CN111770152B - Edge data management method, medium, edge server and system - Google Patents
Edge data management method, medium, edge server and system Download PDFInfo
- Publication number
- CN111770152B CN111770152B CN202010588987.9A CN202010588987A CN111770152B CN 111770152 B CN111770152 B CN 111770152B CN 202010588987 A CN202010588987 A CN 202010588987A CN 111770152 B CN111770152 B CN 111770152B
- Authority
- CN
- China
- Prior art keywords
- edge
- copy
- data
- edge server
- network
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000013523 data management Methods 0.000 title claims abstract description 19
- 238000004364 calculation method Methods 0.000 claims abstract description 32
- 238000013508 migration Methods 0.000 claims abstract description 13
- 230000005012 migration Effects 0.000 claims abstract description 13
- 239000011159 matrix material Substances 0.000 claims description 112
- 239000013598 vector Substances 0.000 claims description 20
- 238000012937 correction Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 7
- 230000004044 response Effects 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 4
- 230000000875 corresponding effect Effects 0.000 description 65
- 238000004891 communication Methods 0.000 description 13
- 230000008569 process Effects 0.000 description 12
- 230000001186 cumulative effect Effects 0.000 description 8
- 238000012545 processing Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 230000006872 improvement Effects 0.000 description 5
- 230000006399 behavior Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 4
- 238000013500 data storage Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000013499 data model Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 238000013467 fragmentation Methods 0.000 description 1
- 238000006062 fragmentation reaction Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1097—Protocols in which an application is distributed across nodes in the network for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1004—Server selection for load balancing
- H04L67/101—Server selection for load balancing based on network conditions
Abstract
The embodiment of the invention provides an edge data management method, a medium, edge servers and a system, wherein in the system, each edge server respectively determines the optimal storage position of each stored copy, and the edge servers use a local edge service network which is formed by all edge servers in a cooperative network corresponding to a data table to which the copies belong and edge servers which have accessed the copies outside the cooperative network as the calculation range of copy migration, and calculate the access weight of each edge server accessing the copies on the basis of accumulated time delay, wherein the edge server corresponding to the maximum access weight is the optimal storage position of the copies in the local edge service network, and stores each copy to the corresponding optimal storage position, thereby effectively reducing the user data access time delay; meanwhile, the method is low in computation complexity, occupation of precious computing resources of the edge server can be reduced, and user experience is improved.
Description
Technical Field
The present invention relates to the field of information technology, in particular to the field of data collaboration and sharing in large-scale edge intelligent computing systems, and more particularly to an edge data management method, medium, edge server and system.
Background
With the development of mobile internet and sensor network technologies, the world of everything interconnection is coming. In recent years, the number of terminal devices of the internet of things is greatly increased, and according to prediction of a well-known internet of things research institution IoT Analytics, the number of globally active internet of things devices in 2020 reaches 100 hundred million, and by 2025, the number of globally active internet of things devices reaches 220 hundred million; according to the forecast of Enterprise CIO, the global Internet of things market will grow to 4570 hundred million in 2020, and the annual composite growth rate will reach 28.5%. With the great improvement of the storage capacity and the sensing capacity of the terminal equipment, how to efficiently realize the processing and analysis of mass data becomes an irrevocable problem. If the data processing is still carried out in a mode of transmitting data back to the cloud data center, not only a large amount of precious network communication resources are occupied, but also the timeliness of data processing analysis cannot meet the requirement of the service quality of the user.
To solve this problem, scholars from the industry and academia have proposed the concept of edge computation around 2015, which is a new computing model for performing computing tasks at the network edge. After about five years of development, edge computing has entered a rapid development period, and compared with a cloud computing centralized technical architecture, edge computing mainly has three primary technical advantages: firstly, the timeliness of data processing is improved, a service request is processed at a position closer to a user and a result is returned, data do not need to be transmitted back to a cloud computing center, network delay is reduced, and service quality is improved; secondly, the consumption of data transmission resources is reduced, and data is analyzed and processed at the edge end, so that the consumption of network resources of operators can be effectively reduced, and the pressure of a cloud data center is further reduced; and finally, the data analysis safety is improved, the private data of the user can be transmitted in a more limited range, the private data do not need to be uploaded to a cloud data center, the risk of data leakage is effectively reduced in the transmission link or the storage link, and the data privacy of the user is ensured.
The current architecture of the edge service network focuses more on the problem of how to efficiently execute a single task or an intelligent task of a single user. However, the elements that make up the edge intelligent network, whether vehicles, buildings, or citizens, are decentralized, mobile, and interconnected. The intelligent computing units at the ends realize interconnection through an edge computing backbone network constructed by an operator, can realize interconnection and cooperation of a plurality of terminal devices through a task cooperation platform, and simultaneously realizes cross-domain sharing of an algorithm model, computing resources and data resources which are necessary for an intelligent task. As shown in fig. 1, the cloud data center is connected to an edge node (edge server) through a network, and provides computing resources, communication resources, and data resources for a user through the edge node. The framework can greatly increase the flexibility of the edge intelligent computing infrastructure and provide the infrastructure for richer application scenes.
An Edge Server Network (Edge Server Network, also called Edge Server Network) shown in fig. 2 is a computer Network composed of a plurality of Edge servers (Edge servers, abbreviated as ES) which are distributed in different areas, physically isolated, connected through a Network, and capable of data communication. ES1, ES2, ES3, ES4, ES5 represent 5 edge servers in an exemplary edge services network, respectively. D12, D13, D14, D15, D23, D24, D35 represent communication distances of two edge servers connected. The edge server can execute intelligent operation tasks, has computing and storage resources and can provide intelligent computing services for the outside. In general, an edge server is used in cooperation with a communication base station, performs data communication through the base station, provides an intelligent operation service to a neighboring mobile user terminal, and stores data executed by an intelligent operation task in a server of an edge server network through a distributed database. In summary, the servers in the edge server network have 3 main functions: intelligent computing services, user data storage and user data communication are provided. The geographical location where the edge servers are placed is generally unchanged, i.e., the distance between the edge servers remains fixed, and the communication delay between the edge servers and the edge servers is generally positively correlated to the distance. The storage of data is typically achieved by a distributed database. In order to improve the storage efficiency of Data, mass Data is generally stored in a fragmentation manner, as shown in fig. 3, a Data Table (Data Table) includes metadata (Meta Data) and Data blocks (blocks), where the metadata is used to describe basic information of the Table, and specific Data contents are respectively stored in different Data blocks, such as Data blocks Block0001, Block0002, Block0003, Block0004, Block0005, Block0006, Block0007, Block0008, and Block0009 in fig. 3. To ensure the reliability of data storage and access, multiple copies of a data fragment are typically stored, e.g., in a distributed fashionFIG. 3 shows that each of the data blocks Block-0001-0003 has 2 copies, namely Block-0001-1, Block-0001-2, Block-0002-1, Block-0002-2, Block-0003-1 and Block-0003-2, wherein the copies Block-0001-1, Block-0002-1 and Block-0003-1 are stored in ES1In the method, the copies Block-0001-2, Block-0002-2 and Block-0003-2 are stored in the ES2In (1).
Due to the outstanding advantages of edge computing in architecture design, artificial intelligence tasks for computing on the cloud sink towards the edge side continuously, better service experience is brought to users, and further development of edge intelligence technology is promoted. Edge intelligence is a technical system integrating intelligent computing, network communication, big data storage and the like. However, due to the limited edge computing resources, it is not possible to sink all the intelligent computing tasks on the cloud to the edge side for operation.
With the improvement of computing power of intelligent chips and the continuous improvement of matching technologies such as network communication, data storage and the like, students are always dedicated to realizing full stack migration of an intelligent computing system on a cloud to an edge side. The application scenes of the edge intelligent service mainly comprise intelligent traffic, intelligent communities, mobile entertainment and the like, and the intelligent processing requirements mainly comprise the aspects of image recognition, target detection, voice recognition, sensing data fusion and the like. At present, the intelligent processing task executed at the edge end mainly takes the training and reasoning of various deep neural networks as the main task. Taking a smart city as an example, the architecture of an edge intelligent service system mainly comprises edge side and object side intelligent computing units, and can realize network communication and data interaction with a cloud computing center.
However, there are challenges to implementing efficient edge intelligence services, mainly with limited computing and storage resources. Compared with the computing power of a cloud computing center, the difference of magnitude order exists in the edge computing resources; in addition, some edge computing units may be deployed in an environment without power supply infrastructure, and the computing power available to the edge computing units per unit time is more limited in consideration of the computing power energy efficiency ratio.
The edge service network mainly provides intelligent computing service to the outside, and a user communicates with an adjacent edge server through an intelligent terminal such as a mobile phone and receives the service. In the process of receiving service, a copy of the data block in the edge services network needs to be accessed. Over time, a copy may be accessed multiple times by different users. If the copy is just accessed by the user adjacent to the edge server storing the copy, the time delay of the access only comprises retrieval time delay, but if the user far away from the edge server storing the copy accesses, the user requests the edge server for data through the base stations of other edge servers, and additional time delay of data transmission is needed. Thus, the latency of accessing the copy at different edge servers may be different. If the copy is not stored in its optimal storage location, a high latency for most users to access the copy may occur, which may cause unreliable quality of service. It can be seen that an important factor for improving the service quality of the edge intelligent service is to store the copy of the data block in an optimal storage location of the edge service network, so as to reduce the time delay for accessing the copy. In the existing method, the optimal storage position of the copy is solved by using a data model through a complex operation process, and the solving process generally needs longer calculation time, so that precious calculation resources of the edge server are excessively occupied, and further the user experience is influenced.
Disclosure of Invention
Accordingly, an object of the present invention is to overcome the above-mentioned drawbacks of the prior art and to provide an edge data management method, medium, edge server and system.
The purpose of the invention is realized by the following technical scheme:
according to a first aspect of the present invention, an edge data management method for managing storage locations of copies of data blocks of a data table in a global edge service network, includes: in response to a request to adjust the storage location of the copies, the edge server performs the following for each copy it stores: s1, identifying, from the global edge service network, a collaborative network corresponding to the data table to which the copy belongs and a local edge network corresponding to the copy, where the collaborative network corresponding to the data table is a network formed by all edge servers storing copies of at least part of data blocks of the data table, and the local edge network corresponding to the copy is a network formed by all edge servers in the collaborative network corresponding to the data table to which the copy belongs and edge servers outside the collaborative network that have accessed the copy; s2, acquiring the accumulated time delay of each edge server in the local edge service network accessing the copy in a preset statistical time period aiming at the copy, and calculating the access weight of each edge server accessing the copy based on the accumulated time delay; and S3, taking the edge server corresponding to the maximum access weight as the optimal storage position of the copy in the local edge service network, and storing each copy to the corresponding optimal storage position.
In some embodiments of the present invention, the step S1 includes: s11, the edge server generates a table name vector formed by the table names of all the data tables stored by the edge server, wherein the edge server considers that the edge server stores the data table if the edge server stores the copy of at least part of the data blocks of the data table; s12, the edge server makes the number of the edge server and the table name vector form even data, the first element of the even data is the number of the edge server, and the second element of the even data is the table name vector; s13, the edge server broadcasts the sequence even data to the global edge service network and receives the sequence even data broadcast by other edge servers in the global edge service network so that the edge server can obtain the global sequence even data; and S14, the edge server takes the table name of the data table to which the copy belongs as a primary key, and inquires all edge servers which store the copies of at least part of data blocks of the data table corresponding to the table name from the global sequence pair data to form a cooperative network.
In some embodiments of the present invention, the step S2 includes: s21, adjusting the obtained accumulated time delay of each edge server in the local edge service network for accessing the copy to increase the numerical difference of each other, and obtaining the time delay weight of each edge server for accessing the copy; and S22, correcting the time delay weight of each edge server for accessing the copy according to the correction function to obtain the access weight of each edge server for accessing the copy.
In some embodiments of the present invention, the step S21 includes:
s211, constructing a time delay matrix corresponding to the copy in the local edge service network, wherein data in a certain row and a certain column in the time delay matrix represent statistical accumulated time delay generated when the copy is accessed through the row and the edge server corresponding to the row in sequence;
s212, calculating a time delay weight matrix of the optimal copy position based on the time delay matrix, wherein data in a certain row and a certain column in the time delay weight matrix represent time delay weights of the copies accessed by the edge servers corresponding to the row and the column in sequence, and the time delay weight matrix is determined according to the following method:
wherein R represents a delay weight matrix, T represents a delay matrix,
representing the s-th in the delay weight matrixxThe k-th column of the row is,
representing the entire s-th in the delay matrixxLine, n represents the number of edge servers in the local edge services network,
is expressed in the s-th of the accumulated time delay matrixxThe maximum value in the row is the maximum value,
s-th representing an accumulated time delay matrixxCumulative time delay, r, of the line k column recordjTo represent
At the s-th of the accumulated delay matrixtIn-line sorting according to numerical value from small to largeThe order of the sorting of (a) and (b),
after the calculation, the numerical value of the row corresponding to the edge server outside the cooperative network of the time delay weight matrix is set as a negative value so as to reduce the probability that the copy migrates outside the cooperative network.
Preferably, the negative value is-1.
In some embodiments of the present invention, the step S22 includes:
s221, constructing a target matrix and initializing all numerical values of the target matrix to be zero, wherein data in a certain row and a certain column of the target matrix represent calculated access weights sequentially accessing the copy through the column and the edge server corresponding to the row, and the target matrix Q is iteratively calculated according to the time delay weight matrix in the following mode until convergence is reached to obtain a final target matrix:
wherein Q denotes an object matrix, Q(s)x,ax) Representing s in the object matrix QxLine axColumn value, R(s)x,ax) Representing s in a delay matrix RxLine axThe value of the column is such that,
representing a correction function, gamma representing a correction coefficient,
s representing the target matrix Qx+1All columns of rows
Maximum value of, sx+1＝ax；
And S222, taking the edge server corresponding to the column with the maximum access weight in the row corresponding to the edge server which currently stores the copy in the final target matrix Q as the optimal storage position of the copy.
Preferably, the correction coefficient γ has a value range of [0.5,1 ].
In some embodiments of the invention, the following condition is met to determine that the iterative computation has converged: and after a certain iterative computation is updated, the change amplitude of all the data in the target matrix is smaller than the change amplitude before updating by a preset amplitude threshold value.
Preferably, the preset amplitude threshold is 1%.
In some embodiments of the present invention, the step S3 includes: and when the optimal storage position corresponding to the copy is different from the current storage position of the copy, executing copy migration operation to migrate the copy to the optimal storage position in the cooperative network to which the data block corresponding to the copy belongs, otherwise, not executing the copy migration operation.
In some embodiments of the invention, the method further comprises: setting an adjusting period for adjusting the storage position of the copy, determining the adjusting time for adjusting the storage position of the copy next time according to the adjusting period, and sending a request for adjusting the storage position of the copy when the adjusting time is reached.
According to a second aspect of the present invention, there is provided an edge server comprising: one or more processors; and a memory, wherein the memory is to store one or more executable instructions; the one or more processors are configured to implement the steps of the method of the first aspect via execution of the one or more executable instructions.
According to a third aspect of the present invention, an edge service system includes a plurality of edge servers, where the plurality of edge servers form a global edge service network, an adjustment period for adjusting a copy storage location is set in the edge service system, an adjustment time for adjusting the copy storage location next time is determined according to the adjustment period, and when the adjustment time is reached, the edge service system issues a request for adjusting the copy storage location, where each edge server in the global edge service network performs the steps of the method according to the first aspect in response to the request for adjusting the copy storage location.
Compared with the prior art, the invention has the advantages that:
in the invention, each edge server respectively determines the optimal storage position of each stored copy, the edge servers use all edge servers in a cooperative network corresponding to a data table to which the copy belongs and a local edge service network formed by the edge servers accessing the copy outside the cooperative network as the calculation range of copy migration, and calculate the access weight of each edge server accessing the copy based on accumulated time delay, wherein the edge server corresponding to the maximum access weight is the optimal storage position of the copy in the local edge service network, and stores each copy to the corresponding optimal storage position, thereby effectively reducing the user data access time delay; meanwhile, the method is low in computation complexity, occupation of precious computing resources of the edge server can be reduced, and user experience is improved.
Drawings
Embodiments of the invention are further described below with reference to the accompanying drawings, in which:
fig. 1 is a schematic diagram of various resources provided by a cloud data center and edge nodes for a user;
FIG. 2 is a schematic diagram of an edge service network;
FIG. 3 is a schematic diagram of distributed storage of data tables on an edge server;
FIG. 4 is a flowchart illustrating an edge data management method according to an embodiment of the invention;
fig. 5 is a diagram illustrating a collaborative network to which a data table is determined according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail by embodiments with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As mentioned in the background section, in the existing method, the optimal storage location of the copy is solved by using the data model through a complex operation process, and the solution process generally requires a long calculation time, and occupies too much precious calculation resources of the edge server, thereby affecting the user experience. In the invention, each edge server determines the optimal storage position of each stored copy, the edge servers use a local edge service network formed by all edge servers in a cooperative network corresponding to a data table to which the copy belongs and edge servers accessing the copy outside the cooperative network as the calculation range of copy migration, and calculate the access weight of each edge server accessing the copy based on accumulated time delay, wherein the edge server corresponding to the maximum access weight is the optimal storage position of the copy in the local edge service network, and stores each copy to the corresponding optimal storage position, so that the calculation complexity is low, the occupation of precious calculation resources of the edge servers is reduced, and the user experience is improved. According to the whole calculation process, the process of obtaining the optimal storage position of the copy in the local edge service network adopts a linear operation mode, the calculation amount is small, the optimal storage position of the copy can be obtained more quickly under the condition of occupying less calculation resources of the edge server, the waiting time of a user is further reduced, the service quality of the edge service network is improved, and better experience is provided for the user.
Before describing embodiments of the present invention in detail, some of the terms used therein will be explained as follows:
the data table is a combination form in which data are stored together according to the definition of the table. For example, in some LBS or map service applications, location information of a user, time information corresponding to the location information, and other user behavior information are stored, and these information form a data table, and with the increase of the user amount and the accumulation of time, the data table may have billions of rows of data, which logically form a huge data table, but during storage, the database may segment and store the data into a plurality of blocks, each of which is a data Block (Block). The data block generally includes header information and data information, where the header information describes what data is stored in the data block and a relationship with other data blocks, and the data information, in the above-mentioned example, is location information of a user and time information and other user behavior information corresponding to the location information.
The cooperative network refers to a logic network for performing edge intelligent task cooperative computing, and each distributed stored data table corresponds to one cooperative network.
The invention provides an edge data management method, which is used for managing the storage position of a copy of a data block of a data table in a global edge service network, and comprises the following steps: in response to the request for adjusting the storage location of the copies, the edge server performs steps S1, S2, S3 for each copy it stores (see fig. 4, described in detail below). The user can set an adjustment period for adjusting the storage position of the copy, the edge service system determines the adjustment time for adjusting the storage position of the copy next time according to the adjustment period, and sends a request for adjusting the storage position of the copy when the adjustment time is reached. For a better understanding of the present invention, each step is described in detail below with reference to specific examples.
In step S1, a collaborative network corresponding to the data table to which the copy belongs and a local edge network corresponding to the copy are identified from the global edge service network, where the collaborative network corresponding to the data table refers to a network formed by all edge servers storing copies of at least some data blocks of the data table, and the local edge network corresponding to the copy is a network formed by all edge servers in the collaborative network corresponding to the data table to which the copy belongs and edge servers outside the collaborative network that have accessed the copy.
According to an embodiment of the present invention, step S1 includes: s11, the edge server generates a table name vector formed by the table names of all the data tables stored by the edge server, wherein the edge server considers that the edge server stores the data table if the edge server stores the copy of at least part of the data blocks of the data table; s12, the edge server makes the number of the edge server and the table name vector form even data, the first element of the even data is the number of the edge server, and the second element of the even data is the table name vector; s13, the edge server broadcasts the sequence even data to the global edge service network and receives the sequence even data broadcast by other edge servers in the global edge service network so that the edge server can obtain the global sequence even data; and S14, the edge server takes the table name of the data table to which the copy belongs as a primary key, and inquires all edge servers which store the copies of at least part of data blocks of the data table corresponding to the table name from the global sequence pair data to form a cooperative network. That is, for a data table stored in a distributed manner, the edge servers where all copies of the data table are located will together form a collaborative network corresponding to the data table. Preferably, the number of the edge server can be any unique ID capable of distinguishing the identity of the edge server, for example, an identity ID, an IP address or a MAC address set by a user.
According to an embodiment of the present invention, step S1 includes: step S101: all the stored data tables of the edge server form a table name vector vDT, { dtname1, dtname2, …, dtname }, and dtname represents the nth data table dtname, wherein the edge server communicates with other edge servers in the global edge service network through the network, and the global edge service network stores user data (distributed stored data tables) based on the distributed database; step S102: all edge servers get the table name vector vDT and sum the server number ESjForm a sequence doll<ESj，vDT>(ii) a Step S103: all edge servers broadcast the even-ordered data to ensure that a communicable node (a communicable edge server) in the global edge service network can receive the even-ordered data so as to ensure that the even-ordered data taken by each edge server in the even-ordered data is finally consistent; step S104: all edge servers in the global edge service network traverse the sequence couple data according to the number of the edge server and the table name vector, and find out the edge server vector corresponding to the data table by taking the table name as a main key, wherein the edge server vector corresponding to the data table is a set formed by the edge servers storing copies of at least a part of data blocks of the data table, and a network formed by all the edge servers in the edge server vector is a cooperative network corresponding to the data table.
According to the bookFor an example of the invention, referring to fig. 5, for the process of determining the collaborative network corresponding to the data table, assume that there is a global edge service network composed of J edge servers, which are numbered ES respectivelyjJ is 1,2,3,4,5, …, J, and each edge server has data blocks (blocks) stored thereon, which form the associated data tables dtname1, dtname2, dtname3, dtname. In the metadata (Meta _ Data) of each edge server, a Data block stored in the edge server and a Data table corresponding to the Data block are recorded. In the beginning, all edge servers first obtain the table name vector vDT and the number ES of the edge serverjForm a sequence even data<ESj，vDT>. All edge servers will broadcast this parity data, which all nodes in the edge server network can receive. Finally, all edge server nodes will master all data table distribution in the network, and the sequence even data taken by each server node will be consistent finally. That is, eventually each edge server receives as shown on the left side of FIG. 5<ES1，vDT>、<ES2，vDT>、<ES3，vDT>、<ES4，vDT>、<ES5，vDT>、…、<ESJ，vDT>And then, based on the sequence data and with the table name as a primary key, finding the edge server vectors corresponding to all the edge servers corresponding to the data tables shown on the right side of fig. 5, wherein the network formed by the edge servers included in the edge server vectors is the cooperative network corresponding to the data tables. The table name is used as a primary Key to refer to a Key value pair in a (keys) form, and the primary Key refers to a Key, which means that the table name is used as the primary Key, and a vector formed by edge server numbers is used as a value. For example, the data table dtname1 in fig. 5 uses dtname1 as the primary Key (Key) to find the vector consisting of ES1 and ES3 as values (values). Finally, it can be obtained that the collaborative network corresponding to the data table dtname1 is the edge server ES1、ES3The formed network, the cooperative network corresponding to the data table dtname2 is the edge server ES1、ES2、ES5The formed network, the cooperative network corresponding to data table dtname3For edge server ES2、ES4、ES5The network formed, the cooperative network corresponding to the data table dtnamen is the edge server ES3、ES4、ES5Forming a network.
In step S2, a cumulative delay of each edge server in the local edge service network accessing the copy within a preset statistical period is obtained for the copy, and an access weight of each edge server accessing the copy is calculated based on the cumulative delay.
According to an embodiment of the present invention, step S2 includes: s21, adjusting the obtained accumulated time delay of each edge server in the local edge service network for accessing the copy to increase the numerical difference of each other, and obtaining the time delay weight of each edge server for accessing the copy; and S22, correcting the time delay weight of each edge server for accessing the copy according to the correction function to obtain the access weight of each edge server for accessing the copy. The preset statistical period may be set by the operator as needed according to the data access frequency, for example, to an hour or a day.
Preferably, step S21 includes: s211, constructing a time delay matrix corresponding to the copy in the local edge service network, wherein data in a certain row and a certain column in the time delay matrix represent statistical accumulated time delay generated when the copy is accessed through the row and the edge server corresponding to the row in sequence; s212, calculating a time delay weight matrix of the optimal copy position based on the time delay matrix, wherein data in a certain row and a certain column in the time delay weight matrix represent the time delay weight of the copy accessed by the column and the edge server corresponding to the row in sequence, and the time delay weight matrix is determined in the following way:
wherein R represents a delay weight matrix, T represents a delay matrix, RsxkRepresenting the s-th in the delay weight matrixxThe k-th column of the row is,
representing the entire s-th in the delay matrixxLine, n represents the number of edge servers in the local edge services network,
is expressed in the s-th of the accumulated time delay matrixxThe maximum value in the row is the maximum value,
s-th representing an accumulated time delay matrixxCumulative time delay, r, of the line k column recordjTo represent
At the s-th of the accumulated delay matrixtAnd setting the numerical values of the rows corresponding to the edge servers outside the cooperative network of the time delay weight matrix as negative values after the calculation according to the sorting order of the numerical values from small to large in the rows so as to reduce the probability of the copy migration outside the cooperative network. Preferably, the negative value is-1. The technical scheme of the embodiment can at least realize the following beneficial technical effects: firstly, the formula adopts linear calculation, so that the occupation of the calculation resources of the edge server is reduced; secondly, after the calculation, the numerical value of the row corresponding to the edge server outside the collaborative network of the delay weight matrix is set to be a negative value so as to reduce the probability of the copy migrating outside the collaborative network, thereby increasing the difficulty of the copy migrating to other edge servers outside the collaborative network, not only reducing the delay, but also keeping the current state of the collaborative network as much as possible, avoiding the user data from frequently migrating to different edge servers, and better protecting the user privacy.
According to an example of the present invention, for the delay matrix T, taking the data Block of the data table dtname2 in fig. 5 as an example, it is assumed that the data Block-0001 is included therein, and the cooperative network corresponding to the data table dtname2 is ES1、ES2、ES5Forming a network, and storing a copy of the data Block Block-0001-1 in the ES1The local edge service network corresponding to the copy Block-0001-1 is ES1、ES2、ES3、ES4、ES5A network of components; for the copy Block-0001-1, the edge server ES1A delay matrix T is established as follows:
according to the foregoing example, the collaborative network corresponding to the data table dtname2 in the delay matrix is the edge server ES1、ES2、ES5The edge server network mainly provides intelligent computing service to the outside, in the process of receiving the service, the data blocks in the edge server network need to be accessed, and a certain data block can be accessed by different users for multiple times along with the time, so that accumulated time delay is generated, and t is the time delay11～t15、t21～t25、t31～t35、t41～t45、t51～t55The cumulative latency of accessing a single copy over a preset statistical period is recorded, and the two subscripts of t indicate which edge servers the user accesses the copy. For ease of understanding and simplicity of illustration, only some elements of the delay matrix that are representative and likely to suffer from an understanding bias are selected for illustration. Such as: t is t11Representing a user directly accessing an edge server ES1Cumulative latency of copy Block-0001-1; t is t12Representing users through edge server ES2Access edge server ES1Cumulative latency of copy Block-0001-1; t is t21Theoretically representing users to sequentially pass through edge server ES1And edge server ES2Access edge server ES1Cumulative latency of copy Block-0001-1 above, but because of if the user is going to the edge server ES1A request for accessing the copy Block-0001-1 is initiated and is directly sent from the edge server ES1Accessing the copy Block-0001-1, so that the value counted here is 0 in practical cases; t is t22Representing users through edge server ES2Access edge server ES1Cumulative latency of copy Block-0001-1; t is t32Indicating user turn-by-turn edge server ES2And edge server ES3Access edge server ES1The cumulative delay of the copy Block-0001-1. It should be noted that the accumulated time delay may be an accumulated communication time delay or an accumulated access time delay. The cumulative communication delay includes only the cumulative delay of transmitting the replica. The accumulated access latency includes the accumulated latency of retrieving and transmitting the copy.
According to an example of the present invention, for the latency weight matrix, continuing the previous example, the edge server ES still takes the copy Block-0001-1 of the data Block-0001 in the data table dtname2 as an example1The form of the obtained delay weight matrix is as follows:
for delay weight matrix
The calculation process of (2) is a local edge service network (n is 5) and ES consisting of the above 5 edge servers1In the behavior example, assume that in the T matrix, T11～t15Representation is stored in ES1The copy of the data Block Block-0001-1 is respectively passed through ES by the user1～ES5The statistical results of the accessed cumulative time delays are 6, 8, 10, 3 and 2 seconds in sequence, and the numerical values are 3,4,5, 2 and 1 in sequence from small to large, so that the access time is shortened, and the access time is shortened
A maximum value 10, r representing the cumulative delay in the row11～r15The calculation process of (2) is as follows:
r11＝5×10+3×6＝68；
r12＝5×10+4×8＝72；
r13＝5×10+5×10＝100；
r14＝5×10+2×3＝56；
r15＝5×10+1×2＝52。
it can be seen that the data after mapping calculated by the formula, as shown by example, has more difference and is distributed on the axis 1.5n (maxT)st) 75, and are distributed more discretely on both sides of the axis, so that the subsequent iterative calculation of the target matrix Q can play a role in accelerating convergence.
According to an embodiment of the present invention, step S22 includes: s221, constructing a target matrix and initializing all numerical values of the target matrix to be zero, wherein data in a certain row and a certain column of the target matrix represent calculated access weights sequentially accessing the copy through the column and the edge server corresponding to the row, and the target matrix Q is iteratively calculated according to the time delay weight matrix in the following mode until convergence is reached to obtain a final target matrix:
wherein Q denotes an object matrix, Q(s)x,ax) Representing s in the object matrix QxLine axColumn value, R(s)x,ax) Representing s in a delay matrix RxLine axThe value of the column is such that,
representing a correction function, gamma representing a correction coefficient,
s representing the target matrix Qx+1All columns of rows
Maximum value of, sx+1＝ax(ii) a And S222, taking the edge server corresponding to the column with the maximum access weight in the row corresponding to the edge server which currently stores the copy in the final target matrix Q as the optimal storage position of the copy. Preferably, the correction coefficient γ takes a value in the range of one decimal in [0.5, 1). When the iterative computation of the target matrix is carried out, judging that the iterative computation has converged if the following conditions are met: after a certain iterative computation is updated, the variation amplitude of all data in the target matrix Q is smaller than the preset amplitude before updatingA degree threshold. The preset amplitude threshold is 1%. Preferably, the method further comprises: and setting the time upper limit of iterative computation of the target matrix, and if the target matrix is not converged when the iterative computation time reaches the time upper limit, outputting the current target matrix as a final target matrix. For example, the upper time limit is set to 10 seconds, and the current target matrix is output if the target matrix is not converged after 10 seconds. The technical scheme of the embodiment can at least realize the following beneficial technical effects: firstly, the target matrix constructed by the method is an enhanced learning matrix, the method adopts a linear iteration mode to update the enhanced learning matrix, can efficiently operate on the edge server, has controllable calculation cost, and cannot cause large occupation of system calculation resources due to introduction of the enhanced learning matrix, and the formula of the iterative calculation shows that the iterative calculation adopts a linear equation, is simple in calculation, does not have complex calculation above a quadratic power, has small cost for a calculation chip of the edge server, and cannot solve the problems that the method cannot be executed or the operation efficiency is low when the method is applied to the edge server; secondly, the method has the characteristic of continuous evolution, performance reduction caused by the change of user group behaviors can be avoided, the generalization capability is strong, and the improvement of the user service performance can be quantified and has continuity along with the time; thirdly, the method of the invention has low computational complexity and better expansibility, and can still maintain stable computational performance on a large-scale local edge service network.
According to an example of the present invention, for the target matrix, continuing the previous example, the local edge service network composed of 5 edge servers is still referred to as a copy Block-0001-1 of the data Block-0001 in the data table dtname 2. Step S22 includes: step S201: edge server ES1Establishing an initial target matrix which is a 5 multiplied by 5 matrix, and initializing all numerical values to 0;
step S202: when the cooperative network provides service to the outside, updating a time delay weight matrix based on a statistical time delay matrix is periodically executed;
step S203: and on the basis of the initial target matrix and the time delay weight matrix, performing optimization iterative calculation on the target matrix according to an iterative formula until the target matrix is converged.
According to an example of the present invention, the iterative calculation process of an element in the target matrix is described based on the element Q (1,5) of the target matrix, where R (1,5) is 100 and γ is 0.8, Q (5,1) calculated last time is 2, Q (5,2) is 5, Q (5,3) is 7, Q (5,4) is 10, and Q (5,5) is 4, and Q (1,5) is R (1,5) +0.8 × Max [ Q (5,1), Q (5,2), Q (5,3), Q (5,4), Q (5,5) ], 100+0.8 × 10 is 108 in the current iterative calculation.
In step S3, the edge server corresponding to the maximum access weight is used as the optimal storage location of the copy in the local edge service network, and each copy is stored in its corresponding optimal storage location.
According to one embodiment of the present invention, storing each copy to its corresponding optimal storage location comprises: and when the optimal storage position corresponding to the copy is different from the current storage position of the copy, executing copy migration operation to migrate the copy to the optimal storage position in the cooperative network to which the data block corresponding to the copy belongs, otherwise, not executing the copy migration operation. The optimal storage location corresponding to the copy is time-varying, and step S3 can determine the optimal storage location of the current copy, that is, the relative optimal location in a future period, and if the current copy is not at the optimal storage location, the copy needs to be migrated to the edge server corresponding to the optimal storage location. However, the optimal storage location may change over time, and thus the selection and correction of the optimal copy location may be configured to run hourly or daily to ensure that the user has access to the relevant data with less latency. The reason for the change in the optimal storage location may be that the geographical distribution of the users has changed; the invention aims to dynamically correct the optimal storage position to adapt to different data request distributions and improve the service quality. For the same data block, only one copy is not necessarily reserved in the cooperative network, and there may be a plurality of copies, the upper limit value is set to N, and if the number of copies in the cooperative network after migration is greater than N, the copy at the original position needs to be deleted.
According to an example of the present invention, step S3 includes: step S301: after the target matrix is converged, the edge server corresponding to the column where the maximum value in the row corresponding to the edge server storing the copy in the target matrix is located is the optimal storage position of the copy; step S302: and judging whether the position of the current copy corresponds to the optimal storage position, if so, not migrating, and if not, performing migration operation on the copy to migrate the copy to the optimal storage position. Continuing with the previous example, taking the copy Block-0001-1 of the data Block-0001 in the data table dtname2 as an example, through the above iterative calculation, assume that the edge server ES1In the final target matrix obtained, Q (1,1) is 10, Q (1,2) is 15, Q (1,3) is 25, Q (1,4) is 0, Q (1,5) is 108, wherein the maximum access weight is 108, and the corresponding edge server is ES5If the optimal storage position of the copy Block-0001-1 is the edge server ES5Stored in the edge server ES1A copy Block-0001-1 of the data Block-0001 will be migrated to the edge server ES5The above.
It should be noted that, although the steps are described in a specific order, the steps are not necessarily performed in the specific order, and in fact, some of the steps may be performed concurrently or even in a changed order as long as the required functions are achieved.
The present invention may be a system, method and/or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions embodied therewith for causing a processor to implement various aspects of the present invention.
The computer readable storage medium may be a tangible device that retains and stores instructions for use by an instruction execution device. The computer readable storage medium may include, for example, but is not limited to, an electronic memory device, a magnetic memory device, an optical memory device, an electromagnetic memory device, a semiconductor memory device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a Static Random Access Memory (SRAM), a portable compact disc read-only memory (CD-ROM), a Digital Versatile Disc (DVD), a memory stick, a floppy disk, a mechanical coding device, such as punch cards or in-groove projection structures having instructions stored thereon, and any suitable combination of the foregoing.
Having described embodiments of the present invention, the foregoing description is intended to be exemplary, not exhaustive, and not limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen in order to best explain the principles of the embodiments, the practical application, or improvements made to the technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (14)
1. An edge data management method for managing storage locations of copies of data blocks of a data table in a global edge services network,
the method comprises the following steps: in response to a request to adjust the storage location of the copies, the edge server performs the following for each copy it stores:
s1, identifying, from the global edge service network, a collaborative network corresponding to the data table to which the copy belongs and a local edge network corresponding to the copy, where the collaborative network corresponding to the data table is a network formed by all edge servers storing copies of at least part of data blocks of the data table, and the local edge network corresponding to the copy is a network formed by all edge servers in the collaborative network corresponding to the data table to which the copy belongs and edge servers outside the collaborative network that have accessed the copy;
s2, acquiring the accumulated time delay of each edge server in the local edge service network accessing the copy in a preset statistical time period aiming at the copy, and calculating the access weight of each edge server accessing the copy based on the accumulated time delay;
and S3, taking the edge server corresponding to the maximum access weight as the optimal storage position of the copy in the local edge service network, and storing each copy to the corresponding optimal storage position.
2. The edge data management method according to claim 1, wherein the step S1 includes:
s11, the edge server generates a table name vector formed by the table names of all the data tables stored by the edge server, wherein the edge server considers that the edge server stores the data table if the edge server stores the copy of at least part of the data blocks of the data table;
s12, the edge server makes the number of the edge server and the table name vector form even data, the first element of the even data is the number of the edge server, and the second element of the even data is the table name vector;
s13, the edge server broadcasts the sequence even data to the global edge service network and receives the sequence even data broadcast by other edge servers in the global edge service network so that the edge server can obtain the global sequence even data;
and S14, the edge server takes the table name of the data table to which the copy belongs as a primary key, and inquires all edge servers which store the copies of at least part of data blocks of the data table corresponding to the table name from the global sequence pair data to form a cooperative network.
3. The edge data management method according to claim 2, wherein the step S2 includes:
s21, adjusting the obtained accumulated time delay of each edge server in the local edge service network for accessing the copy to increase the numerical difference of each other, and obtaining the time delay weight of each edge server for accessing the copy;
and S22, correcting the time delay weight of each edge server for accessing the copy according to the correction function to obtain the access weight of each edge server for accessing the copy.
4. The edge data management method according to claim 3, wherein the step S21 includes:
s211, constructing a time delay matrix corresponding to the copy in the local edge service network, wherein data in a certain row and a certain column in the time delay matrix represent statistical accumulated time delay generated when the copy is accessed through the row and the edge server corresponding to the row in sequence;
s212, calculating a time delay weight matrix of the optimal copy position based on the time delay matrix, wherein data in a certain row and a certain column in the time delay weight matrix represent time delay weights of the copies accessed by the edge servers corresponding to the row and the column in sequence, and the time delay weight matrix is determined according to the following method:
wherein R represents a delay weight matrix, T represents a delay matrix,
representing the s-th in the delay weight matrixxThe k-th column of the row is,
representing the entire s-th in the delay matrixxLine, n represents the number of edge servers in the local edge services network,
representing in an accumulated time delay matrixS thxThe maximum value in the row is the maximum value,
s-th representing an accumulated time delay matrixxCumulative time delay, r, of the line k column recordjTo represent
At the s-th of the accumulated delay matrixtThe rows are sorted according to the sorting order of numerical values from small to large,
after the calculation, the numerical value of the row corresponding to the edge server outside the cooperative network of the time delay weight matrix is set as a negative value so as to reduce the probability that the copy migrates outside the cooperative network.
5. The edge data management method of claim 4 wherein the negative value is-1.
6. The edge data management method according to claim 4, wherein the step S22 includes:
s221, constructing a target matrix and initializing all numerical values of the target matrix to be zero, wherein data in a certain row and a certain column of the target matrix represent calculated access weights sequentially accessing the copy through the column and the edge server corresponding to the row, and the target matrix Q is iteratively calculated according to the time delay weight matrix in the following mode until convergence is reached to obtain a final target matrix:
wherein Q denotes an object matrix, Q(s)x,ax) Representing s in the object matrix QxLine axColumn value, R(s)x,ax) Representing s in a delay matrix RxLine axThe value of the column is such that,
representing a correction function, gammaWhich is indicative of the correction factor(s),
s representing the target matrix Qx+1All columns of rows
Maximum value of, sx+1＝ax；
And S222, taking the edge server corresponding to the column with the maximum access weight in the row corresponding to the edge server which currently stores the copy in the final target matrix Q as the optimal storage position of the copy.
7. The edge data management method according to claim 6, wherein the correction coefficient γ has a value in a range of [0.5,1 ].
8. The edge data management method according to claim 6 or 7, wherein the iterative computation is determined to have converged when the following condition is met:
and after a certain iterative computation is updated, the change amplitude of all the data in the target matrix is smaller than the change amplitude before updating by a preset amplitude threshold value.
9. The edge data management method of claim 8 wherein the preset magnitude threshold is 1%.
10. The edge data management method according to any one of claims 1 to 7, wherein the step S3 includes:
and when the optimal storage position corresponding to the copy is different from the current storage position of the copy, executing copy migration operation to migrate the copy to the optimal storage position in the cooperative network to which the data block corresponding to the copy belongs, otherwise, not executing the copy migration operation.
11. The edge data management method according to any one of claims 1 to 7, wherein the method further comprises: setting an adjusting period for adjusting the storage position of the copy, determining the adjusting time for adjusting the storage position of the copy next time according to the adjusting period, and sending a request for adjusting the storage position of the copy when the adjusting time is reached.
12. A computer-readable storage medium, having embodied thereon a computer program, the computer program being executable by a processor to perform the steps of the method of any one of claims 1 to 11.
13. An edge server, comprising:
one or more processors; and
a memory, wherein the memory is to store one or more executable instructions;
the one or more processors are configured to implement the steps of the method of any one of claims 1-11 via execution of the one or more executable instructions.
14. An edge service system, comprising a plurality of edge servers, wherein the plurality of edge servers form a global edge service network, an adjustment period for adjusting the storage location of the copy is provided in the system, an adjustment time for adjusting the storage location of the copy next time is determined according to the adjustment period, and a request for adjusting the storage location of the copy is issued when the adjustment time is reached, and each edge server in the global edge service network performs the steps of the method according to any one of claims 1 to 11 in response to the request for adjusting the storage location of the copy.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010588987.9A CN111770152B (en) | 2020-06-24 | 2020-06-24 | Edge data management method, medium, edge server and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010588987.9A CN111770152B (en) | 2020-06-24 | 2020-06-24 | Edge data management method, medium, edge server and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111770152A CN111770152A (en) | 2020-10-13 |
| CN111770152B true CN111770152B (en) | 2021-07-20 |
Family
ID=72722480
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010588987.9A Active CN111770152B (en) | 2020-06-24 | 2020-06-24 | Edge data management method, medium, edge server and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111770152B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112584439A (en) * | 2020-11-27 | 2021-03-30 | 重庆邮电大学 | Caching method in edge calculation |
| CN113489787B (en) * | 2021-07-06 | 2023-01-17 | 北京邮电大学 | Method and device for collaborative migration of mobile edge computing service and data |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104317669A (en) * | 2014-10-17 | 2015-01-28 | 南京邮电大学 | Cloud heterogeneous storage system and data copy management method thereof |
| CN107463342A (en) * | 2017-08-28 | 2017-12-12 | 北京奇艺世纪科技有限公司 | A kind of storage method and device of CDN fringe nodes file |
| CN109257193A (en) * | 2017-07-11 | 2019-01-22 | 中国移动通信有限公司研究院 | Edge cache management method, personal cloud system and computer readable storage medium |
| CN109982104A (en) * | 2019-01-25 | 2019-07-05 | 武汉理工大学 | The video of mobile awareness prefetches and caching Replacement Decision method in a kind of mobile edge calculations |
| CN110914808A (en) * | 2017-08-04 | 2020-03-24 | 国际商业机器公司 | Copying and migrating files to secondary storage sites |
| CN111104069A (en) * | 2019-12-20 | 2020-05-05 | 北京金山云网络技术有限公司 | Multi-region data processing method and device of distributed storage system and electronic equipment |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6844414B2 (en) * | 2017-05-23 | 2021-03-17 | 富士通株式会社 | Distributed data management program, distributed data management method and distributed data management device |
| CN108121512B (en) * | 2017-12-22 | 2020-04-07 | 苏州大学 | Edge computing service caching method, system and device and readable storage medium |
-
2020
- 2020-06-24 CN CN202010588987.9A patent/CN111770152B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104317669A (en) * | 2014-10-17 | 2015-01-28 | 南京邮电大学 | Cloud heterogeneous storage system and data copy management method thereof |
| CN109257193A (en) * | 2017-07-11 | 2019-01-22 | 中国移动通信有限公司研究院 | Edge cache management method, personal cloud system and computer readable storage medium |
| CN110914808A (en) * | 2017-08-04 | 2020-03-24 | 国际商业机器公司 | Copying and migrating files to secondary storage sites |
| CN107463342A (en) * | 2017-08-28 | 2017-12-12 | 北京奇艺世纪科技有限公司 | A kind of storage method and device of CDN fringe nodes file |
| CN109982104A (en) * | 2019-01-25 | 2019-07-05 | 武汉理工大学 | The video of mobile awareness prefetches and caching Replacement Decision method in a kind of mobile edge calculations |
| CN111104069A (en) * | 2019-12-20 | 2020-05-05 | 北京金山云网络技术有限公司 | Multi-region data processing method and device of distributed storage system and electronic equipment |
Non-Patent Citations (2)
| Title |
|---|
| A Scalable Testbed for Task Offloading and Deployment of Heterogeneous Edge Computing;Boyu Diao 等;《2019 IEEE International Conferences on Ubiquitous Computing & Communications (IUCC) and Data Science and Computational Intelligence (DSCI) and Smart Computing, Networking and Services (SmartCNS)》;20191023;第586-590页 * |
| 移动边缘计算中数据缓存和计算迁移的智能优化技术;汪海霞 等;《中兴通讯技术》;20180317;第15-18页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111770152A (en) | 2020-10-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11410046B2 (en) | Learning-based service migration in mobile edge computing | |
| CN114116198A (en) | Asynchronous federal learning method, system, equipment and terminal for mobile vehicle | |
| CN110728317A (en) | Training method and system of decision tree model, storage medium and prediction method | |
| CN109104464B (en) | Distributed data updating method for collaborative storage in edge computing environment | |
| CN114338504B (en) | Micro-service deployment and routing method based on network edge system | |
| US20170366409A1 (en) | Dynamic Acceleration in Content Delivery Network | |
| CN111770152B (en) | Edge data management method, medium, edge server and system | |
| CN106464669B (en) | Intelligent file prefetching based on access patterns | |
| CN111049903B (en) | Edge network load distribution algorithm based on application perception prediction | |
| CN101873224A (en) | Cloud computing load balancing method and equipment | |
| CN113315978B (en) | Collaborative online video edge caching method based on federal learning | |
| CN113191530B (en) | Block link point reliability prediction method and system with privacy protection function | |
| CN115941790A (en) | Edge collaborative content caching method, device, equipment and storage medium | |
| WO2023207411A1 (en) | Traffic determination method and apparatus based on spatio-temporal data, and device and medium | |
| CN111488528A (en) | Content cache management method and device and electronic equipment | |
| CN102480502A (en) | I/O load equilibrium method and I/O server | |
| CN111044062B (en) | Path planning and recommending method and device | |
| CN111465057B (en) | Edge caching method and device based on reinforcement learning and electronic equipment | |
| CN112492026B (en) | Hybrid self-adaptive copy consistency updating method in dynamic cloud storage environment | |
| Gao et al. | A deep learning framework with spatial-temporal attention mechanism for cellular traffic prediction | |
| CN114039871A (en) | Method, system, device and medium for cellular traffic prediction | |
| CN113301126A (en) | Edge calculation method suitable for heterogeneous networking gateway | |
| WO2022062777A1 (en) | Data management method, data management apparatus, and storage medium | |
| CN115580613A (en) | Mobile edge computing server cooperation caching method based on space-time graph convolution model | |
| CN116228206A (en) | Data center operation and maintenance management method and device, electronic equipment and operation and maintenance management system |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240305 Address after: 518133, 7th Floor, Building 8, Qiugu, Meisheng Huigu Science and Technology Park, No. 83 Dabao Road, Shanghe Community, Xin'an Street, Bao'an District, Shenzhen City, Guangdong Province Patentee after: SHENZHEN GUOKE YIDAO TECHNOLOGY Co.,Ltd. Country or region after: China Address before: 100190 No. 6 South Road, Zhongguancun Academy of Sciences, Beijing, Haidian District Patentee before: Institute of Computing Technology, Chinese Academy of Sciences Country or region before: China |
|
| TR01 | Transfer of patent right |