CN108632170A - A kind of method and device for realizing bandwidth allocation - Google Patents
A kind of method and device for realizing bandwidth allocation Download PDFInfo
- Publication number
- CN108632170A CN108632170A CN201710152461.4A CN201710152461A CN108632170A CN 108632170 A CN108632170 A CN 108632170A CN 201710152461 A CN201710152461 A CN 201710152461A CN 108632170 A CN108632170 A CN 108632170A
- Authority
- CN
- China
- Prior art keywords
- bandwidth
- current limliting
- memory access
- application
- collection 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.)
- Withdrawn
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/50—Queue scheduling
- H04L47/52—Queue scheduling by attributing bandwidth to queues
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/50—Queue scheduling
- H04L47/52—Queue scheduling by attributing bandwidth to queues
- H04L47/525—Queue scheduling by attributing bandwidth to queues by redistribution of residual bandwidth
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/70—Admission control; Resource allocation
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
A kind of realization bandwidth allocation methods and device, including:Bandwidth allocation is carried out to each application;Bandwidth current limliting processing is carried out to the application for completing bandwidth allocation according to the memory access density collection information of application.The embodiment of the present invention is handled by bandwidth current limliting realizes Bandwidth adjustment, improves the performance of system.
Description
Technical field
Present document relates to but be not limited to multiprocessor systems on chips technology, espespecially a kind of method and dress for realizing bandwidth allocation
It sets.
Background technology
In multi-processor system-on-chip field, the System on Chip/SoC of multiple nucleus system generally uses synchronous DRAM
(SDRAM, Synchronous Dynamic Random Access Memory) or its upgrade the memory of type as memory,
Multinuclear memory access interference problem drastically influences the memory access performance of system and monokaryon or application.Wherein, the execution of accessing operation is suitable
Sequence determines the number of memory switching row address and read-write type, to affect memory access performance;From different IPs or application
Access request, the limited main memory resource of competition being interweaved, the access request sequence pair for destroying different IPs/application answer main memory
In original locality, so as to cause access request sequence row locality and body (bank) concurrency reduce;In addition, multinuclear
Competition main memory resource also results in access request congestion.The memory access performance for the system that reduces is interfered in multinuclear memory access, how to be solved more
Core memory access interference problem is always multiprocessor systems on chips (MPSoC, multiprocessor system-on-chip) research
Hot spot.Memory access transaction is needed from initiating to being finally completed by main equipment, main memory controller and dynamic random access memory
(DRAM) etc. several components solve memory access interference problem and carry out mainly for above-mentioned three.In the related technology, processing memory access interference
The method of problem includes:Main memory divides, main memory is dispatched and bandwidth current limliting management.
Main memory divides mainly different channels or the body (bank) by the way that the access request of different main equipments to be mapped to main memory
In, to reduce memory access interference.The access request of different main equipments is mapped to different channels or the body of main memory, can be isolated not
With the access request of main equipment, it is effectively retained memory access sequence locality characteristics.Main memory division includes that channel divides and bank divisions.
Wherein, it is a kind of channel division side of application perception that main memory access, which divides (MCP, Memory Channel Partitioning),
Method, this method are then highest by annoyance level by assessing the annoyance levels of different main equipment access request sequence pair systems
Main equipment is mapped to individual channel.It is limited to the reduction of memory access interference but since most threads remain unchanged sharing channel.
Bank is divided is isolated interference by interfering memory access the access request sequence of high core to be mapped to different bank, reduces multinuclear
It influences each other.Bank is divided to be divided flexibly than channel, and the scope of application is also relatively wide, but bank divisions equally reduce access request
The bank level parallelisms of sequence, and can not effectively solve core/application and the annoyance level low type of the high type of annoyance level
Interfering with each other between core/application.
Main memory scheduling by access request sequence reorder reduce interference to the original row locality of thread and
The influence of bank level parallelisms.The method of the usual connected applications perception of current memory access optimisation technique improves main memory controller pair
The perception degree of access request sequence.For example parallel perception batch dispatching algorithm (PAR-BS) is improved by the way of batch splitting scheduling
Performance;The linear algebra algorithm (ATLAS) of automatic adjustment is carried by preferentially obtaining the access request that minimum DRAM is served by
The throughput of high system, but need to sacrifice fairness to be cost.
The management of bandwidth current limliting is mainly competed by coordinating the memory access of each main equipment, slows down bus Congestion Level SPCC, is effectively reduced
Multinuclear access request sequence staggeredly and congestion, to reduce memory access interference effect.Existing bandwidth current limliting management is general to be used
Current limliting is carried out at main equipment port (the processor end for sending access request), by sending access request inside control processor
Rate is realized (generally by controlling its internal MSHR register).The characteristic of access request is generally sent by monitoring main equipment
To be adjusted to MSHR registers.
The management of bandwidth current limliting divides flexibly than main memory scheduling and main memory, while hardware complexity is small.Recent study is main
There are main equipment end bandwidth current limliting and main memory controller end bandwidth current limliting.The method that FST proposes source current limliting, by main equipment
Control memory access rate in end improves system fairness to reduce memory access interference.FST establishes a set of effective feedback loop mechanism
Multinuclear memory access fairness situation of change is monitored, by the thread of preferential fairness difference come lifting system fairness.But FST is needed
Processor structure is changed, hardware realization complexity is high, and this method does not account for system throughput demand.
Chip multiprocessors (CMP) system is a kind of system in multiprocessor systems on chips, CMP system bandwidth allocation
Mainly use one kind in following two modes:One is be allocated bandwidth in strict accordance with the bandwidth demand of application;It is another
By bandwidth strict preference memory access intensive applications when kind;Wherein, bandwidth is allocated in strict accordance with the bandwidth demand of application, meeting
Leading to memory access, intensive applications do not occur dying of hunger phenomenon;And if strict preference memory access intensive applications, it can be under some scenes
(for example most applications is memory access not intensive applications) influences the memory access performance of memory access intensive applications.Therefore, above-mentioned CMP systems
There is certain defect in the bandwidth allocation methods in system, affect the performance of CMP system.
Invention content
It is the general introduction of the theme to being described in detail herein below.This general introduction is not to limit the protection model of claim
It encloses.
The embodiment of the present invention provides a kind of method and device for realizing bandwidth allocation, can carry out reasonable distribution to bandwidth,
The performance of lifting system.
An embodiment of the present invention provides a kind of methods for realizing bandwidth allocation, including:
Bandwidth allocation is carried out to each application;
Bandwidth current limliting processing is carried out to the application for completing bandwidth allocation according to the memory access density collection information of application.
Optionally, further include before this method:
To each application, determination apply the access request amount in predetermined period, using determining access request amount as
The memory access density collection information of application.
Optionally, the progress bandwidth allocation includes:
The bandwidth demand that each application is determined according to the access request amount in the memory access density collection information of each application, according to
Determining bandwidth demand carries out bandwidth allocation to all applications;
Wherein, the bandwidth demand of application is directly proportional to the access request amount.
Optionally, the progress bandwidth current limliting, which is handled, includes:
By all applications, it is divided into according to the memory access density collection information and one or more preset sub-clustering threshold values
Two or more bandwidth sub-clusterings;
Each bandwidth sub-clustering to division carries out bandwidth current limliting processing according to preset current limliting strategy.
Optionally, the progress bandwidth current limliting, which is handled, includes:
By all applications, bandwidth sensitive cluster is divided into according to the memory access density collection information and preset first sub-clustering threshold value
With delay-sensitive cluster;
Bandwidth current limliting processing is carried out according to preset current limliting strategy to bandwidth sensitive cluster.
Optionally, described to include according to the progress bandwidth current limliting processing of memory access density collection information:
According to the memory access frequency of the memory access frequency and delay-sensitive cluster of memory access density collection information computation bandwidth sensitivity cluster, and count
Calculate the memory access frequency ratio of the memory access frequency of bandwidth sensitive cluster and the memory access frequency of delay-sensitive cluster;
Calculated according to memory access density collection information all applications access request amount it is cumulative and;
When access request amount is cumulative and corresponding bandwidth demand is more than system bandwidth, according to the memory access frequency for calculating acquisition
Ratio and access request amount add up and carry out real-time bandwidth current limliting to bandwidth sensitive cluster and handle.
Optionally, the method further includes:
When carrying out access request without other application in addition to the application for carrying out bandwidth current limliting processing, release to carrying out bandwidth
The bandwidth current limliting of one or more applications of current limliting processing.
On the other hand, the embodiment of the present invention also provides a kind of device for realizing bandwidth allocation, including:Allocation unit and current limliting
Processing unit;Wherein,
Allocation unit is used for, and bandwidth allocation is carried out to each application;
Current limliting processing unit is used for, and bandwidth is carried out to the application for completing bandwidth allocation according to the memory access density collection information of application
Current limliting processing.
Optionally, which further includes:
Determination unit, for each application, the access request amount in predetermined period to be applied in determination, with determining visit
Deposit the memory access density collection information of the request amount as application.
Optionally, the allocation unit is specifically used for:
The bandwidth demand that each application is determined according to the access request amount in the memory access density collection information of each application, according to
Determining bandwidth demand carries out bandwidth allocation to all applications;
Wherein, the bandwidth demand of application is directly proportional to the access request amount.
Optionally, the current limliting processing unit is specifically used for:
By all applications, it is divided into according to the memory access density collection information and one or more preset sub-clustering threshold values
Two or more bandwidth sub-clusterings;
Each bandwidth sub-clustering to division carries out bandwidth current limliting processing according to preset current limliting strategy.
Optionally, the current limliting processing unit is specifically used for:
By all applications, bandwidth sensitive cluster is divided into according to the memory access density collection information and preset first sub-clustering threshold value
With delay-sensitive cluster;
Bandwidth current limliting processing is carried out according to preset current limliting strategy to bandwidth sensitive cluster.
Optionally, the current limliting processing unit includes according to the progress bandwidth current limliting processing of memory access density collection information:
According to the memory access frequency of the memory access frequency and delay-sensitive cluster of memory access density collection information computation bandwidth sensitivity cluster, and count
Calculate the memory access frequency ratio of the memory access frequency of bandwidth sensitive cluster and the memory access frequency of delay-sensitive cluster;
Calculated according to memory access density collection information all applications access request amount it is cumulative and;
When access request amount is cumulative and corresponding bandwidth demand is more than system bandwidth, according to the memory access frequency for calculating acquisition
Ratio and access request amount add up and carry out real-time bandwidth current limliting to bandwidth sensitive cluster and handle.
Optionally, described device further includes lifting unit, for working as in addition to the application for carrying out bandwidth current limliting processing without it
When his application carries out access request, the bandwidth current limliting of one or more applications to carrying out bandwidth current limliting processing is released.
Compared with the relevant technologies, technical scheme includes:Bandwidth allocation is carried out to each application;According to the memory access of application
Density collection information carries out bandwidth current limliting processing to the application for completing bandwidth allocation.The embodiment of the present invention is handled real by bandwidth current limliting
Show Bandwidth adjustment, improves the performance of system.
Other features and advantages of the present invention will be illustrated in the following description, also, partly becomes from specification
It obtains it is clear that understand through the implementation of the invention.The purpose of the present invention and other advantages can be by specification, rights
Specifically noted structure is realized and is obtained in claim and attached drawing.
Description of the drawings
Attached drawing is used for providing further understanding technical solution of the present invention, and a part for constitution instruction, with this
The embodiment of application technical solution for explaining the present invention together, does not constitute the limitation to technical solution of the present invention.
Fig. 1 is the flow chart for the method that the embodiment of the present invention realizes bandwidth allocation;
Fig. 2 is the structure diagram for the device that the embodiment of the present invention realizes bandwidth allocation.
Specific implementation mode
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with attached drawing to the present invention
Embodiment be described in detail.It should be noted that in the absence of conflict, in the embodiment and embodiment in the application
Feature mutually can arbitrarily combine.
Step shown in the flowchart of the accompanying drawings can be in the computer system of such as a group of computer-executable instructions
It executes.Also, although logical order is shown in flow charts, and it in some cases, can be with suitable different from herein
Sequence executes shown or described step.
Fig. 1 is the flow chart for the method that the embodiment of the present invention realizes bandwidth allocation, as shown in Figure 1, including:
Step 100 carries out bandwidth allocation to each application;
Optionally, further include before step 100 of the embodiment of the present invention:
To each application, determination apply the access request amount in predetermined period, using determining access request amount as
The memory access density collection information of application.
It should be noted that predetermined period can be period regular time, predetermined period can be defined as the time
Piece, since the data of the too short acquisition of predetermined period are unstable, can not accurate response application memory access situation, predetermined period is long to be obtained
Data can not react the changing rule of memory access situation, therefore experience that predetermined period can be based on those skilled in the art,
And the effect in application process in conjunction with bandwidth allocation process is adjusted.
Optionally, progress of embodiment of the present invention bandwidth allocation includes:
The bandwidth demand that each application is determined according to the access request amount in the memory access density collection information of each application, according to
Determining bandwidth demand carries out bandwidth allocation to all applications;
Wherein, the bandwidth demand of application is directly proportional to access request amount.
It should be noted that since memory access density collection information is obtained according to predetermined period in real time, bandwidth
Distribution can carry out period update using predetermined period as unit;It can be in the access request amount pair of previous predetermined period
The application in current preset period carries out bandwidth allocation.
Step 101 carries out bandwidth current limliting processing according to the memory access density collection information of application to the application for completing bandwidth allocation.
Optionally, progress of embodiment of the present invention bandwidth current limliting, which is handled, includes:
By all applications, two are divided into according to memory access density collection information and one or more preset sub-clustering threshold values
Or more than two bandwidth sub-clusterings;
Each bandwidth sub-clustering to division carries out bandwidth current limliting processing according to preset current limliting strategy.
It should be noted that current limliting strategy may include:One or more bandwidth sub-clusterings more to bandwidth allocation
It carries out bandwidth current limliting and provides bandwidth by the bandwidth current limliting of the more application of bandwidth allocation for the less application of bandwidth allocation.
Optionally, progress of embodiment of the present invention bandwidth current limliting, which is handled, includes:
By all applications, bandwidth sensitive cluster is divided into according to memory access density collection information and preset first sub-clustering threshold value and is prolonged
Sensitivity cluster late;
Bandwidth current limliting processing is carried out according to preset current limliting strategy to bandwidth sensitive cluster.
It should be noted that the first sub-clustering threshold value chooses the general access request amount for considering system bandwidth and application.The
One sub-clustering threshold value is excessively high, can be that excessive application is divided into delay-sensitive cluster, and memory access interference can be caused to increase;If first point
Cluster threshold value is too low, and more applications are divided into bandwidth sensitive cluster, and system throughput can be affected.The embodiment of the present invention can be with
According to system bandwidth and access request amount, the setting that value carries out the first sub-clustering threshold value is incorporated experience into, the first sub-clustering threshold value can root
It is handled according to the effect of bandwidth current limliting.
In addition, the embodiment of the present invention includes bandwidth sensitive cluster and delay-sensitive cluster, one is carried out to the application of bandwidth sensitive cluster
Fixed current limliting can provide bandwidth for the application in delay-sensitive cluster, improve systematic entirety energy;Current limliting width of the embodiment of the present invention
There are certain relationships for difference degree between the selection of degree and system bandwidth and the access request amount of application.To bandwidth sensitive cluster
In application carry out current limliting processing after, delay-sensitive cluster can obtain more memory access chances by relatively preferential, to more
Its calculating process is promoted soon, increases system throughput;If current limliting amplitude is too small, delay-sensitive cluster cannot obtain enough bands
It is wide;If current limliting amplitude is excessive, the memory access interference inside bandwidth sensitive cluster will increase, to influence system performance.The present invention is real
Example is applied in the case of system bandwidth scarcity of resources, the effect for carrying out bandwidth current limliting acquisition is more notable.The embodiment of the present invention may be used also
To carry out bandwidth current limliting processing according to other strategies, such as the application in bandwidth sensitive cluster is gradually carried out according to default step-length
The processing of bandwidth current limliting, current limliting amplitude is determined according to the result of bandwidth current limliting;The embodiment of the present invention can also be to bandwidth current limliting amplitude
It is recorded, the current bandwidth current limliting amplitude that adjustment is obtained is as the bandwidth current limliting width of posterior predetermined period one by one
The initial value of degree.The processing of bandwidth current limliting can be executed by bandwidth demand monitor, and bandwidth demand monitor is in the related technology
Existing structure, this will not be repeated here.
Optionally, the embodiment of the present invention includes according to the progress bandwidth current limliting processing of memory access density collection information:
According to the memory access frequency of the memory access frequency and delay-sensitive cluster of memory access density collection information computation bandwidth sensitivity cluster, and count
Calculate the memory access frequency ratio of the memory access frequency of bandwidth sensitive cluster and the memory access frequency of delay-sensitive cluster;
Calculated according to memory access density collection information all applications access request amount it is cumulative and;
When access request amount is cumulative and corresponding bandwidth demand is more than system bandwidth, according to the memory access frequency for calculating acquisition
Ratio and access request amount add up and carry out real-time bandwidth current limliting to bandwidth sensitive cluster and handle.
It should be noted that the processing of bandwidth current limliting may include:Memory access response is limited by not giving bus arbitration mandate,
And then the bandwidth current limliting realized.
In addition, when the bandwidth of application application is beyond the bandwidth distributed, the embodiment of the present invention can be in the current preset period
Interior to carry out bandwidth current limliting processing to application, control application obtains the bandwidth resources in addition to bandwidth allocation.
Optionally, present invention method further includes:
When carrying out access request without other application in addition to the application for carrying out bandwidth current limliting processing, release to carrying out bandwidth
The bandwidth current limliting of one or more applications of current limliting processing.
It should be noted that all applications can be carried out to release bandwidth current limliting by releasing bandwidth current limliting, it can also be according to place
Bandwidth current limliting is determined whether to lift in the access request amount of the application of bandwidth current limliting processing, such as only less to access request amount
Certain applications carry out release bandwidth current limliting processing.
Compared with the relevant technologies, technical scheme includes:Bandwidth allocation is carried out to each application;According to the memory access of application
Density collection information carries out bandwidth current limliting processing to the application for completing bandwidth allocation.The embodiment of the present invention is handled real by bandwidth current limliting
Show Bandwidth adjustment, improves the performance of system.
Fig. 2 is the structure diagram for the device that the embodiment of the present invention realizes bandwidth allocation, as shown in Fig. 2, including:Allocation unit
With current limliting processing unit;Wherein,
Allocation unit is used for, and bandwidth allocation is carried out to each application;
Optionally, allocation unit of the embodiment of the present invention is specifically used for:
The bandwidth demand that each application is determined according to the access request amount in the memory access density collection information of each application, according to
Determining bandwidth demand carries out bandwidth allocation to all applications;
Wherein, the bandwidth demand of application is directly proportional to access request amount.
It should be noted that since memory access density collection information is obtained according to predetermined period in real time, bandwidth
Distribution can carry out period update using predetermined period as unit;It can be in the access request amount pair of previous predetermined period
The application in current preset period carries out bandwidth allocation.
Current limliting processing unit is used for, and bandwidth is carried out to the application for completing bandwidth allocation according to the memory access density collection information of application
Current limliting processing.
Optionally, current limliting of embodiment of the present invention processing unit is specifically used for:
By all applications, two are divided into according to memory access density collection information and one or more preset sub-clustering threshold values
Or more than two bandwidth sub-clusterings;
Each bandwidth sub-clustering to division carries out bandwidth current limliting processing according to preset current limliting strategy.
It should be noted that current limliting strategy may include:One or more bandwidth sub-clusterings more to bandwidth allocation
It carries out bandwidth current limliting and provides bandwidth by the bandwidth current limliting of the more application of bandwidth allocation for the less application of bandwidth allocation.
Optionally, current limliting of embodiment of the present invention processing unit is specifically used for:
By all applications, bandwidth sensitive cluster is divided into according to memory access density collection information and preset first sub-clustering threshold value and is prolonged
Sensitivity cluster late;
Bandwidth current limliting processing is carried out according to preset current limliting strategy to bandwidth sensitive cluster.
It should be noted that the first sub-clustering threshold value chooses the general access request amount for considering system bandwidth and application.The
One sub-clustering threshold value is excessively high, can be that excessive application is divided into delay-sensitive cluster, and memory access interference can be caused to increase;If first point
Cluster threshold value is too low, and more applications are divided into bandwidth sensitive cluster, and system throughput can be affected.The embodiment of the present invention can be with
According to system bandwidth and access request amount, the setting that value carries out the first sub-clustering threshold value is incorporated experience into, the first sub-clustering threshold value can root
It is handled according to the effect of bandwidth current limliting.
In addition, the embodiment of the present invention includes bandwidth sensitive cluster and delay-sensitive cluster, one is carried out to the application of bandwidth sensitive cluster
Fixed current limliting can provide bandwidth for the application in delay-sensitive cluster, improve systematic entirety energy;Current limliting width of the embodiment of the present invention
There are certain relationships for difference degree between the selection of degree and system bandwidth and the access request amount of application.To bandwidth sensitive cluster
In application carry out current limliting processing after, delay-sensitive cluster can obtain more memory access chances by relatively preferential, to more
Its calculating process is promoted soon, increases system throughput;If current limliting amplitude is too small, delay-sensitive cluster cannot obtain enough bands
It is wide;If current limliting amplitude is excessive, the memory access interference inside bandwidth sensitive cluster will increase, to influence system performance.The present invention is real
Example is applied in the case of system bandwidth scarcity of resources, the effect for carrying out bandwidth current limliting acquisition is more notable.The embodiment of the present invention may be used also
To carry out bandwidth current limliting processing according to other strategies, such as the application in bandwidth sensitive cluster is gradually carried out according to default step-length
The processing of bandwidth current limliting, current limliting amplitude is determined according to the result of bandwidth current limliting;The embodiment of the present invention can also be to bandwidth current limliting amplitude
It is recorded, the current bandwidth current limliting amplitude that adjustment is obtained is as the bandwidth current limliting width of posterior predetermined period one by one
The initial value of degree.The processing of bandwidth current limliting can be executed by bandwidth demand monitor, and bandwidth demand monitor is in the related technology
Existing structure, this will not be repeated here.
Optionally, current limliting of embodiment of the present invention processing unit carries out bandwidth current limliting processing packet according to memory access density collection information
It includes:
According to the memory access frequency of the memory access frequency and delay-sensitive cluster of memory access density collection information computation bandwidth sensitivity cluster, and count
Calculate the memory access frequency ratio of the memory access frequency of bandwidth sensitive cluster and the memory access frequency of delay-sensitive cluster;
Calculated according to memory access density collection information all applications access request amount it is cumulative and;
When access request amount is cumulative and corresponding bandwidth demand is more than system bandwidth, according to the memory access frequency for calculating acquisition
Ratio and access request amount add up and carry out real-time bandwidth current limliting to bandwidth sensitive cluster and handle.
It should be noted that the processing of bandwidth current limliting may include:Memory access response is limited by not giving bus arbitration mandate,
And then the bandwidth current limliting realized.
In addition, when the bandwidth of application application is beyond the bandwidth distributed, the embodiment of the present invention can be in the current preset period
Interior to carry out bandwidth current limliting processing to application, control application obtains the bandwidth resources in addition to bandwidth allocation.
Optionally, device of the embodiment of the present invention further includes:
Determination unit, for each application, the access request amount in predetermined period to be applied in determination, with determining visit
Deposit memory access density collection information of the request amount as application.
It should be noted that predetermined period can be period regular time, predetermined period can be defined as the time
Piece, since the data of the too short acquisition of predetermined period are unstable, can not accurate response application memory access situation, predetermined period is long to be obtained
Data can not react the changing rule of memory access situation, therefore experience that predetermined period can be based on those skilled in the art,
And the effect in application process in conjunction with bandwidth allocation process is adjusted.
Optionally, device of the embodiment of the present invention further includes lifting unit, for when except the application for carrying out bandwidth current limliting processing
When carrying out access request without other application outside, the bandwidth of one or more applications to carrying out bandwidth current limliting processing is released
Current limliting.
It should be noted that all applications can be carried out to release bandwidth current limliting by releasing bandwidth current limliting, it can also be according to place
Bandwidth current limliting is determined whether to lift in the access request amount of the application of bandwidth current limliting processing, such as only less to access request amount
Certain applications carry out release bandwidth current limliting processing.
Compared with the relevant technologies, technical scheme includes:Bandwidth allocation is carried out to each application;According to the memory access of application
Density collection information carries out bandwidth current limliting processing to the application for completing bandwidth allocation.The embodiment of the present invention is handled real by bandwidth current limliting
Show Bandwidth adjustment, improves the performance of system.
One of ordinary skill in the art will appreciate that all or part of step in the above method can be instructed by program
Related hardware (such as processor) is completed, and described program can be stored in computer readable storage medium, as read-only memory,
Disk or CD etc..Optionally, all or part of step of above-described embodiment can also be come using one or more integrated circuits
It realizes.Correspondingly, the form realization of hardware may be used in each module/unit in above-described embodiment, such as passes through integrated electricity
Its corresponding function is realized on road, can also be realized in the form of software function module, such as is stored in by processor execution
Program/instruction in memory realizes its corresponding function.The present invention is not limited to the hardware and softwares of any particular form
In conjunction with.
Although disclosed herein embodiment it is as above, the content only for ease of understanding the present invention and use
Embodiment is not limited to the present invention.Technical staff in any fields of the present invention is taken off not departing from the present invention
Under the premise of the spirit and scope of dew, any modification and variation, but the present invention can be carried out in the form and details of implementation
Scope of patent protection, still should be subject to the scope of the claims as defined in the appended claims.
Claims (14)
1. a kind of method for realizing bandwidth allocation, which is characterized in that including:
Bandwidth allocation is carried out to each application;
Bandwidth current limliting processing is carried out to the application for completing bandwidth allocation according to the memory access density collection information of application.
2. according to the method described in claim 1, it is characterized in that, further including before this method:
To each application, the access request amount in predetermined period is applied in determination, using determining access request amount as application
The memory access density collection information.
3. according to the method described in claim 1, it is characterized in that, the progress bandwidth allocation includes:
The bandwidth demand that each application is determined according to the access request amount in the memory access density collection information of each application, according to determination
Bandwidth demand to all applications carry out bandwidth allocation;
Wherein, the bandwidth demand of application is directly proportional to the access request amount.
4. according to claims 1 to 3 any one of them method, which is characterized in that the progress bandwidth current limliting, which is handled, includes:
By all applications, two are divided into according to the memory access density collection information and one or more preset sub-clustering threshold values
Or more than two bandwidth sub-clusterings;
Each bandwidth sub-clustering to division carries out bandwidth current limliting processing according to preset current limliting strategy.
5. according to claims 1 to 3 any one of them method, which is characterized in that the progress bandwidth current limliting, which is handled, includes:
By all applications, bandwidth sensitive cluster is divided into according to the memory access density collection information and preset first sub-clustering threshold value and is prolonged
Sensitivity cluster late;
Bandwidth current limliting processing is carried out according to preset current limliting strategy to bandwidth sensitive cluster.
6. according to the method described in claim 5, it is characterized in that, described carry out according to memory access density collection information at bandwidth current limliting
Reason includes:
According to the memory access frequency of the memory access frequency and delay-sensitive cluster of memory access density collection information computation bandwidth sensitivity cluster, and calculate band
The memory access frequency ratio of the memory access frequency and the memory access frequency of delay-sensitive cluster of wide sensitivity cluster;
Calculated according to memory access density collection information all applications access request amount it is cumulative and;
When access request amount is cumulative and corresponding bandwidth demand is more than system bandwidth, according to the memory access frequency ratio for calculating acquisition
And access request amount is cumulative and carries out real-time bandwidth current limliting processing to bandwidth sensitive cluster.
7. according to claims 1 to 3 any one of them method, which is characterized in that the method further includes:
When carrying out access request without other application in addition to the application for carrying out bandwidth current limliting processing, release to carrying out bandwidth current limliting
The bandwidth current limliting of one or more applications of processing.
8. a kind of device for realizing bandwidth allocation, which is characterized in that including:Allocation unit and current limliting processing unit;Wherein,
Allocation unit is used for, and bandwidth allocation is carried out to each application;
Current limliting processing unit is used for, and bandwidth current limliting is carried out to the application for completing bandwidth allocation according to the memory access density collection information of application
Processing.
9. device according to claim 8, which is characterized in that the device further includes:
Determination unit, for each application, determination to be applied the access request amount in predetermined period, asked with determining memory access
The memory access density collection information of the amount of asking as application.
10. device according to claim 8, which is characterized in that the allocation unit is specifically used for:
The bandwidth demand that each application is determined according to the access request amount in the memory access density collection information of each application, according to determination
Bandwidth demand to all applications carry out bandwidth allocation;
Wherein, the bandwidth demand of application is directly proportional to the access request amount.
11. according to claim 8~10 any one of them device, which is characterized in that the current limliting processing unit is specifically used for:
By all applications, two are divided into according to the memory access density collection information and one or more preset sub-clustering threshold values
Or more than two bandwidth sub-clusterings;
Each bandwidth sub-clustering to division carries out bandwidth current limliting processing according to preset current limliting strategy.
12. according to claim 8~10 any one of them device, which is characterized in that the current limliting processing unit is specifically used for:
By all applications, bandwidth sensitive cluster is divided into according to the memory access density collection information and preset first sub-clustering threshold value and is prolonged
Sensitivity cluster late;
Bandwidth current limliting processing is carried out according to preset current limliting strategy to bandwidth sensitive cluster.
13. device according to claim 12, which is characterized in that the current limliting processing unit is according to memory access density collection information
Carrying out the processing of bandwidth current limliting includes:
According to the memory access frequency of the memory access frequency and delay-sensitive cluster of memory access density collection information computation bandwidth sensitivity cluster, and calculate band
The memory access frequency ratio of the memory access frequency and the memory access frequency of delay-sensitive cluster of wide sensitivity cluster;
Calculated according to memory access density collection information all applications access request amount it is cumulative and;
When access request amount is cumulative and corresponding bandwidth demand is more than system bandwidth, according to the memory access frequency ratio for calculating acquisition
And access request amount is cumulative and carries out real-time bandwidth current limliting processing to bandwidth sensitive cluster.
14. according to claim 8~10 any one of them device, which is characterized in that described device further includes lifting unit, is used
In when carrying out access request without other application in addition to the application for carrying out bandwidth current limliting processing, release to carrying out at bandwidth current limliting
The bandwidth current limliting of one or more applications of reason.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710152461.4A CN108632170A (en) | 2017-03-15 | 2017-03-15 | A kind of method and device for realizing bandwidth allocation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710152461.4A CN108632170A (en) | 2017-03-15 | 2017-03-15 | A kind of method and device for realizing bandwidth allocation |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108632170A true CN108632170A (en) | 2018-10-09 |
Family
ID=63686251
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710152461.4A Withdrawn CN108632170A (en) | 2017-03-15 | 2017-03-15 | A kind of method and device for realizing bandwidth allocation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108632170A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109995669A (en) * | 2019-04-09 | 2019-07-09 | 深圳前海微众银行股份有限公司 | Distributed current-limiting method, device, equipment and readable storage medium storing program for executing |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070106879A1 (en) * | 2005-11-08 | 2007-05-10 | Hitachi, Ltd. | Semiconductor device |
CN103139099A (en) * | 2012-12-31 | 2013-06-05 | 北京北大众志微系统科技有限责任公司 | Self-adapting method and device for bandwidth regulation |
CN104679593A (en) * | 2015-03-13 | 2015-06-03 | 浪潮集团有限公司 | Task scheduling optimization method based on SMP system |
-
2017
- 2017-03-15 CN CN201710152461.4A patent/CN108632170A/en not_active Withdrawn
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070106879A1 (en) * | 2005-11-08 | 2007-05-10 | Hitachi, Ltd. | Semiconductor device |
CN103139099A (en) * | 2012-12-31 | 2013-06-05 | 北京北大众志微系统科技有限责任公司 | Self-adapting method and device for bandwidth regulation |
CN104679593A (en) * | 2015-03-13 | 2015-06-03 | 浪潮集团有限公司 | Task scheduling optimization method based on SMP system |
Non-Patent Citations (1)
Title |
---|
刘阳国 等: "面向异构多核系统芯片的高效动态带宽划分方法", 《计算机辅助设计与图形学学报》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109995669A (en) * | 2019-04-09 | 2019-07-09 | 深圳前海微众银行股份有限公司 | Distributed current-limiting method, device, equipment and readable storage medium storing program for executing |
CN109995669B (en) * | 2019-04-09 | 2024-05-03 | 深圳前海微众银行股份有限公司 | Distributed current limiting method, device, equipment and readable storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110096349B (en) | Job scheduling method based on cluster node load state prediction | |
CN104750557B (en) | A kind of EMS memory management process and memory management device | |
US20170083474A1 (en) | Distributed memory controller | |
US20130212594A1 (en) | Method of optimizing performance of hierarchical multi-core processor and multi-core processor system for performing the method | |
CN103870341A (en) | Method and system of adjusting resources of virtual machine | |
US9684366B2 (en) | Distributed power management system with plurality of power management controllers controlling zone and component power caps of respective zones by determining priority of other zones | |
CN103595651B (en) | Distributed data stream processing method and system | |
KR20140134190A (en) | Multicore system and job scheduling method thereof | |
DE112012003701T5 (en) | Dynamically allocate a power budget across multiple domains of a processor | |
KR102430934B1 (en) | Memory bandwidth scheduling based on lower bound quality of service | |
CN107145388B (en) | Task scheduling method and system under multi-task environment | |
US20140189181A1 (en) | On-Chip Bus Arbitration Method and Device Thereof | |
US9746909B2 (en) | Computer program product and a node implementing power management by associated power management controllers based on distributed node power consumption and priority data | |
CN110018781B (en) | Disk flow control method and device and electronic equipment | |
CN108959111A (en) | Data storage device and method for flow management | |
US20230376419A1 (en) | Reconfigurable cache architecture and methods for cache coherency | |
CN113037800B (en) | Job scheduling method and job scheduling device | |
US11220688B2 (en) | Oversubscription scheduling | |
KR20100074920A (en) | Apparatus and method for load balancing in multi-core system | |
US20220300323A1 (en) | Job Scheduling Method and Job Scheduling Apparatus | |
CN113032102A (en) | Resource rescheduling method, device, equipment and medium | |
CN115586961A (en) | AI platform computing resource task scheduling method, device and medium | |
CN106293736B (en) | Two-stage programmer and its calculation method for coarseness multicore computing system | |
US20210019258A1 (en) | Performance Telemetry Aided Processing Scheme | |
CN107203256B (en) | Energy-saving distribution method and device under network function virtualization scene |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20181009 |
|
WW01 | Invention patent application withdrawn after publication |