：The data distribution unique identity is given according to ad hoc rules；Do not reallocated if the identity and other data identity for being stored in database repeat task number, knot Beam handling process.
- 6. according to the method for claim 1, it is characterised in that distribute the sub-step to each distributed node by pre-defined rule Suddenly include：According to load-balancing algorithm the sub-step is distributed to each distributed node.
- 7. according to the method for claim 1, it is characterised in that will need data to be processed according to the execution sequence in institute State and final result is stored in database after each distributed node is handled.
- 8. according to the method for claim 1, it is characterised in that distribute the sub-step to each distributed node by pre-defined rule Suddenly include：If distributed node breaks down, the sub-step for distributing to the malfunctioning node is reassigned to other nodes.
- A kind of 9. device of distributed algorithm processing data, it is characterised in that including：Module is decoupled, for the algorithm pre-set to be divided into independent son according to the execution sequence of the algorithm process data Step；Distribute module, for distributing the sub-step to each distributed node by pre-defined rule；Scheduler module, for data to be processed will to be needed to be handled according to the execution sequence in each distributed node, wherein, The characteristic value after the sub-step processing is transmitted between each distributed node.
- 10. device according to claim 9, other are also to include：Database module, for storing algorithm sign, the result of data identity and data after algorithm process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610843366.4A CN107870813A (en) | 2016-09-22 | 2016-09-22 | A kind of method and device of distributed algorithm processing data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610843366.4A CN107870813A (en) | 2016-09-22 | 2016-09-22 | A kind of method and device of distributed algorithm processing data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107870813A true CN107870813A (en) | 2018-04-03 |
Family
ID=61750668
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610843366.4A Withdrawn CN107870813A (en) | 2016-09-22 | 2016-09-22 | A kind of method and device of distributed algorithm processing data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107870813A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108874548A (en) * | 2018-07-11 | 2018-11-23 | 深圳市东微智能科技股份有限公司 | Data processing dispatching method, device, computer equipment and data processing system |
| CN111352712A (en) * | 2020-02-25 | 2020-06-30 | 程瑞萍 | Cloud computing task tracking processing method and device, cloud computing system and server |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101291245A (en) * | 2007-04-16 | 2008-10-22 | 阿里巴巴集团控股有限公司 | Distributed processing method, system and device thereof |
| CN102073546A (en) * | 2010-12-13 | 2011-05-25 | 北京航空航天大学 | Task-dynamic dispatching method under distributed computation mode in cloud computing environment |
| JP2013025425A (en) * | 2011-07-19 | 2013-02-04 | Nec Corp | Distributed data management system, distributed data management method, and distributed data management program |
| CN103455518A (en) * | 2012-06-04 | 2013-12-18 | 中兴通讯股份有限公司 | Data processing method and device |
| CN105163380A (en) * | 2015-09-23 | 2015-12-16 | 浙江理工大学 | Distributed cross-layer optimization method for MIMO (Multiple Input Multiple Output) wireless multi-hop network |
-
2016
- 2016-09-22 CN CN201610843366.4A patent/CN107870813A/en not_active Withdrawn
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101291245A (en) * | 2007-04-16 | 2008-10-22 | 阿里巴巴集团控股有限公司 | Distributed processing method, system and device thereof |
| CN102073546A (en) * | 2010-12-13 | 2011-05-25 | 北京航空航天大学 | Task-dynamic dispatching method under distributed computation mode in cloud computing environment |
| JP2013025425A (en) * | 2011-07-19 | 2013-02-04 | Nec Corp | Distributed data management system, distributed data management method, and distributed data management program |
| CN103455518A (en) * | 2012-06-04 | 2013-12-18 | 中兴通讯股份有限公司 | Data processing method and device |
| CN105163380A (en) * | 2015-09-23 | 2015-12-16 | 浙江理工大学 | Distributed cross-layer optimization method for MIMO (Multiple Input Multiple Output) wireless multi-hop network |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108874548A (en) * | 2018-07-11 | 2018-11-23 | 深圳市东微智能科技股份有限公司 | Data processing dispatching method, device, computer equipment and data processing system |
| CN111352712A (en) * | 2020-02-25 | 2020-06-30 | 程瑞萍 | Cloud computing task tracking processing method and device, cloud computing system and server |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105900064B (en) | The method and apparatus for dispatching data flow task | |
| CN109471727A (en) | A kind of task processing method, apparatus and system | |
| US8954497B2 (en) | Parallel distributed processing method and computer system | |
| US7076545B2 (en) | Load balancing the servicing of received packets | |
| CN112486648A (en) | Task scheduling method, device, system, electronic equipment and storage medium | |
| US8434085B2 (en) | Scalable scheduling of tasks in heterogeneous systems | |
| CN114741207B (en) | GPU resource scheduling method and system based on multi-dimensional combination parallelism | |
| US20180246765A1 (en) | System and method for scheduling jobs in distributed datacenters | |
| CN108111554B (en) | Control method and device for access queue | |
| EP3090329A1 (en) | Event-driven data processing system | |
| CN109271243B (en) | Cluster task management system | |
| WO2006123177A1 (en) | Data processing network | |
| CN111464659A (en) | Node scheduling method, node pre-selection processing method, device, equipment and medium | |
| CN111880939A (en) | Container dynamic migration method and device and electronic equipment | |
| CN110287009A (en) | A kind of working node selection method, device, storage medium and server | |
| CN113660231A (en) | Message parsing method, device, equipment and storage medium | |
| CN107870813A (en) | A kind of method and device of distributed algorithm processing data | |
| CN113672500B (en) | Deep learning algorithm testing method and device, electronic device and storage medium | |
| CA2631255A1 (en) | Scalable scheduling of tasks in heterogeneous systems | |
| CN114490086A (en) | Method, device, electronic equipment, medium and program product for dynamically adjusting resources | |
| CN114564313A (en) | Load adjustment method and device, electronic equipment and storage medium | |
| CN114327894A (en) | Resource allocation method, device, electronic equipment and storage medium | |
| CN112307046A (en) | Data acquisition method and device, computer readable storage medium and electronic equipment | |
| CN105162837B (en) | The method and system of I/O throughputs are promoted under mass data storage environment | |
| CN107046581A (en) | A kind of monitoring method, device and the server of service operation state |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20180403 |
|
| WW01 | Invention patent application withdrawn after publication |