CN105630690B - The method and device that big data is handled in isomerization hardware - Google Patents

The method and device that big data is handled in isomerization hardware Download PDF

Info

Publication number
CN105630690B
CN105630690B CN201410608826.6A CN201410608826A CN105630690B CN 105630690 B CN105630690 B CN 105630690B CN 201410608826 A CN201410608826 A CN 201410608826A CN 105630690 B CN105630690 B CN 105630690B
Authority
CN
China
Prior art keywords
host
data
target data
address
address information
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201410608826.6A
Other languages
Chinese (zh)
Other versions
CN105630690A (en
Inventor
解壁伟
高云伟
詹剑锋
张立新
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huawei Technologies Co Ltd
Institute of Computing Technology of CAS
Original Assignee
Huawei Technologies Co Ltd
Institute of Computing Technology of CAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huawei Technologies Co Ltd, Institute of Computing Technology of CAS filed Critical Huawei Technologies Co Ltd
Priority to CN201410608826.6A priority Critical patent/CN105630690B/en
Publication of CN105630690A publication Critical patent/CN105630690A/en
Application granted granted Critical
Publication of CN105630690B publication Critical patent/CN105630690B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The embodiment of the invention discloses the method and devices that big data in a kind of isomerization hardware is handled, it is related to field of communication technology, it can solve isomerization hardware processing big data application, the data of loading and the data needed for program operation are inconsistent, the problem of occupying the memory of isomerization hardware, and influencing the normal operation of program.The method of the present invention includes:Receive program and address information that the Host is sent;Establish the mapping relations of virtual address and described address information that the source data is locally stored;During the isomerization hardware runs described program when invocation target data, target data request is sent to the Host;After receiving the target data that the Host is replied, described program is run using the target data.The present invention is suitable for isomerization hardware.

Description

The method and device that big data is handled in isomerization hardware
Technical field
The present invention relates to the method and devices that big data in field of communication technology more particularly to a kind of isomerization hardware is handled.
Background technology
Isomerization hardware be interconnected by PCI-E (pci Express, universal serial bus) as expansion card and host Host it is hard Part can provide the abilities such as more powerful vector processing, floating-point operation or multi-threaded parallel for Host.Memory on isomerization hardware Smaller and be difficult extension, when handling data, data can not disposably be loaded into different isomerization hardware by the applications of some big datas Structure hardware executes, the case where being much smaller than data processing amount for the memory on this isomerization hardware, and the method for generally use is, will The memory of isomerization hardware is loaded into after big data section in batches, i.e., after last consignment of data are completed in isomerization hardware processing, then by second Batch data is loaded on isomerization hardware, for example, the data of 16G to be divided into the data block of 4 4G, is loaded into isomerization hardware in batches Memory on handled.
At least there are the following problems in the prior art:In the data handling procedure of some big datas application, program logarithm According to demand be it is dynamic, i.e., program operation before not can determine which partial data it needs, so isomerization hardware handles this Big data is planted in application, data to be loaded into isomerization hardware memory in batches, can be made needed for data and the present procedure operation of loading Data it is inconsistent, present procedure, which runs unwanted data and is stored in isomerization hardware, occupies a large amount of memory, not only wastes The memory headroom of isomerization hardware also affects the normal operation of program.
Invention content
The embodiment of the present invention provides the method and device that big data is handled in a kind of isomerization hardware, and it is hard can to solve isomery Part handle this big data in application, data and the present procedure operation for being loaded into isomerization hardware memory it is required data it is inconsistent, Not only waste the memory headroom of isomerization hardware, the problem of also affecting the normal operation of program.
In order to achieve the above objectives, the embodiment of the present invention adopts the following technical scheme that:
In a first aspect, the embodiment of the present invention provides a kind of method that big data is handled in isomerization hardware, isomerization hardware with Host Host connections, including:
Program and address information that the Host is sent are received, described program is the program run on isomerization hardware, institute It is the address information that is stored in the Host of source data to state address information, and the source data is runs institute on the isomerization hardware State the data needed for program;
Establish the mapping relations of virtual address and described address information that the source data is locally stored;
During the isomerization hardware runs described program when invocation target data, target data is sent to the Host Request, the target data request carries the address information of the target data, in order to which the Host is according to the number of targets According to address information search the target data, the address information of the target data is the isomerization hardware according to the mapping The virtual address of relationship and the target data determines that the target data belongs to the source data;
After receiving the target data that the Host is replied, described program is run using the target data.
With reference to first aspect, in the first possible realization method of first aspect, when the source data is stored in institute When stating the memory of Host, described address information is the physical address that the Host stores the source data;
When the source data is stored in the disk of the Host, described address information is that the Host stores the source number According to virtual address.
With reference to first aspect, in second of possible realization method of first aspect, the number of targets is used described Before operation described program, the method further includes:
The target data is stored in the memory of the isomerization hardware;
The mapping being locally stored between the virtual address of the target data and the address information of the target data is closed System is revised as that the virtual address of the target data and target data physics in the isomerization hardware memory is locally stored Mapping relations between address.
Second of possible realization method with reference to first aspect, in the third possible realization method of first aspect In, when the source data is stored in the disk of the Host, the target data is stored in the isomerization hardware described Memory in before, the method further includes:
Receive the physical address that the target data that the Host is replied stores in the memory of the Host, the object Reason address is that the Host is stored in the physical address in the memory of the Host after transferring the target data in disk;
The mapping being locally stored between the virtual address of the target data and the address information of the target data is closed System is revised as being locally stored between the virtual address of the target data and the physical address stored in the memory of the Host Mapping relations;
It is asked to the Host transmitting portions target data, the partial target request of data carries partial target data and exists The physical address stored in the memory of the Host, in order to the Host according to the partial target data the Host's The physical address stored in memory searches the partial target data;
It is then described after receiving the target data that the Host is replied, run described program packet using the target data It includes:
After receiving the partial target data that the Host is replied, the partial target data run described program is used.
With reference to first aspect, in the 4th kind of possible realization method of first aspect, work as the isomerization hardware described During operation described program when invocation target data, before being asked to the Host transmission datas, the method further includes:
It adds and identifies for described address information, the mark expression described address information is the data in the Host The address information of storage.
Second aspect, the embodiment of the present invention provide the device that big data is handled in a kind of isomerization hardware, including:
Receiving unit, the program and address information sent for receiving the Host, described program are on isomerization hardware The program of operation, described address information are the address information that source data stores in the Host, and the source data is described different The data needed for described program are run on structure hardware;
Unit is established, the mapping for establishing virtual address and described address information that the source data is locally stored is closed System;
Request unit, for during running described program when the isomerization hardware when invocation target data, to described Host sends target data request, and the target data request carries the address information of the target data, in order to described Host searches the target data according to the address information of the target data, and the address information of the target data is described different Structure hardware determines that the target data belongs to the source number according to the virtual address of the mapping relations and the target data According to;
It using unit, is used for after receiving the target data that the Host is replied, described in target data operation Program.
In conjunction with second aspect, in the first possible realization method of second aspect, when the source data is stored in institute When stating the memory of Host, described address information is the physical address that the Host stores the source data;
When the source data is stored in the disk of the Host, described address information is that the Host stores the source number According to virtual address.
In conjunction with second aspect, in second of possible realization method of second aspect, described device further includes:
Storage unit, for the target data to be stored in the memory of the isomerization hardware;
Change unit, the address information of virtual address and the target data for the target data will to be locally stored Between mapping relations be revised as that the virtual address of the target data is locally stored and the target data is hard in the isomery Mapping relations in part memory between physical address.
In conjunction with second of possible realization method of second aspect, in the third possible realization method of second aspect In, the receiving unit is additionally operable to receive the object that the target data that the Host is replied stores in the memory of the Host Address is managed, the physical address is that the Host is stored in after transferring the target data in disk in the memory of the Host Physical address;
The modification unit is additionally operable to that the ground of the virtual address and the target data of the target data will be locally stored Mapping relations between the information of location are revised as the virtual address of the target data being locally stored and in the memory of the Host Mapping relations between the physical address of storage;
The request unit is additionally operable to ask to the Host transmitting portions target data, the partial target request of data The physical address that partial target data store in the memory of the Host is carried, in order to which the Host is according to the part mesh The physical address that mark data store in the memory of the Host searches the partial target data;
It is described to be additionally operable to after receiving the partial target data that the Host is replied using unit, use the partial target Data run described program.
In conjunction with second aspect, in the 4th kind of possible realization method of second aspect, described device further includes:
Adding device, for being identified for the addition of described address information, the mark indicates that described address information is the number According to the address information stored in the Host.
In a kind of isomerization hardware provided in an embodiment of the present invention big data handle method and device, the present invention in isomery it is hard The address information that source data stores in Host needed for the program to be run and the program operation that part receiving host Host is sent, ground Location information is the address information of source data storage;The mapping for establishing virtual address and address information that source data is locally stored is closed System;When isomerization hardware runs the target data called in program process in source data, isomerization hardware is according to mapping relations and mesh The virtual address of mark data determines the address information of target data, and the data for carrying target data address information are sent to Host Request, in order to which Host searches target data according to the address information of target data;After receiving the target data that Host is replied, Program is run using target data.In this way, when isomerization hardware handles big data in application, isomerization hardware need not be loaded data into Memory, it is only necessary to be loaded into the address information that data store in Host, while establishing data in LVA local virtual address and address The mapping of information, when isomerization hardware needs to call data during running program, can by LVA local virtual address and Mapping determines the address information that data store in Host, and the request for carrying address information is then sent to Host, is returned in Host After returning data, program is continued to run with using data, both isomerization hardware can be loaded into avoid redundant data, and waste the interior of isomerization hardware It deposits, and does not influence the normal operation of program.
Description of the drawings
It to describe the technical solutions in the embodiments of the present invention more clearly, below will be to needed in the embodiment Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for ability For the those of ordinary skill of domain, without creative efforts, it can also be obtained according to these attached drawings other attached Figure.
Fig. 1 is the method flow diagram that one embodiment of the invention provides;
Fig. 2 is the method flow diagram that further embodiment of this invention provides;
Fig. 3 is the method flow diagram that further embodiment of this invention provides;
Fig. 4, Fig. 5 are the apparatus structure schematic diagram that further embodiment of this invention provides.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts all other Embodiment shall fall within the protection scope of the present invention.
One embodiment of the invention provides a kind of method that big data is handled in isomerization hardware, is used for isomerization hardware, and isomery is hard Part is linked with Host, as shown in Figure 1, method includes:
101, the program and address information that isomerization hardware receiving host Host is sent.
Wherein, program is the program run on isomerization hardware, and described address information is that source data is deposited in the Host The address information of storage, the source data are the data needed for operation described program on the isomerization hardware.
Optionally, when source data is stored in the memory of Host, address information is the physical address that Host stores source data; When source data is stored in the disk of Host, address information is the virtual address that Host stores source data, and void is stored in Host The correspondence for the physical address that quasi- address is stored with source data in disk, Host can be looked by virtual address and correspondence Find the source data in disk.
102, the virtual address of source data and the mapping relations of address information is locally stored in isomerization hardware foundation.
103, when isomerization hardware runs invocation target data in program process, isomerization hardware is asked to Host transmission datas It asks.
Wherein, request of data carries the address information of target data, in order to which Host is according to the address information of target data Search target data.The address information of target data is that isomerization hardware is true according to the virtual address of mapping relations and target data Fixed, target data is a part for source data.
104, after receiving the target data that Host is replied, isomerization hardware runs program using target data.
Optionally, before isomerization hardware is using target data, target data can also be stored in isomery by isomerization hardware In the memory of hardware, then it will be locally stored between the virtual address of the target data and the address information of the target data Mapping relations be revised as being locally stored the virtual address of the target data with the target data in the isomerization hardware Deposit the mapping relations between physical address.In this way, isomerization hardware is in running program process, it would be desirable to data be loaded into itself Memory, avoid extra data, save memory headroom.
Further, when source data is stored in the disk of Host, in the memory that target data is stored in isomerization hardware In before, if the data that isomerization hardware is transferred are larger, target data after disk is transferred, can be first stored in Host by Host Memory in, the physical address of storage is then sent to isomerization hardware, isomerization hardware receives the target data that Host is replied and exists The physical address stored in the memory of Host, and the virtual address of the target data and the target data will be locally stored Mapping relations between address information are revised as that the virtual address of the target data and the memory in the Host is locally stored Mapping relations between the physical address of middle storage, then isomerization hardware from the memory of Host transfer target data by several times, i.e., to Host, which is sent, carries the partial target request of data that partial target data store physical address in the memory of Host, in order to Host searches the partial target data according to the physical address that partial target data store in the memory of Host.
It should be noted that when isomerization hardware transfers the data in disk file, it will usually once to by relevant data It all transfers, but these data are not moved while being used, it is possible to the data disposably transferred first are stored in Host's In memory, called data is being implemented according to the operating condition of program, need not mass data disposably called in into isomerization hardware, kept away Exempt from memory waste.
Optionally, when isomerization hardware invocation target data, before being asked to Host transmission datas, isomerization hardware also needs To be address information addition mark, mark indicates that address information is the address information that data store in Host, isomerization hardware It can determine whether data are stored in the memory of isomerization hardware according to mark.
In a kind of isomerization hardware provided in an embodiment of the present invention big data handle method and device, the present invention in isomery it is hard The address information that source data stores in Host needed for the program to be run and the program operation that part receiving host Host is sent, ground Location information is the address information of source data storage;The mapping for establishing virtual address and address information that source data is locally stored is closed System;When calling the target data in source data during isomerization hardware operation described program, isomerization hardware is according to mapping relations The address information of target data is determined with the virtual address of target data, and is sent to Host and carried target data address information Request of data, in order to which Host searches target data according to the address information of target data;In the number of targets for receiving Host replies According to rear, target data is used to run program.In this way, when isomerization hardware handles big data in application, need not load data into different The memory of structure hardware, it is only necessary to be loaded into the address information that data store in Host, while establish data in LVA local virtual address It can pass through local virtual when isomerization hardware needs to call data during running program with the mapping of address information Address and mapping determine the address information that data store in Host, and the request for carrying address information is then sent to Host, After Host returned datas, program is continued to run with using data, isomerization hardware can be both loaded into avoid redundant data, waste isomery is hard The memory of part, and the normal operation of program is not influenced.
Further embodiment of this invention provides a kind of method that big data is handled in isomerization hardware, big for isomerization hardware processing The scene of data, isomerization hardware are interconnected with Host by PCI-E, and big data is stored in the memory of Host, as shown in Fig. 2, institute The method of stating includes:
201, Host extracts the pointer for the big data for needing to transmit from big data program segment, and is translated into Host On physical address.
Wherein, big data is source data, and big data program segment is to need the program run on isomerization hardware in Host, Big data is data needed for program operation, and Host, which can be extracted by corresponding program from big data program segment, to be needed to pass The pointer for the big data passed, i.e. virtual address, then convert virtual address to physical address, and physical address is in Host memories The address of big data storage.
202, Host creates the state table for safeguarding big data lock status.
Wherein, the state of data lock status illustrates the use state of the Physical Page corresponding to current data, for example, with " 0 " indicates to be used, and " 1 " indicates not use.In follow-up data processing procedure, variation that state table is stored according to data It is updated.
203, the physical address of program and Host storage data is sent to isomerization hardware by Host.
Wherein, Host will need the pointer of data needed for the program run on isomerization hardware and program operation to send To isomerization hardware, isomerization hardware receives its being executed for task that determines after program, after reception pointer, can store number by Host According to physical address obtain data.
204, isomerization hardware creates task after receiving program.
205, isomerization hardware stores the local page table of physical address establishment of data according to Host, establishes virtual in local page table The mapping relations of address and the physical address of data storage on Host.
Wherein, isomerization hardware stores the local page table of physical address establishment of data according to Host, is established in local page table The mapping relations of the virtual address and the physical address of data storage on Host of data are locally stored.
206, isomerization hardware is the physical address addition mark that Host stores data in local page table.
207, when isomerization hardware, which is run, to be needed to call data in program process, it is hard to judge whether the data are stored in isomery The memory of part obtains data if the data are stored in the memory of isomerization hardware from memory, executes step 214;When the data It is not stored in the memory of isomerization hardware, thens follow the steps 208.
Wherein, isomerization hardware can learn that the data whether there is in different by judging that the address in local page table identifies In the physical memory of structure hardware.
208, isomerization hardware generates interruption, is asked to Host transmission datas.
Wherein, isomerization hardware can obtain the data depositing in Host according to the mapping relations recorded from local page table Storage address carries the physical address that the data of acquisition store in Host when being asked to Host transmission datas.
209, after Host receives request of data, judge whether the address of the data is lock-out state by lock status table, if The address of the data is not lock-out state, thens follow the steps 210;If the address of the data is lock-out state, step is continued to execute Rapid 209.
Wherein, after Host receives request of data, judge whether the corresponding Physical Page of the data is locking by lock status table State, if the corresponding Physical Page of the data is lock-out state, the address of the data is lock-out state, needs to wait for its unlock, Continue to execute step 209;If the corresponding Physical Page of the data is not lock-out state, the address of the data is not lock-out state, Then follow the steps 210.
210, Host judges whether isomerization hardware is copy to the request method of data, if request of the isomerization hardware to data Mode is copy, thens follow the steps 211;If isomerization hardware is not copy to the request method of data, 212 are thened follow the steps.
Wherein, isomerization hardware can be copy to the request method of data, can also be to call in.
211, Host returns to its data asked to isomerization hardware.
It should be noted that after having executed this step, step 214 is executed.
212, Host returns to its data asked to isomerization hardware, and discharges the memory of data occupancy.
213, isomerization hardware storage receives the data that Host is returned, and changes local page table.
Wherein, it after isomerization hardware receives the data that Host is returned, stores data in local memory, and by native page The storage address of data is revised as the address in local memory in table, cancels mark.
214, isomerization hardware continues to execute program using the data of calling.
In the prior art, isomerization hardware handles this big data in application, data to be loaded into isomerization hardware memory in batches, Data and the data needed for program operation of loading can be made inconsistent, in this way, not only taking up the memory of isomerization hardware, can also be influenced To the normal operation of program.Compared with prior art, when isomerization hardware handles big data in application, need not be by number in the present invention According to the memory for being loaded into isomerization hardware, it is only necessary to be loaded into the address information that data store in Host, while establish data in local The mapping of virtual address and address information can pass through when isomerization hardware needs to call data during running program LVA local virtual address and mapping determine the address information that data store in Host, are then sent to Host and carry address information Request, after Host returned datas, program is continued to run with using data, and isomerization hardware, wave can be both loaded into avoid redundant data Take the memory of isomerization hardware, and does not influence the normal operation of program.
Further embodiment of this invention provides a kind of method that big data is handled in isomerization hardware, big for isomerization hardware processing The scene of data, isomerization hardware are interconnected with Host by PCI-E, the disk text that big data is mapped by maps virtual page mmap Part, big data are source data, i.e., the data only have virtual address in Host, are not stored in Host memories, but are stored In disk file, as shown in figure 3, the method includes:
301, Host calls mmap mapped files, and it is literary to disk to establish in Host the virtual address for storing data by page table The mapping relations of the physical address of data are stored in part.
Wherein, the physical address for storing data in Host in the virtual address to disk file of storage data is recorded in page table Mapping relations.
302, program and page table are sent to isomerization hardware by Host.
Wherein, Host will need the program run on isomerization hardware and page table to be sent to isomerization hardware, and isomerization hardware receives Its being executed for task is determined after program, after receiving page table, can obtain data by page table.
303, isomerization hardware creates task after receiving program.
304, isomerization hardware receives the local page table that the virtual address in page table is added to after page table isomerization hardware.
Wherein, after isomerization hardware receives the page table of Host, local page table is changed according to the page table of Host, in local page table The middle mapping relations for establishing the virtual address of storage data in the virtual address that data are locally stored and Host.
305, isomerization hardware is the virtual address addition mark of storage data in the Host in page table.
Wherein, mark indicates that this virtual address is the virtual address of certain process on Host, object corresponding with virtual address It manages on the disk that address is Host.
306, when isomerization hardware, which is run, to be needed to call data in program process, it is hard to judge whether the data are stored in isomery The memory of part obtains data if the data are stored in the memory of isomerization hardware from memory, executes step 313;When the data It is not stored in the memory of isomerization hardware, thens follow the steps 307.
Wherein, isomerization hardware can learn that the data whether there is in different by judging that the address in local page table identifies In the physical memory of structure hardware.
307, isomerization hardware generates interruption, is asked to Host transmission datas.
Wherein, isomerization hardware can obtain the data and stored in Host according to the mapping relations recorded from local page table The virtual address of data, when being asked to Host transmission datas, the data for carrying acquisition store the virtual of data in Host Address.
308, Host is according to the virtual address and mmap mapped files of storage data, by the data of isomerization hardware request from magnetic Disk is called in and is stored in memory.
309, Host adds the state of the data in the state table for safeguarding lock status.
Wherein, after data storage to memory, by the addition of its lock status in the state table for safeguarding lock status.If there is no dimensions The state table for protecting lock status then creates state table to safeguard the lock status of the data.
310, the physical address that Host is stored to isomerization hardware returned data in Host memories.
311, the local page table of physical address modification that isomerization hardware is stored according to data in Host memories.
Wherein, after isomerization hardware receives the physical address that data store in Host memories, by data in local page table The corresponding address of virtual address be revised as the physical address stored in Host memories.
312, the physical address called data that isomerization hardware is stored according to data in Host memories.
Wherein, the process for the physical address called data that isomerization hardware is stored according to data in Host memories and Fig. 2 institutes Show that the step 207 of embodiment is identical to step 212 process, details are not described herein.
It should be noted that the data that isomerization hardware is transferred in this step can be isomerization hardware number of request in step 307 According to whole, can also be a part for isomerization hardware request data in step 307.
313, isomerization hardware continues to execute program using the data of calling.
It should be noted that the scene of isomerization hardware processing big data, big data can also partly be stored in the interior of Host In depositing, be partly stored in the disk of Host, in such cases, isomerization hardware by flow chart of data processing shown in Fig. 2 as shown in figure 3 Flow chart of data processing combine.The data in disk can also be called in memory before Host sends program to isomerization hardware Storage, then directly executes flow chart of data processing shown in Fig. 2, and data are called from the disk of Host to save isomerization hardware Time.
In the prior art, isomerization hardware handles this big data in application, data to be loaded into isomerization hardware memory in batches, Data and the data needed for program operation of loading can be made inconsistent, in this way, not only taking up the memory of isomerization hardware, can also be influenced To the normal operation of program.Compared with prior art, when isomerization hardware handles big data in application, need not be by number in the present invention According to the memory for being loaded into isomerization hardware, it is only necessary to be loaded into the address information that data store in Host, while establish data in local The mapping of virtual address and address information can pass through when isomerization hardware needs to call data during running program LVA local virtual address and mapping determine the address information that data store in Host, are then sent to Host and carry address information Request, after Host returned datas, program is continued to run with using data, and isomerization hardware, wave can be both loaded into avoid redundant data Take the memory of isomerization hardware, and does not influence the normal operation of program.
Further embodiment of this invention provides the device 40 that big data is handled in a kind of isomerization hardware, as shown in figure 4, the dress Setting 40 includes:
Receiving unit 41, the program and address information sent for receiving the Host, described program are in isomerization hardware The program of upper operation, described address information are the address information that source data stores in the Host, and the source data is described The data needed for described program are run on isomerization hardware;
Unit 42 is established, the mapping for establishing virtual address and described address information that the source data is locally stored is closed System;
Request unit 43, for during running described program when the isomerization hardware when invocation target data, to described Host sends target data request, and the target data request carries the address information of the target data, in order to described Host searches the target data according to the address information of the target data, and the address information of the target data is described different Structure hardware determines that the target data belongs to the source number according to the virtual address of the mapping relations and the target data According to;
Using unit 44, for after receiving the target data that the Host is replied, institute to be run using the target data State program.
Wherein, when the source data is stored in the memory of the Host, described address information stores institute for the Host State the physical address of source data;When the source data is stored in the disk of the Host, described address information is the Host Store the virtual address of the source data.
Further, as shown in figure 5, described device 40 can also include:
Storage unit 45, for the target data to be stored in the memory of the isomerization hardware;
Unit 46 is changed, the address for virtual address and the target data that the target data will be locally stored is believed Mapping relations between breath are revised as that the virtual address of the target data is locally stored with the target data in the isomery Mapping relations in hardware memory between physical address.
Further, as shown in figure 5, the receiving unit 41 is additionally operable to receive the target data that the Host is replied The physical address stored in the memory of the Host, the physical address are that the Host transfers the number of targets from disk The physical address being stored in after in the memory of the Host;
The modification unit 46 is additionally operable to that the virtual address of the target data and the target data will be locally stored Mapping relations between address information are revised as that the virtual address of the target data and the memory in the Host is locally stored Mapping relations between the physical address of middle storage;
The request unit 43 is additionally operable to ask to the Host transmitting portions target data, and the partial target data are asked It asks and carries the physical address that partial target data store in the memory of the Host, in order to which the Host is according to the part The physical address that target data stores in the memory of the Host searches the partial target data;
It is described to be additionally operable to after receiving the partial target data that the Host is replied using unit 44, use the part mesh Mark data run described program.
Further, as shown in figure 5, described device 40 can also include:
Adding device 47, for being identified for the addition of described address information, the mark indicates that described address information is described The address information that data store in the Host.
In the prior art, isomerization hardware handles this big data in application, data to be loaded into isomerization hardware memory in batches, Data and the data needed for program operation of loading can be made inconsistent, in this way, not only taking up the memory of isomerization hardware, can also be influenced To the normal operation of program.Compared with prior art, the program to be run that 40 receiving host Host of device is sent in the present invention The address information that required source data stores in Host is run with program, address information is the address information of data source storage;It builds Deposit the mapping relations of the virtual address and address information of source storage data in vertical local;Source number is called in program process when device 40 is run When target data in, device 40 determines that the address of target data is believed according to the virtual address of mapping relations and target data Breath, and the request of data for carrying target data address information is sent to Host, in order to which Host believes according to the address of target data Breath searches target data;After receiving the target data that Host is replied, program is run using target data.In this way, working as device 40 Big data is handled in application, the memory of device 40 need not be loaded data into, it is only necessary to be loaded into the ground that data store in Host Location information, while mapping of the data in LVA local virtual address and address information is established, when device 40 is during running program When needing to call data, the address information that data store in Host can be determined by LVA local virtual address and mapping, then It is sent to Host and carries the request of address information and continue to run with program using data after Host returned datas, it both can be to avoid Redundant data device for loading 40, wastes the memory of device 40, and does not influence the normal operation of program.
The method that above-mentioned offer may be implemented in the device that big data is handled in isomerization hardware provided in an embodiment of the present invention is real Example is applied, concrete function realizes the explanation referred in embodiment of the method, and details are not described herein.Isomery provided in an embodiment of the present invention The method and device that big data is handled in hardware can be adapted for isomerization hardware, but be not limited only to this.
Each embodiment in this specification is described in a progressive manner, identical similar portion between each embodiment Point just to refer each other, and each embodiment focuses on the differences from other embodiments.Especially for equipment reality For applying example, since it is substantially similar to the method embodiment, so describing fairly simple, related place is referring to embodiment of the method Part explanation.
One of ordinary skill in the art will appreciate that realizing all or part of flow in above-described embodiment method, being can be with Relevant hardware is instructed to complete by computer program, the program can be stored in a computer read/write memory medium In, the program is when being executed, it may include such as the flow of the embodiment of above-mentioned each method.Wherein, the storage medium can be magnetic Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access Memory, RAM) etc..
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any Those familiar with the art in the technical scope disclosed by the present invention, all answer by the change or replacement that can be readily occurred in It is included within the scope of the present invention.Therefore, protection scope of the present invention should be subject to the protection scope in claims.

Claims (10)

1. a kind of method that big data is handled in isomerization hardware, isomerization hardware are connect with host Host, which is characterized in that including:
Program and address information that the Host is sent are received, described program is the program run on isomerization hardware, describedly Location information is the address information that source data stores in the Host, and the source data is to run the journey on the isomerization hardware Data needed for sequence;
Establish the mapping relations of virtual address and described address information that the source data is locally stored;
During the isomerization hardware runs described program when invocation target data, sends target data to the Host and ask It asks, the target data request carries the address information of the target data, in order to which the Host is according to the target data Address information search the target data, the address information of the target data is that the isomerization hardware is closed according to the mapping The virtual address of system and the target data determines that the target data belongs to the source data;
After receiving the target data that the Host is replied, described program is run using the target data.
2. according to the method described in claim 1, it is characterized in that, when the source data is stored in the memory of the Host, The address information of the source data is the physical address that the Host stores the source data;
When the source data is stored in the disk of the Host, the address information of the source data is described in Host storages The virtual address of source data.
3. according to the method described in claim 1, it is characterized in that, it is described using the target data operation described program it Before, the method further includes:
The target data is stored in the memory of the isomerization hardware;
Mapping relations between the virtual address that the target data is locally stored and the address information of the target data are repaiied It is changed to that the virtual address of the target data and target data physical address in the isomerization hardware memory is locally stored Between mapping relations.
4. according to the method described in claim 3, it is characterized in that, when the source data is stored in the disk of the Host, It is described the target data is stored in the memory of the isomerization hardware before, the method further includes:
Receive the physical address that the target data that the Host is replied stores in the memory of the Host, it is described physically Location is that the Host is stored in the physical address in the memory of the Host after transferring the target data in disk;
Mapping relations between the virtual address that the target data is locally stored and the address information of the target data are repaiied It is changed to that reflecting between the virtual address of the target data and the physical address stored in the memory of the Host is locally stored Penetrate relationship;
It is asked to the Host transmitting portions target data, the partial target request of data carries partial target data described The physical address stored in the memory of Host, in order to the Host according to the partial target data the Host memory The physical address of middle storage searches the partial target data;
It is then described after receiving the target data that the Host is replied, include using target data operation described program:
After receiving the partial target data that the Host is replied, the partial target data run described program is used.
5. according to the method described in claim 1, it is characterized in that, described when the isomerization hardware runs described program process When middle invocation target data, before being asked to the Host transmission datas, the method further includes:
It adds and identifies for described address information, the mark indicates that described address information is that the data store in the Host Address information.
6. the device that big data is handled in a kind of isomerization hardware, which is characterized in that including:
Receiving unit, program and address information for receiving Host transmissions, described program is the journey run on isomerization hardware Sequence, described address information are the address information that source data stores in the Host, and the source data is on the isomerization hardware Run the data needed for described program;
Unit is established, the mapping relations for establishing virtual address and described address information that the source data is locally stored;
Request unit, for when invocation target data, being sent out to the Host during running described program when the isomerization hardware Send target data to ask, target data request carries the address information of the target data, in order to the Host according to The address information of the target data searches the target data, and the address information of the target data is the isomerization hardware root It is determined according to the virtual address of the mapping relations and the target data, the target data belongs to the source data;
Using unit, for after receiving the target data that the Host is replied, described program to be run using the target data.
7. device according to claim 6, which is characterized in that when the source data is stored in the memory of the Host, The address information of the source data is the physical address that the Host stores the source data;
When the source data is stored in the disk of the Host, the address information of the source data is described in Host storages The virtual address of source data.
8. device according to claim 6, which is characterized in that described device further includes:
Storage unit, for the target data to be stored in the memory of the isomerization hardware;
Unit is changed, between the virtual address and the address information of the target data for the target data will to be locally stored Mapping relations be revised as being locally stored the virtual address of the target data with the target data in the isomerization hardware Deposit the mapping relations between physical address.
9. device according to claim 8, which is characterized in that the receiving unit is additionally operable to receive what the Host was replied The physical address that the target data stores in the memory of the Host, the physical address are that the Host is adjusted from disk Take the physical address being stored in after the target data in the memory of the Host;
Believe the address that the modification unit is additionally operable to the virtual address and the target data that the target data will be locally stored The virtual address that mapping relations between breath are revised as being locally stored the target data is stored in the memory of the Host Physical address between mapping relations;
The request unit is additionally operable to ask to the Host transmitting portions target data, and the partial target request of data carries The physical address that partial target data store in the memory of the Host, in order to which the Host is according to the partial target number The partial target data are searched according to the physical address stored in the memory of the Host;
It is described to be additionally operable to after receiving the partial target data that the Host is replied using unit, use the partial target data Run described program.
10. device according to claim 6, which is characterized in that described device further includes:
Adding device, for being identified for the addition of described address information, the mark indicates that described address information is that the data exist The address information stored in the Host.
CN201410608826.6A 2014-10-31 2014-10-31 The method and device that big data is handled in isomerization hardware Active CN105630690B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410608826.6A CN105630690B (en) 2014-10-31 2014-10-31 The method and device that big data is handled in isomerization hardware

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410608826.6A CN105630690B (en) 2014-10-31 2014-10-31 The method and device that big data is handled in isomerization hardware

Publications (2)

Publication Number Publication Date
CN105630690A CN105630690A (en) 2016-06-01
CN105630690B true CN105630690B (en) 2018-10-12

Family

ID=56045659

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410608826.6A Active CN105630690B (en) 2014-10-31 2014-10-31 The method and device that big data is handled in isomerization hardware

Country Status (1)

Country Link
CN (1) CN105630690B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110737608B (en) * 2018-07-03 2024-03-15 阿里巴巴集团控股有限公司 Data operation method, device and system
CN112306718B (en) * 2020-11-18 2023-02-28 山东云海国创云计算装备产业创新中心有限公司 Communication method, system and related device between local equipment and heterogeneous equipment
CN115454681B (en) * 2022-11-10 2023-01-20 维塔科技(北京)有限公司 Batch processing program execution method, device and system

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5907864A (en) * 1995-06-07 1999-05-25 Texas Instruments Incorporated Data processing device with time-multiplexed memory bus
CN101013404A (en) * 2007-01-26 2007-08-08 浙江大学 Heterogeneous multi-core system-oriented management method of paging memory space
CN101430651A (en) * 2007-11-05 2009-05-13 中兴通讯股份有限公司 Access method for peripheral devices in heterogeneous multi-core system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5907864A (en) * 1995-06-07 1999-05-25 Texas Instruments Incorporated Data processing device with time-multiplexed memory bus
CN101013404A (en) * 2007-01-26 2007-08-08 浙江大学 Heterogeneous multi-core system-oriented management method of paging memory space
CN101430651A (en) * 2007-11-05 2009-05-13 中兴通讯股份有限公司 Access method for peripheral devices in heterogeneous multi-core system

Also Published As

Publication number Publication date
CN105630690A (en) 2016-06-01

Similar Documents

Publication Publication Date Title
CN104679669B (en) The method of cache cache accumulator systems and access cache row cache line
CN103544269B (en) Methods and node controllers for storing and enquiring directories
US7870536B2 (en) Computer implemented method and system for sharing resources among hierarchical containers of resources
US20200364106A1 (en) Storage system with coordinated recovery across multiple input-output journals of different types
CN103559319B (en) The cache synchronization method and equipment of distributed cluster file system
CN105630690B (en) The method and device that big data is handled in isomerization hardware
CN109491928A (en) Buffer control method, device, terminal and storage medium
CN105408895A (en) Latch-free, log-structured storage for multiple access methods
JP2001526815A (en) File system primitives that provide native file system support for remote storage
CN102693164B (en) Equipment and method for preventing buffer overflow
CN103885894A (en) Information Coherency Maintenance Systems And Methods
CN102141925A (en) Applicability detection using third party target state
CN110019112A (en) Data transactions method, apparatus and electronic equipment
CN109656886B (en) Key value pair-based file system implementation method, device, equipment and storage medium
CN108595503A (en) Document handling method and server
CN103677878A (en) Method and device for patching
CN105049530A (en) Adaption device and method for plurality of distributed cache systems
CN106250061A (en) File download processing method and processing device
CN105446889B (en) A kind of EMS memory management process, device and Memory Controller Hub
CN109656889A (en) File system implementation method, device, equipment and storage medium based on block chain
CN105701020B (en) A kind of method of internal storage access, relevant apparatus and system
CN106649146A (en) Memory release method and apparatus
WO2015130282A1 (en) Communication between integrated graphics processing units
JP2004362570A (en) Shadow paging
US7653797B1 (en) Optimizing a marking phase in mark-sweep garbage collectors by reducing paging activity

Legal Events

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