CN107450968A - Load restoring method, device and equipment - Google Patents

Load restoring method, device and equipment Download PDF

Info

Publication number
CN107450968A
CN107450968A CN201610377787.2A CN201610377787A CN107450968A CN 107450968 A CN107450968 A CN 107450968A CN 201610377787 A CN201610377787 A CN 201610377787A CN 107450968 A CN107450968 A CN 107450968A
Authority
CN
China
Prior art keywords
kernel function
dimension
playback unit
load
resource
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201610377787.2A
Other languages
Chinese (zh)
Other versions
CN107450968B (en
Inventor
韩锐
王振涛
袁泉
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 CN201610377787.2A priority Critical patent/CN107450968B/en
Publication of CN107450968A publication Critical patent/CN107450968A/en
Application granted granted Critical
Publication of CN107450968B publication Critical patent/CN107450968B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/48—Program initiating; Program switching, e.g. by interrupt
    • G06F9/4806—Task transfer initiation or dispatching
    • G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
    • G06F9/485—Task life-cycle, e.g. stopping, restarting, resuming execution

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Debugging And Monitoring (AREA)

Abstract

The embodiment of the present invention provides a kind of load restoring method, device and equipment.This method includes：According to the load behavioural information recorded in public cloud daily record, the first kernel function set is obtained from kernel function storehouse；According to the evaluation and test demand of user and the public cloud daily record, obtain load and reset daily record；The load behavior of each kernel function in total resources utilization rate and the first kernel function set of each playback unit in each resource dimension, each playback unit is obtained from the first kernel function set in the second kernel function set corresponding to each resource dimension；Second kernel function set of all playback units in each resource dimension in daily record is reset according to the load, obtains the mixed load for evaluating system performance.This method substantially increases the precision of load reduction.

Description

Load restoring method, device and equipment
Technical field
The present embodiments relate to computer technology, more particularly to a kind of load restoring method, device and equipment.
Background technology
In modern publicly-owned cloud platform, substantial amounts of tenant (i.e. cloud computing user) shares cloud computing infrastructure resources, should Resource includes central processing unit (Central Processing Unit, abbreviation CPU) resource, internal memory and input and output (Input/ Output, abbreviation I/O) resource, for performing various types of operations (such as CPU intensive type, the operation of I/O intensities).Therefore, very More operations (or load) often concurrently perform in the same node of data center, form mixed load.By counting accordingly Data understand that public cloud mixed load often has two characteristic features：First, the load behavior each loaded in mixed load Difference in height alienation, different loads are all each in the pattern of submission, run time, resource requirement, architectural feature even load behavior Differ；Second, the load of a large amount of short-term operations is constantly submitted and completed, the frequent change of the load behavior of mixed load.
In general, when developing some new system, it is necessary to be evaluated and tested to the performance of the system.Therefore, right The load behavior according to the mixed load recorded in public cloud daily record is generally required when system is evaluated and tested, is restored really mixed Load is closed, and it is really behaved in system, by load behavior or the fortune of checking the mixed load actual motion The row time carrys out the performance of evaluating system.Load restoring method of the prior art, it is specifically to be based on Facebook and Yahoo！ Cloud platform daily record, the method reduced to the Hadoop MapReduce operations being widely present in cloud platform, it is specially：
It is that load behavior to MapReduce operations is described first, then evaluation and test time to be required according to user Section, is sampled to true daily record, obtains load and resets daily record；Finally, daily record generation MapReduce is reset based on the load to bear Input data is carried, further the MapReduce operations based on the generation emulation of I/O resource dimensions in Hadoop platform.
But load restoring method of the prior art, the reduction of load is only considered from I/O resource dimension, it does not have Consider the actual execution logic of load, i.e., do not consider CPU, memory source utilization rate, and corresponding architecture load Feature, it is not high that it loads the accuracy of reduction.
The content of the invention
The embodiment of the present invention provides a kind of load restoring method, device and equipment, to solve to load in the prior art also The technical problem that former consideration dimension is single, load reduction accuracy is not high
In a first aspect, the embodiment of the present invention provides a kind of load restoring method, including：
According to the load behavioural information recorded in public cloud daily record, the first kernel function set is obtained from kernel function storehouse；Its In, the first kernel function set includes CPU core function set and space kernel function set；The CPU core function set includes Multiple CPU core functions, the space kernel function set include the set of internal memory kernel function and/or IO kernel function set, the internal memory Kernel function set includes multiple internal memory kernel functions, and the IO kernel functions set includes multiple IO kernel functions；
According to the evaluation and test demand of user and the public cloud daily record, obtain load and reset daily record；Wherein, the load is reset Each playback unit in daily record includes all loads in current playback unit total money in each resource dimension respectively Source utilization rate；
According to total resources utilization rate of each playback unit in each resource dimension and the first kernel function collection The load behavior of each kernel function in conjunction, each playback unit is obtained from the first kernel function set in each institute State the second kernel function set corresponding to resource dimension；Wherein, the same kernel function in the second kernel function set is described Resource utilization sum in resource dimension is no more than the total resources utilization rate in the resource dimension；
Second kernel function set of all playback units in each resource dimension in daily record is reset according to the load, Obtain the mixed load for evaluating system performance.
The load restoring method that above-mentioned first aspect is provided, by being believed according to the load behavior recorded in public cloud daily record Breath, the first kernel function set is obtained from kernel function storehouse, and weight is loaded according to the evaluation and test demand and public cloud log acquisition of user Daily record is put, so that load reduction apparatus is according to total resources utilization rate of each playback unit in each resource dimension and the The load behavior of each kernel function in one kernel function set, each playback unit is obtained from the first kernel function set each Second kernel function set corresponding to resource dimension, and then all playback units in daily record are reset according to load and tieed up in each resource The second kernel function set on degree, obtains the mixed load for evaluating system performance.Method provided in an embodiment of the present invention, When carrying out load reduction, it is contemplated that the load restraint of multiple resource dimensions, its mixed load accuracy restored is high, and Load is reduced using kernel function in the embodiment of the present application, can be by shortening the run time of kernel function itself so that The time of system evaluation shortens, and substantially increases the evaluation and test efficiency of system.
Also include all loads point in current playback unit in a kind of possible design, in each playback unit Target architecture characteristic value not in each resource dimension；All playback lists reset according to the load in daily record Second kernel function set of the member in each resource dimension, obtains the mixed load for evaluating system performance, specifically includes：
The second kernel function set according to corresponding to each playback unit in each resource dimension determines each institute State kernel function sequence of the playback unit in each resource dimension；
According to target architecture characteristic value of each playback unit in each resource dimension, from the core Optimal kernel function combination of each playback unit in each resource dimension is determined in the sequence of function；Wherein, it is described All kernel functions in optimal kernel function combination are tieed up in architectural feature value caused by the system operation and the resource The error of target architecture characteristic value on degree is minimum；
Optimal kernel function of all playback units in each resource dimension reset according to the load in daily record combines, Obtain the mixed load for evaluating system performance.
In a kind of possible design, each target of the playback unit in each resource dimension of the basis Architectural feature value, determine each playback unit in each resource dimension most from the kernel function sequence Excellent kernel function combination, is specifically included：
According to target architecture characteristic value of each playback unit in each resource dimension, using dynamic The method of planning determines each optimal core of the playback unit in each resource dimension from the kernel function sequence Combination of function.
The there is provided load restoring method of possible design, by according to each playback unit in each resource dimension pair The the second kernel function set answered determines kernel function sequence of each playback unit in each resource dimension, and according to each playback Target architecture characteristic value of the unit in each resource dimension, determines each playback unit each from kernel function sequence Optimal kernel function combination in resource dimension, and then all playback units in load playback daily record are in each resource dimension On optimal kernel function combination, obtain the mixed load for evaluating system performance.By fully being examined when carrying out load reduction The architectural feature of worry load, select and reset the target architecture characteristic value error recorded in daily record minimum with load The combination of optimal kernel function so that the mixed load restored run in system to be evaluated caused by architecture The error of characteristic value and the target architecture characteristic value in respective resources dimension is minimum, substantially increases the reduction of mixed load Accuracy rate.
In a kind of possible design, when obtaining optimal core letter of each playback unit in space resources dimension After array is closed, methods described also includes：
It is corresponding in cpu resource to obtain optimal kernel function combination of each playback unit in the space resources dimension The load behavioural information of dimension；Wherein, the space resources dimension includes memory source dimension and/or I/O resource dimension, described Optimal kernel function combination corresponding load behavioural information bag in cpu resource dimension of the playback unit in the space resources dimension Include CPU spatial contents and the caused load estimated performance CPI that the optimal kernel function combination need to take when being run in system Value.
In a kind of possible design, the space kernel function collection is combined into the set of internal memory kernel function or IO kernel function set；
Each total resources utilization rate and the first core letter of the playback unit in each resource dimension of the basis The load behavior of each kernel function in manifold conjunction, each playback unit is obtained from the first kernel function set every The second kernel function set, is specifically included corresponding to the individual resource dimension：
Step A：According to total resources utilization rate of the playback unit in space resources dimension, the playback unit is determined The first spatial content, and according to the sky of every kind of space kernel function in first spatial content and the space kernel function set Between resource utilization, it is determined that the maximum of every kind of space kernel function uses number under first spatial content；
Step B：Number is used according to the maximum of every kind of space kernel function under first spatial content, determined described heavy Unit is put in the second kernel function set corresponding to the space resources dimension；
Step C：Combined according to optimal kernel function of the playback unit in the space resources dimension in system operation When total resources utilization rate in cpu resource dimension of the CPU spatial contents and the playback unit that need to take, determine the playback The second space capacity of unit；
Step D：According to the cpu resource of every kind of CPU core function in the second space capacity and the CPU core function set Utilization rate, it is determined that the maximum of every kind of CPU core function uses number under the second space capacity；
Step E：Number is used according to the maximum of every kind of CPU core function under the second space capacity, determined described heavy Unit is put in the second kernel function set corresponding to the cpu resource dimension；
Step F：Each playback unit in daily record is reset to the load and performs the step A to the step E respectively, All playback units in the load playback daily record are obtained respectively second corresponding to space resources dimension and cpu resource dimension Kernel function set.
In a kind of possible design, each target of the playback unit in each resource dimension of the basis Architectural feature value, determine each playback unit each from the kernel function sequence using the method for Dynamic Programming Optimal kernel function combination in the resource dimension, is specifically included：
Step G：According to first object architectural feature value of the playback unit in space resources dimension, using dynamic The method of state planning determines the playback unit in space from the playback unit in the kernel function sequence of space resources dimension Optimal kernel function combination in resource dimension；
Step H：Combined according to optimal kernel function of the playback unit in the space resources dimension in system operation The the second target architecture characteristic value of caused CPI value and the playback unit in cpu resource dimension, it is determined that described Threeth target architecture characteristic value of the playback unit in cpu resource dimension；
Step I：According to threeth target architecture characteristic value of the playback unit in cpu resource dimension, using dynamic The method of state planning determines that the playback unit provides in CPU from the playback unit in the kernel function sequence of cpu resource dimension Optimal kernel function combination in the dimension of source；
Step J：Each playback unit in daily record is reset to the load and performs the step G to the step I respectively, All playback units in the load playback daily record are obtained respectively optimal corresponding to space resources dimension and cpu resource dimension Kernel function combines.
The there is provided load restoring method of the possible design, by considering the information of public cloud tenant, associative function storehouse In kernel function take out different types of operation (or load), from the distribution situation of multiple resource dimensions analysis load, and The playback unit in daily record is reset with reference to load, the optimal kernel function group of different resource dimension is obtained for each playback unit Close, load reduction is carried out based on the optimal kernel function combination, the mixed load restored can reappear the true of data center Scene, precision is higher, can quantify and evaluation and test foundation with a high credibility so as to provide neck for evaluating system；Meanwhile using dynamic The method of state planning obtains each playback unit and combined in the optimal kernel function of each resource dimension, substantially increases optimal core letter The acquisition efficiency that array is closed, and can also be by adjusting the run time of kernel function, to shorten the evaluation and test time of system, necessarily The evaluation and test efficiency of systematic function is improved in degree.
In a kind of possible design, the load behavioural information that is recorded in the daily record according to public cloud, from kernel function storehouse The first kernel function set of middle acquisition, is specifically included：
Load behavioural information according to being recorded in the public cloud daily record determines load behavior scope；
The load behavior of every kind of kernel function in the kernel function storehouse and the load behavior scope, from the core The kernel function for meeting the load behavior scope is obtained in function library, forms the first kernel function set.
The there is provided load restoring method of the possible design, by being believed according to the load behavior recorded in public cloud daily record Breath determines a load behavior scope, and the load behavior of every kind of kernel function in kernel function storehouse and the load behavior model Enclose, the kernel function for meeting the load behavior scope is obtained from kernel function storehouse, the first kernel function set is formed, so as to ensure first Kernel function in kernel function set has certain typicalness, ensures subsequently it is determined that optimal kernel function in each resource dimension During combination, the first kernel function set can provide more preferable foundation, ensure that the validity and accuracy of load reduction.
In a kind of possible design, in machine type that the evaluation and test demand of the user includes recording in daily record, daily record The number of machine and at least one in the period of daily record corresponding to the machine type of record.
In a kind of possible design, the evaluation and test demand according to user and the public cloud daily record, load weight is obtained Daily record is put, is specifically included：
Daily record fragment is chosen from the public cloud daily record according to the evaluation and test demand of the user, wrapped in the daily record fragment Multiple first playback units are included, first playback unit includes each being supported on each resource dimension in current first playback unit Resource utilization on degree；
The resource utilization being each supported in each resource dimension in each first playback unit is entered Row sum operation, obtain total resources utilization rate of each first playback unit in each resource dimension；
To the architectural feature being each supported in each resource dimension in each first playback unit Value carries out mean value computation, obtains target architecture feature of each first playback unit in each resource dimension Value；
According to total resources utilization rate of each first playback unit in each resource dimension, each described the Target architecture characteristic value of one playback unit in each resource dimension, obtain the load and reset daily record.
The there is provided load restoring method of possible design, by the evaluation and test demand according to the user from public cloud day Daily record fragment is chosen in will, and the content of each first playback unit in the daily record fragment is changed, obtains load weight The content of the playback unit in daily record is put, and then obtains load and resets daily record, so that load reduction apparatus is being loaded During reduction, it can be reduced based on fine-grained playback unit so that the precision of the load restored is higher.
Second aspect, the embodiment of the present invention provide a kind of load reduction apparatus, including：
Kernel function analyzes acquisition module, for according to the load behavioural information recorded in public cloud daily record, from kernel function storehouse The first kernel function set of middle acquisition；Wherein, the first kernel function set includes CPU core function set and space kernel function collection Close；The CPU core function set includes multiple CPU core functions, the space kernel function set include the set of internal memory kernel function and/ Or IO kernel function set, the internal memory kernel function set include multiple internal memory kernel functions, the IO kernel functions set includes multiple IO kernel functions；
Log acquisition module is reset, for the evaluation and test demand according to user and the public cloud daily record, load is obtained and resets Daily record；Wherein, each playback unit that the load is reset in daily record includes all loads difference in current playback unit Total resources utilization rate in each resource dimension；
Kernel function matching module, for according to each total resources utilization rate of the playback unit in each resource dimension With the load behavior of each kernel function in the first kernel function set, each institute is obtained from the first kernel function set Playback unit is stated in the second kernel function set corresponding to each resource dimension；Wherein, in the second kernel function set The total resources that resource utilization sum of the same kernel function in the resource dimension is no more than in the resource dimension use Rate；
Recovery module is loaded, for resetting all playback units in daily record according to the load in each resource dimension The second kernel function set, obtain the mixed load for evaluating system performance.
Also include all loads point in current playback unit in a kind of possible design, in each playback unit Target architecture characteristic value not in each resource dimension；The load recovery module, is specifically included：
Sequence determination unit, for each playback unit of basis in the second core letter corresponding to each resource dimension Manifold, which is closed, determines each kernel function sequence of the playback unit in each resource dimension；
Combinatorial Optimization unit, for according to each objective body tying of the playback unit in each resource dimension Structure characteristic value, each optimal core letter of the playback unit in each resource dimension is determined from the kernel function sequence Array is closed；Wherein, all kernel functions in the optimal kernel function combination are special in architecture caused by the system operation The error of value indicative and the target architecture characteristic value in the resource dimension is minimum；
Reduction unit is loaded, for resetting all playback units in daily record according to the load in each resource dimension Optimal kernel function combination, obtain the mixed load for evaluating system performance.
In a kind of possible design, the Combinatorial Optimization unit, specifically for each playback unit of basis every Target architecture characteristic value in the individual resource dimension, determined using the method for Dynamic Programming from the kernel function sequence Optimal kernel function combination of each playback unit in each resource dimension.
In a kind of possible design, described device also includes：
Acquisition module, for obtaining each playback unit in space resources dimension in the Combinatorial Optimization unit The combination of optimal kernel function after, obtain optimal kernel function combination of each playback unit in the space resources dimension The corresponding load behavioural information in cpu resource dimension；Wherein, the space resources dimension includes memory source dimension and/or IO Resource dimension, optimal kernel function combination of the playback unit in the space resources dimension are corresponding in cpu resource dimension Load behavioural information includes CPU spatial contents that the optimal kernel function combination need to take when being run in system and caused negative Carry estimated performance CPI value.
In a kind of possible design, the space kernel function collection is combined into the set of internal memory kernel function or IO kernel function set； The kernel function matching module, is specifically included：
Determining unit, for the total resources utilization rate according to the playback unit in space resources dimension, it is determined that described First spatial content of playback unit, and according to every kind of space core in first spatial content and the space kernel function set The space resources utilization rate of function, it is determined that the maximum of every kind of space kernel function uses number under first spatial content；With And number is used according to the maximum of every kind of space kernel function under first spatial content, determine the playback unit in institute State the second kernel function set corresponding to space resources dimension, and according to the playback unit in the space resources dimension most The CPU spatial contents and the playback unit that excellent kernel function combination need to take in system operation are total in cpu resource dimension Resource utilization, determine the second space capacity of the playback unit；And according to the second space capacity and the CPU The cpu resource utilization rate of every kind of CPU core function in kernel function set, it is determined that every kind of CPU core letter under the second space capacity Several maximums uses number, and uses number according to the maximum of every kind of CPU core function under the second space capacity, determines institute Playback unit is stated in the second kernel function set corresponding to the cpu resource dimension；
Cycling element, for indicating that the determining unit obtains all playback units difference in the load playback daily record In the second kernel function set corresponding to space resources dimension and cpu resource dimension.
In a kind of possible design, the Combinatorial Optimization unit, specifically include：
Dynamic optimization subelement, for the first object architecture according to the playback unit in space resources dimension Characteristic value, determined using the method for Dynamic Programming in the kernel function sequence of space resources dimension from the playback unit described heavy Optimal kernel function combination of the unit in space resources dimension is put, and according to the playback unit in the space resources dimension Optimal kernel function combination CPI caused by system operation value and the playback unit in cpu resource dimension second Target architecture characteristic value, determine threeth target architecture characteristic value of the playback unit in cpu resource dimension；With And the 3rd target architecture characteristic value according to the playback unit in cpu resource dimension, using the method for Dynamic Programming Determine the playback unit in cpu resource dimension most in the kernel function sequence of cpu resource dimension from the playback unit Excellent kernel function combination；
Subelement is circulated, for indicating that it is all heavy in the load playback daily record that the dynamic optimization unit obtains respectively Unit is put respectively in optimal kernel function combination corresponding to space resources dimension and cpu resource dimension.
In a kind of possible design, the kernel function analyzes acquisition module, specifically for according to the public cloud daily record The load behavioural information of middle record determines load behavior scope, and the load row of every kind of kernel function in the kernel function storehouse For and the load behavior scope, obtained from the kernel function storehouse and meet the kernel function of the load behavior scope, formed The first kernel function set.
In a kind of possible design, in machine type that the evaluation and test demand of the user includes recording in daily record, daily record The number of machine and at least one in the period of daily record corresponding to the machine type of record.
In a kind of possible design, the playback log acquisition module, specifically include：
Daily record fragment chooses unit, and daily record is chosen from the public cloud daily record for the evaluation and test demand according to the user Fragment, the daily record fragment include multiple first playback units, and first playback unit includes current first playback unit In be each supported on resource utilization in each resource dimension；
Processing unit, for being each supported in each resource dimension in each first playback unit Resource utilization carries out sum operation, and obtaining total resources of each first playback unit in each resource dimension makes With rate, and to the architectural feature value being each supported in each resource dimension in each first playback unit Mean value computation is carried out, obtains target architecture feature of each first playback unit in each resource dimension Value；
Log acquisition unit is reset, for total in each resource dimension according to each first playback unit Resource utilization, each target architecture characteristic value of first playback unit in each resource dimension, are obtained Daily record is reset in the load.
The there is provided load reduction apparatus of each possible design of above-mentioned second aspect and second aspect, its advantage Beneficial effect caused by being referred to each possible design of above-mentioned first aspect and first aspect, will not be repeated here.
The third aspect, the embodiment of the present invention provide a kind of load reduction apparatus, including：
Processor, for according to the load behavioural information recorded in public cloud daily record, the first core to be obtained from kernel function storehouse Function set, and the evaluation and test demand according to user and the public cloud daily record, obtain load and reset daily record；And according to each Total resources utilization rate of the playback unit in each resource dimension and each kernel function in the first kernel function set it is negative Load behavior, each playback unit is obtained from the first kernel function set second corresponding to each resource dimension Kernel function set, and second kernel function of all playback units in the load playback daily record in each resource dimension Set, obtains the mixed load for evaluating system performance；Wherein, the first kernel function set includes CPU core function set With space kernel function set；The CPU core function set includes multiple CPU core functions, and the space kernel function set includes interior Kernel function set and/or IO kernel function set are deposited, the internal memory kernel function set includes multiple internal memory kernel functions, the IO cores letter Manifold, which is closed, includes multiple IO kernel functions；Each playback unit that the load is reset in daily record includes in current playback unit It is all to load the total resources utilization rate in each resource dimension respectively；Same kernel function in the second kernel function set Resource utilization sum in the resource dimension is no more than the total resources utilization rate in the resource dimension.
Also include all loads point in current playback unit in a kind of possible design, in each playback unit Target architecture characteristic value not in each resource dimension；
The processor, specifically for each playback unit of basis in the second core corresponding to each resource dimension Function set determines each kernel function sequence of the playback unit in each resource dimension, and according to each described heavy Target architecture characteristic value of the unit in each resource dimension is put, is determined from the kernel function sequence each described Optimal kernel function combination of the playback unit in each resource dimension；And the institute in the load playback daily record There is optimal kernel function combination of the playback unit in each resource dimension, obtain the mixed load for evaluating system performance；Its In, all kernel functions in the optimal kernel function combination architectural feature value caused by the system operation with it is described The error of target architecture characteristic value in resource dimension is minimum.
In a kind of possible design, the processor, specifically for each playback unit of basis each described Target architecture characteristic value in resource dimension, each institute is determined from the kernel function sequence using the method for Dynamic Programming State optimal kernel function combination of the playback unit in each resource dimension.
In a kind of possible design, the processor, it is additionally operable to obtaining each playback unit in space money After optimal kernel function combination in the dimension of source, each optimal core of the playback unit in the space resources dimension is obtained The corresponding load behavioural information in cpu resource dimension of combination of function；Wherein, the space resources dimension includes memory source dimension And/or I/O resource dimension, optimal kernel function combination of the playback unit in the space resources dimension are corresponding in cpu resource The load behavioural information of dimension includes CPU spatial contents and the production that the optimal kernel function combination need to take when being run in system Raw load estimated performance CPI value.
In a kind of possible design, the space kernel function collection is combined into the set of internal memory kernel function or IO kernel function set；
The processor, specifically for the total resources utilization rate according to the playback unit in space resources dimension, really First spatial content of the fixed playback unit, and according to every kind of in first spatial content and the space kernel function set The space resources utilization rate of space kernel function, it is determined that the maximum use of every kind of space kernel function is individual under first spatial content Number, and number is used according to the maximum of every kind of space kernel function under first spatial content, determine that the playback unit exists Second kernel function set corresponding to the space resources dimension；And according to the playback unit in the space resources dimension On the optimal kernel function CPU spatial contents that need to be taken in system operation of combination and the playback unit in cpu resource dimension On total resources utilization rate, determine the second space capacity of the playback unit, and according to the second space capacity and described The cpu resource utilization rate of every kind of CPU core function in CPU core function set, it is determined that every kind of CPU core under the second space capacity The maximum of function uses number, and then uses number according to the maximum of every kind of CPU core function under the second space capacity, really The fixed playback unit obtains the load and resets daily record in the second kernel function set corresponding to the cpu resource dimension In all playback units respectively in the second kernel function set corresponding to space resources dimension and cpu resource dimension.
In a kind of possible design, the processor, specifically for according to the playback unit in space resources dimension On first object architectural feature value, using Dynamic Programming method from the playback unit space resources dimension core Optimal kernel function combination of the playback unit in space resources dimension is determined in the sequence of function, and according to the playback unit Optimal kernel function combination in the space resources dimension exists in the value of CPI caused by system operation and the playback unit The second target architecture characteristic value in cpu resource dimension, determine threeth mesh of the playback unit in cpu resource dimension Mark system Structural Eigenvalue, and the 3rd target architecture characteristic value according to the playback unit in cpu resource dimension, In the kernel function sequence of cpu resource dimension determine that the playback unit exists from the playback unit using the method for Dynamic Programming Optimal kernel function combination in cpu resource dimension, and then obtain all playback units that the load is reset in daily record and exist respectively Optimal kernel function combination corresponding to space resources dimension and cpu resource dimension.
In a kind of possible design, the processor, specifically for according to the load recorded in the public cloud daily record Behavioural information determines load behavior scope, and the load behavior of every kind of kernel function in the kernel function storehouse and described negative Load behavior scope, the kernel function for meeting the load behavior scope is obtained from the kernel function storehouse, form the first core letter Manifold is closed.
In a kind of possible design, in machine type that the evaluation and test demand of the user includes recording in daily record, daily record The number of machine and at least one in the period of daily record corresponding to the machine type of record.
In a kind of possible design, the processor, specifically for the evaluation and test demand according to the user from the public affairs There is selection daily record fragment in cloud daily record, and to being each supported in each resource dimension in each first playback unit Resource utilization carries out sum operation, and obtaining total resources of each first playback unit in each resource dimension makes With rate；And to the architectural feature being each supported in each resource dimension in each first playback unit Value carries out mean value computation, obtains target architecture feature of each first playback unit in each resource dimension Value, and according to total resources utilization rate of each first playback unit in each resource dimension, each described first Target architecture characteristic value of the playback unit in each resource dimension, obtain the load and reset daily record；The day Master chip section includes multiple first playback units, and first playback unit includes each being supported in current first playback unit Resource utilization in each resource dimension.
The there is provided load reduction apparatus of each possible design of the above-mentioned third aspect and the third aspect, its advantage Beneficial effect caused by being referred to each possible design of above-mentioned first aspect and first aspect, will not be repeated here.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing There is the required accompanying drawing used in technology description to be briefly described, it should be apparent that, drawings in the following description are this hairs Some bright embodiments, for those of ordinary skill in the art, without having to pay creative labor, can be with Other accompanying drawings are obtained according to these accompanying drawings.
Fig. 1 is the schematic flow sheet of load restoring method embodiment one provided in an embodiment of the present invention；
Fig. 2 is the schematic flow sheet of load restoring method embodiment two provided in an embodiment of the present invention；
Fig. 3 is the schematic flow sheet of dynamic programming method provided in an embodiment of the present invention；
Fig. 4 is the schematic flow sheet of load restoring method embodiment three provided in an embodiment of the present invention；
Fig. 5 is the schematic flow sheet of load restoring method example IV provided in an embodiment of the present invention；
Fig. 6 is the schematic flow sheet of load restoring method embodiment five provided in an embodiment of the present invention；
Fig. 7 is the structural representation of load reduction apparatus embodiment one provided in an embodiment of the present invention；
Fig. 8 is the structural representation of load reduction apparatus embodiment two provided in an embodiment of the present invention；
Fig. 9 is the structural representation of load reduction apparatus embodiment three provided in an embodiment of the present invention；
Figure 10 is the structural representation of load reduction apparatus example IV provided in an embodiment of the present invention；
Figure 11 is the structural representation of load reduction apparatus embodiment one provided in an embodiment of the present invention.
Embodiment
To make the purpose, technical scheme and advantage of the embodiment of the present invention clearer, below in conjunction with the embodiment of the present invention In accompanying drawing, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that described embodiment is Part of the embodiment of the present invention, rather than whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art The every other embodiment obtained under the premise of creative work is not made, belongs to the scope of protection of the invention.
It should be appreciated that term "and/or" used herein is only a kind of incidence relation for describing affiliated partner, represent There may be three kinds of relations, for example, A and/or B, can be represented：Individualism A, while A and B be present, individualism B these three Situation.In addition, character "/" herein, it is a kind of relation of "or" to typically represent forward-backward correlation object.
It will be appreciated that though XXX may be described using term first, second, third, etc. in embodiments of the present invention, but These XXX should not necessarily be limited by these terms.These terms are only used for XXX being distinguished from each other out.For example, implementation of the present invention is not being departed from In the case of example scope, the first XXX can also be referred to as the 2nd XXX, and similarly, the 2nd XXX can also be referred to as the first XXX.
Depending on linguistic context, word as used in this " if ", " if " can be construed to " ... when " or " when ... " or " in response to determining " or " in response to detection ".Similarly, depending on linguistic context, phrase " if it is determined that " or " such as Fruit detects (condition or event of statement) " can be construed to " when it is determined that when " or " in response to determine " or " when detection (statement Condition or event) when " or " in response to detect (condition or event of statement) ".
It should also be noted that, term " comprising ", "comprising" or its any other variant are intended to nonexcludability Comprising, so that commodity or system including a series of elements not only include those key elements, but also including without clear and definite The other element listed, or also include for this commodity or the intrinsic key element of system.In the feelings not limited more Under condition, the key element that is limited by sentence "including a ...", it is not excluded that in the commodity including the key element or system also Other identical element be present.
Load restoring method, device and equipment provided in an embodiment of the present invention, go for being based on publicly-owned cloud platform day The load reduction of will, pass through reduced load to evaluate and test the scene of the performance of developed system.Optionally, reduced here Load out can be mixed load, and the mixed load can include multiple loads (or multiple operations), each load tool There is the load behavior of its own.Optionally, the load behavior can include：Submission pattern, run time, resource is loaded to use Rate, architectural feature etc..Optionally, resource here can include at least one in cpu resource, memory source and I/O resource Kind；Also, same to be supported on the type of resource utilization and architectural feature corresponding to different resource dimensions be different , optionally, for same load, in cpu resource dimension, its resource utilization is the utilization rate of cpu resource, its Architectural feature is load estimated performance (Cycles Per Instruction, abbreviation CPI), and the CPI refers to that CPU is each Clock cycle number required for instruction；In internal memory resource dimension, its resource utilization is the utilization rate of memory source, its body Architecture is characterized as load memory access characteristic (Memory Accesses per Instruction, abbreviation MAI), and the MAI is referred to The internal memory memory access number of each instructions of CPU；In I/O resource dimension, its resource utilization is the utilization rate of I/O resource, its system Architectural feature is load IO characteristics (Input/Output Operations Per Second, abbreviation IOPS), what the IOPS referred to It is read-write IO per second number.
Load restoring method, device and equipment involved by the embodiment of the present invention, it aims to solve the problem that and loaded in the prior art The technical problem that the consideration dimension of reduction is single, load reduction accuracy is not high.Optionally, the method for the present embodiment can be negative Reduction apparatus is carried, can also be load reduction apparatus, the load reduction apparatus can be integrated in load reduction apparatus.Following realities Applying example to have executive agent is illustrated exemplified by load reduction apparatus.
Technical scheme is described in detail with specifically embodiment below.These specific implementations below Example can be combined with each other, and may be repeated no more for same or analogous concept or process in some embodiments.
Fig. 1 is the schematic flow sheet of load restoring method embodiment one provided in an embodiment of the present invention.The present embodiment is related to Be load reduction apparatus by from different resource dimensions to load reduce, improve load reduction accuracy specific mistake Journey.As shown in figure 1, this method includes：
S101：According to the load behavioural information recorded in public cloud daily record, the first kernel function collection is obtained from kernel function storehouse Close.
Wherein, the first kernel function set includes CPU core function set and space kernel function set；The CPU core letter Manifold, which is closed, includes multiple CPU core functions, and the space kernel function set includes the set of internal memory kernel function and/or IO kernel function collection Close, the internal memory kernel function set includes multiple internal memory kernel functions, and the IO kernel functions set includes multiple IO kernel functions.
Specifically, public cloud daily record can record the load behavior letter of each load on each node under publicly-owned cloud platform Breath, when the load behavioural information can include being supported on resource utilization in different resource dimension, architectural feature, operation Between, the information such as submission pattern.CPU core function set and space kernel function set, the CPU core can be included in above-mentioned kernel function storehouse Function set can include multiple CPU core functions, the certain load behavior in cpu resource dimensional representation of each CPU functions, example Such as characterize cpu resource utilization rate and CPI；Above-mentioned space kernel function set can include the set of internal memory kernel function and/or IO cores Function set, the internal memory kernel function set include multiple internal memory kernel functions, and each internal memory kernel function is in internal memory resource dimension also table Certain load behavior is levied, such as characterize memory source utilization rate and MAI, the IO kernel functions set to include multiple IO Kernel function, each IO kernel functions also characterize certain load behavior in I/O resource dimension, for example, characterize I/O resource utilization rate with And IOPS.
Therefore load reduction apparatus can analyze the load behavior (load characteristic) of each kernel function in kernel function storehouse, by it Matched with the load behavioural information recorded in public cloud daily record, the kernel function of matching is screened, form the first core letter Manifold is closed.Optionally, matching operation here can be that load behavior is less than to the maximum load row recorded in public cloud daily record Screen to form the first kernel function set for all kernel functions of information, can also be and load behavior is more than public cloud daily record The minimum load behavioural information of middle record and all core letters for being less than the maximum load behavioural information recorded in public cloud daily record Number sieve is elected to form the first kernel function set, can also be other matching ways, how the embodiment of the present invention is to filter out The mode of first kernel function set does not limit.
S102：According to the evaluation and test demand of user and the public cloud daily record, obtain load and reset daily record.
Wherein, each playback unit that the load is reset in daily record includes all loads point in current playback unit Total resources utilization rate not in each resource dimension.
Specifically, after load reduction apparatus obtains the first kernel function set, load reduction apparatus obtains commenting for user Survey demand, optionally, user, which can directly input the evaluation and test demand for treating evaluating system, gives load reduction apparatus, can also be User will be write the evaluation and test demand for treating evaluating system as a formatted file, and load reduction apparatus can be come by resolution file The evaluation and test demand of user is obtained, in a word, the evaluation and test demand of user of the embodiment of the present invention to how to obtain does not limit.It is optional , the evaluation and test demand is used to characterize the requirement that user resets load daily record, and optionally, the evaluation and test demand can include load weight Put the machine type recorded in daily record, the number of machine corresponding to the machine type recorded in daily record is reset in load and negative Load-carrying is put at least one in the period of daily record.
After loading reduction apparatus and getting the evaluation and test demand of user, according to the evaluation and test demand of the user from public cloud day The load playback daily record for meeting that user evaluates and tests demand is filtered out in will, daily record is reset in the load can be according to certain playback unit (such as being divided by the second) is divided into multiple playback units, and each playback unit includes all loads point in current playback unit Total resources utilization rate not in each resource dimension, including all be supported in current playback unit in cpu resource dimension Overall availability, the overall availability in internal memory resource dimension and the overall availability in I/O resource dimension.
S103：According to total resources utilization rate of each playback unit in each resource dimension and the first core letter The load behavior of each kernel function in manifold conjunction, each playback unit is obtained from the first kernel function set every Second kernel function set corresponding to the individual resource dimension.
Wherein, resource utilization of the same kernel function in the second kernel function set in the resource dimension it With no more than the total resources utilization rate in the resource dimension.
Technical scheme for convenience of description, illustrated here with playback unit A, other reset single The operation of member is referred to the playback unit to carry out, i.e., the embodiment of the present invention is identical for the operation of each playback unit 's.
Specifically, when load reduction apparatus get total resources utilization rates of the playback unit A in each resource dimension it Afterwards, the load behavior of each kernel function in reduction apparatus syncaryon function library is loaded, for example, with reference to each kernel function in different moneys The resource utilization of source dimension, playback unit A is obtained from the first kernel function set in the second core corresponding to each resource dimension Function set, resource utilization sum of the same kernel function in the resource dimension in the second kernel function set are no more than Total resources utilization rate in resource dimension.
For example, it is assumed that all cpu resource overall availabilities for being supported on cpu resource dimension are 30% in playback unit A, inside The memory source overall availability for depositing resource dimension is 40%, is 20% in the I/O resource overall availability of I/O resource dimension, the first core Function set, which includes CPU core function set, the set of internal memory kernel function and IO kernel function set, CPU core function set, includes 3 Individual CPU core function, it is CPU core function 1, CPU core function 2, CPU core function 3 respectively；The set of internal memory kernel function includes 2 internal memories Kernel function, it is internal memory kernel function 1 and internal memory kernel function 2 respectively；The set of IO kernel functions includes 2 IO kernel functions, is IO cores respectively Function 1 and IO kernel functions 2.Wherein, the cpu resource utilization rate of CPU core function 1 is that the cpu resource of 8%, CPU core function 2 uses Rate is 12%, the cpu resource utilization rate of CPU core function 3 is 10%, and the memory source utilization rate of internal memory kernel function 1 is 15%, interior The memory source utilization rate for depositing kernel function 2 is that the memory source utilization rate of 20%, IO kernel functions 1 is the interior of 5%, IO kernel functions 2 Resource utilization is deposited as 10%.
Based on data above, load reduction apparatus can determine playback unit A in second function corresponding to cpu resource dimension Set A includes 3 CPU core functions, 1,2 CPU core function 2 and 3 CPU core functions 3, i.e., should for CPU core function 1 Resource utilization sum of the CPU core function 1 in cpu resource dimension is that 8%+8%+8%=24% is less than 30%, CPU core letter Resource utilization sum of the number 2 in cpu resource dimension is that 12%+12%=24% is less than 30%, and CPU core function 3 provides in CPU Resource utilization sum in the dimension of source is also no more than 30% for 10%+10%+10%=30%；Similarly, including playback unit A Depositing the second kernel function set B corresponding to resource dimension includes 2 internal memory kernel functions 1 and 2 internal memory kernel functions 2, i.e., for internal memory For kernel function 1, resource utilization sum of the internal memory kernel function 1 in internal memory resource dimension is that 15%+15%=30% is small In 40%, resource utilization sum of the internal memory kernel function 2 in internal memory resource dimension is that 20%+20%=40% is also no more than 40%；Playback unit A includes 4 IO kernel functions 1 and 2 IO kernel functions in the second kernel function set C corresponding to I/O resource dimension 2, i.e., for IO kernel functions 1, resource utilization sum of the IO kernel functions 1 in I/O resource dimension is 5%+5%+5%+ 5%=20% is 10%+10%=20% no more than resource utilization sum of 20%, the IO kernel functions 2 in I/O resource dimension Also it is no more than 20%.
In summary, playback unit A corresponds to a second kernel function set A in cpu resource dimension, tieed up in memory source A second kernel function set B is corresponded on degree, one second and function set C is corresponded in I/O resource dimension.According to above-mentioned side Method, load reduction apparatus can obtain each playback unit in the second kernel function set corresponding to each resource dimension.
It should be noted that working as in above-mentioned first kernel function set only includes CPU core function set and internal memory kernel function collection Close, then corresponding resource dimension is just cpu resource dimension and memory source dimension, that is, each playback will be obtained by loading reduction apparatus Unit is in cpu resource dimension and in the second kernel function set corresponding to internal memory resource dimension；When in above-mentioned first kernel function set Only include CPU core function set and IO kernel function set, then corresponding resource dimension is just tieed up for cpu resource dimension and I/O resource Degree, that is, each playback unit will be obtained in cpu resource dimension and in the second core letter corresponding to I/O resource dimension by loading reduction apparatus Manifold is closed；When above-mentioned first kernel function set includes CPU core function set, the set of internal memory kernel function and IO kernel function set, Then corresponding resource dimension is just cpu resource dimension, memory source dimension and I/O resource dimension, that is, loading reduction apparatus will obtain Each playback unit is in the second kernel function set corresponding to cpu resource dimension, memory source dimension and I/O resource dimension.
S104：Second kernel function of all playback units in each resource dimension in daily record is reset according to the load Set, obtains the mixed load for evaluating system performance.
Specifically, when load reduction apparatus obtains second kernel function set of each playback unit in each resource dimension Afterwards, kernel function of the reduction apparatus can based on any combination in the second kernel function set in each resource dimension is loaded Load reduction is carried out, so as to obtain the mixed load for evaluating system performance.The mixing restored using the above method is born Carry, it is contemplated that the load restraint of multiple resource dimensions, its mixed load accuracy restored is high, and the embodiment of the present application It is middle that load is reduced using kernel function, can be by shortening the run time of kernel function itself so that system evaluation when Between shorten, substantially increase the evaluation and test efficiency of system.
Load restoring method provided in an embodiment of the present invention, by being believed according to the load behavior recorded in public cloud daily record Breath, the first kernel function set is obtained from kernel function storehouse, and weight is loaded according to the evaluation and test demand and public cloud log acquisition of user Daily record is put, so that load reduction apparatus is according to total resources utilization rate of each playback unit in each resource dimension and the The load behavior of each kernel function in one kernel function set, each playback unit is obtained from the first kernel function set each Second kernel function set corresponding to resource dimension, and then all playback units in daily record are reset according to load and tieed up in each resource The second kernel function set on degree, obtains the mixed load for evaluating system performance.Method provided in an embodiment of the present invention, When carrying out load reduction, it is contemplated that the load restraint of multiple resource dimensions, its mixed load accuracy restored is high, and Load is reduced using kernel function in the embodiment of the present application, can be by shortening the run time of kernel function itself so that The time of system evaluation shortens, and substantially increases the evaluation and test efficiency of system.
Fig. 2 is the schematic flow sheet of load restoring method embodiment two provided in an embodiment of the present invention.The present embodiment is related to Be by the second kernel function set obtain with kernel function sequence, and according to the kernel function sequence carry out load reduction specific mistake Journey.On the basis of above-described embodiment, above-mentioned S104 can specifically include：
S201：The second kernel function set determines every according to corresponding to each playback unit in each resource dimension Kernel function sequence of the individual playback unit in each resource dimension.
Specifically, when load reduction apparatus obtains each playback unit in the second kernel function collection corresponding to each resource dimension After conjunction, the kernel function sequence in the resource dimension can be determined according to the second kernel function set.Equally with above-mentioned playback Exemplified by unit A, other playback units perform with reference to playback unit A process, specifically：
Corresponding three the second kernel function set of above-mentioned playback unit A, it is the second kernel function set of cpu resource dimension respectively A, the second kernel function set B of memory source dimension and the second kernel function set C of I/O resource dimension, second function set A Including 3 CPU core functions 1 (being set to c1), 2 CPU core functions 2 (being set to c2) and 3 CPU core functions 3 (being set to c3), the second core Function set B includes 2 internal memory kernel functions 1 (being set to m1) and 2 internal memory kernel functions 2 (being set to m1), the second kernel function set C bags Include 4 IO kernel functions 1 (being set to x1) and 2 IO kernel functions 2 (being set to x2).Loading reduction apparatus can be according to certain sequence group Determine that kernel function sequence corresponding to the second kernel function set A is (c1, c1, c1, c2, c2, c3, c3, c3) into rule, it is optional to go back The sequence of other sequences can be formed, the embodiment of the present invention is not limited to this；Similarly, load reduction apparatus determines the second core Kernel function sequence corresponding to function set B is (m1, m1, m2, m2), determines kernel function sequence corresponding to the second kernel function set C For (x1, x1, x1, x1, x2, x2).
After the same method, load reduction apparatus can obtain core letter of each playback unit in each resource dimension Number Sequence.
S202：According to target architecture characteristic value of each playback unit in each resource dimension, from Optimal kernel function combination of each playback unit in each resource dimension is determined in the kernel function sequence.
Wherein, all kernel functions in the optimal kernel function combination are special in architecture caused by the system operation The error of value indicative and the target architecture characteristic value is minimum.
Specifically, in the embodiment of the present invention, each playback unit also includes all loads difference in current playback unit Target architecture characteristic value in each resource dimension, by taking above-mentioned retransmission unit A as an example, except bag in playback unit A Include all cpu resource overall availabilities being supported in cpu resource dimension in current playback unit, in internal memory resource dimension Memory source overall availability, the I/O resource overall availability in I/O resource dimension, further comprises the target in cpu resource dimension Architectural feature value (for example, target CPI values), target architecture characteristic value (such as the target in internal memory resource dimension MAI values), the target architecture characteristic value (such as target IOPS values) in I/O resource dimension.
By taking above-mentioned playback unit A, cpu resource dimension as an example, kernel function sequence corresponding to the second kernel function set A for (c1, C1, c1, c2, c2, c3, c3, c3), its corresponding kernel function combination can have 28Individual, load reduction apparatus can be from 28Individual core letter Array chooses an optimal kernel function combination (the optimal kernel function combination can include at least one kernel function) in closing, as long as really The optimal kernel function combination is protected in the architectural feature value (CPI) of cpu resource dimension caused by system operation with being provided in CPU The error minimum of the target architecture characteristic value of source dimension.
Optionally, can be that each kernel function combination of estimation operation one by one is looked into when choosing above-mentioned optimal kernel function combination It is seen in CPI values caused by cpu resource dimension, then selects one and the target architecture characteristic value in cpu resource dimension The minimum kernel function combination of error.
Optionally, load reduction apparatus can also obtain optimal kernel function combination using the method for Dynamic Programming, specifically The flow chart shown in Fig. 3 is may refer to, on how by the method for Dynamic Programming to obtain playback unit A in cpu resource dimension On optimal kernel function combination, be referred to the introduction about Fig. 3 in following embodiments, will not be repeated here.Advised by dynamic The method drawn obtains optimal kernel function and combined, and can greatly shorten optimal kernel function combination and obtain the time, so as to which effectively lifting is The efficiency of system evaluation and test.
According to above-mentioned introduced method, load reduction apparatus can be obtained by playback unit A in internal memory resource dimension The combination of optimal kernel function and the optimal kernel function combination in I/O resource dimension, also, can also obtain it is each reset it is single Optimal kernel function combination of the member in each resource dimension.
S203：Optimal kernel function of all playback units in each resource dimension in daily record is reset according to the load Combination, obtains the mixed load for evaluating system performance.
Specifically, load reduction apparatus can combine above-mentioned obtained each playback unit in each optimal kernel function As each playback scripts, then these playback scripts are combined, load reduction is carried out based on the playback script after combination, So that the architectural feature value of the mixed load restored and the target architecture in respective resources dimension are special The error of value indicative is minimum, substantially increases the reduction accuracy rate of mixed load.
Load restoring method provided in an embodiment of the present invention, by corresponding in each resource dimension according to each playback unit The second kernel function set determine kernel function sequence of each playback unit in each resource dimension, and according to it is each reset it is single Target architecture characteristic value of the member in each resource dimension, determines each playback unit in each money from kernel function sequence Optimal kernel function combination in the dimension of source, and then all playback units in load playback daily record are in each resource dimension Optimal kernel function combination, obtain the mixed load for evaluating system performance.By being taken into full account when carrying out load reduction The architectural feature of load, select and reset the target architecture characteristic value error recorded in daily record minimum most with load The combination of excellent kernel function so that the mixed load restored run in system to be evaluated caused by architecture spy The error of value indicative and the target architecture characteristic value in respective resources dimension is minimum, and the reduction for substantially increasing mixed load is accurate True rate.
Fig. 4 is the schematic flow sheet of load restoring method embodiment three provided in an embodiment of the present invention.The present embodiment is related to Be load reduction apparatus according to the load behavioural information recorded in public cloud daily record, the first kernel function is obtained from kernel function storehouse The detailed process of set.On the basis of above-described embodiment, further, above-mentioned S101 can specifically include：
S301：Load behavioural information according to being recorded in public cloud daily record determines a load behavior scope.
Specifically, load reduction apparatus can parse the loaded load behavioural information of institute recorded in public cloud daily record, It is then based on each resource dimension and determines load behavior scope corresponding to each resource dimension.Optionally, with cpu resource dimension Exemplified by, the load behavior scope can be [P, Q], wherein, P can be minimum cpu resource utilization rate or minimum CPI, Q Can be the CPI of corresponding maximum cpu resource utilization rate or maximum, certainly, P and A can also be other value ranges, only P is wanted to be less than Q.
S302：The load behavior of every kind of kernel function in the kernel function storehouse and the load behavior scope, from The kernel function for meeting the load behavior scope is obtained in the kernel function storehouse, forms the first kernel function set.
It is CPU core function set respectively, interior specifically, the kernel function set of three types can be included in kernel function storehouse Kernel function set and IO kernel function set are deposited, various kernel function, such as CPU core are may each comprise below the kernel function each gathered Different types of CPU core function can be included under function set (such as including CPU core function 1, CPU core function 2 ... CPU core Function n), different types of internal memory kernel function can be included under internal memory kernel function set (such as including internal memory kernel function 1, internal memory Different types of IO kernel functions can be included under kernel function 2 ... internal memory kernel function n), IO kernel function set (such as including IO cores Function 1, IO kernel functions 2 ... IO kernel functions n).Load reduction apparatus needs to analyze the load of every kind of kernel function in kernel function storehouse Behavior, that is, analyze the cpu resource utilization rate of every kind of kernel function, memory source utilization rate, I/O resource utilization rate, architectural feature CPI, MAI, IOPS etc., reduction apparatus load behavior scope based on determined by above-mentioned is then loaded, which core letter then judged Several load behaviors is located in the range of the load behavior, and then these are met into the kernel function of load behavior scope screens, Form the first kernel function set.
Load restoring method provided in an embodiment of the present invention, by according to the load behavioural information recorded in public cloud daily record Determine a load behavior scope, and the load behavior of every kind of kernel function in kernel function storehouse and the load behavior model Enclose, the kernel function for meeting the load behavior scope is obtained from kernel function storehouse, the first kernel function set is formed, so as to ensure first Kernel function in kernel function set has certain typicalness, ensures subsequently it is determined that optimal kernel function in each resource dimension During combination, the first kernel function set can provide more preferable foundation, ensure that the validity and accuracy of load reduction.
Fig. 5 is the schematic flow sheet of load restoring method example IV provided in an embodiment of the present invention.The present embodiment is related to Be to load evaluation and test demand and the public cloud daily record of the reduction apparatus according to user, obtain the specific mistake that daily record is reset in load Journey.On the basis of above-described embodiment, further, above-mentioned S102 specifically may include steps of：
S401：Daily record fragment, the day master chip are chosen from the public cloud daily record according to the evaluation and test demand of the user Section includes multiple first playback units, and first playback unit is each including being each supported in current first playback unit Resource utilization in resource dimension.
Specifically, the evaluation and test demand of user is wanted to obtain that daily record is reset in what kind of load for characterizing user, i.e., The evaluation and test demand of user can include load and reset in the machine type that should be recorded in daily record, load playback daily record and should record The machine type corresponding to machine number and load reset daily record period in it is at least one.Then, load Reduction apparatus chooses the daily record fragment for meeting the evaluation and test demand, this day master chip according to the evaluation and test demand of user from public cloud daily record Section includes multiple first playback units, and each first playback unit includes each being supported on each money in current first playback unit Resource utilization in the dimension of source.
S402：The resource being each supported in each resource dimension in each first playback unit is used Rate carries out sum operation, obtains total resources utilization rate of each first playback unit in each resource dimension.
Specifically, after load reduction apparatus acquisition meets the daily record fragment of evaluation and test demand of user, due to day master chip What each first playback unit of section included is each be supported in current first playback unit in each resource dimension Resource utilization, therefore, load reduction apparatus carry out the resource utilization in same resource dimension in first playback unit It is cumulative, and then obtain all total resources utilization rates being supported in the first playback unit in each resource dimension.For example, it is assumed that Daily record fragment includes 100 the first playback units, includes 3 loads, this 3 loads in one of them first playback unit Cpu resource utilization rate in first retransmission unit is respectively 20%, 15%, 10%, in first retransmission unit in It is respectively 30%, 25%, 20% to deposit resource utilization；Therefore, all CPU of these three loads can be provided by loading reduction apparatus Source utilization rate is overlapped, and the memory source utilization rate of these three loads is overlapped, and finally obtains the first playback unit Cpu resource overall availability in cpu resource dimension is 45%, and the memory source overall availability in internal memory resource dimension is 75%.
S403：To the architecture being each supported in each resource dimension in each first playback unit Characteristic value carries out mean value computation, obtains target architecture of each first playback unit in each resource dimension Characteristic value.
Specifically, continuing according to the example above, loading reduction apparatus can be by being each supported in the first playback unit Architectural feature value in same resource dimension carries out mean value computation, so as to obtain first playback unit in the resource dimension On target architecture characteristic value.Optionally, mean value computation here can be arithmetic mean, weighted average, can also be Other modes are averaged, and the embodiment of the present invention is not limited this.
According to the method described above, this operation is performed both by each first playback unit, obtains each first playback unit every Target architecture characteristic value in individual resource dimension.
Optionally, there is no the restriction of sequential relationship between above-mentioned S402 and S403, the two can also be carried out simultaneously, S403 It can be carried out before S402.
S404：According to total resources utilization rate of each first playback unit in each resource dimension, each Target architecture characteristic value of first playback unit in each resource dimension, obtain the load and reset day Will.
Specifically, load reduction apparatus is obtaining total resources use of each first playback unit in each resource dimension After the target architecture characteristic value of rate, each first playback unit in each resource dimension, same first is reset Total resources utilization rate of the unit in each resource dimension, the target architecture characteristic value in each resource dimension are as negative The content of a playback unit in daily record is put in load-carrying, by that analogy, obtains multiple playback units that daily record is reset in load, and then Obtain load and reset daily record.
Load restoring method provided in an embodiment of the present invention, by the evaluation and test demand according to the user from public cloud daily record Middle selection daily record fragment, and the content of each first playback unit in the daily record fragment is changed, obtain load and reset The content of playback unit in daily record, and then obtain load and reset daily record, so that load reduction apparatus load also When former, it can be reduced based on fine-grained playback unit so that the precision of the load restored is higher.
For the technical scheme of the more specific description embodiment of the present invention, following embodiments go out from playback unit A angle Send the overall process for introducing mixed load reduction.Wherein, the resource dimension of the load involved by playback unit A provides including CPU Source dimension, memory source dimension, I/O resource dimension, the first kernel function set include CPU core function set, internal memory kernel function collection Close and IO kernel function set.Referring specifically to following embodiments：
Fig. 6 is the schematic flow sheet of load restoring method embodiment five provided in an embodiment of the present invention.The present embodiment is related to Be based on playback unit A obtain playback unit A in each resource dimension optimal kernel function combination detailed process.Such as figure Shown in 6, this method includes：
S501：Load behavioural information according to being recorded in public cloud daily record determines a load behavior scope.
S502：The load behavior of every kind of kernel function in the kernel function storehouse and the load behavior scope, from The kernel function for meeting the load behavior scope is obtained in the kernel function storehouse, forms the first kernel function set.
S503：Daily record fragment, the day master chip are chosen from the public cloud daily record according to the evaluation and test demand of the user Section includes multiple first playback units.
First playback unit includes the resource being each supported in current first playback unit in each resource dimension Utilization rate.
S504：The resource being each supported in each resource dimension in each first playback unit is used Rate carries out sum operation, obtains total resources utilization rate of each first playback unit in each resource dimension.
S505：To the architecture being each supported in each resource dimension in each first playback unit Characteristic value carries out mean value computation, obtains target architecture of each first playback unit in each resource dimension Characteristic value.
S506：According to total resources utilization rate of each first playback unit in each resource dimension, each Target architecture characteristic value of first playback unit in each resource dimension, obtain the load and reset day Will.
Specifically, above-mentioned S501 to S506 detailed process may refer to the description of above-described embodiment three and example IV, It will not be repeated here.
Based on this, load reduction apparatus obtains load and resets daily record, and each playback unit that the load is reset in daily record is equal Including the total resources utilization rate and current playback list in each resource dimension respectively of all loads in current playback unit All loads in member target architecture characteristic value in each resource dimension respectively.It is following using a playback unit A as Example illustrates, it is assumed that the first kernel function set includes M internal memory kernel function, P IO kernel function, Q CPU core function.
S507：According to total resources utilization rates of the playback unit A in space resources dimension, the first of playback unit A is determined Spatial content, and according to the space resources utilization rate of every kind of space kernel function in the first spatial content and space kernel function set, It is determined that the maximum of every kind of space kernel function uses number under first spatial content.
Optionally, the space resources dimension can be memory source dimension, can also be I/O resource dimension.
Specifically, in the present embodiment, space kernel function set can be internal memory kernel function set, can also be IO kernel functions Set, no matter for any kernel function set in both set, it can perform S507.Due to first in the present embodiment Kernel function set includes CPU core function set, the set of internal memory kernel function and IO kernel function set, therefore load reduction apparatus needs The set of internal memory kernel function and IO kernel function set are directed to respectively, performs the process of the S507, be specially：
Reduction apparatus is loaded according to total resources utilization rates of the playback unit A in internal memory resource dimension (i.e. in playback unit A Memory source overall availability), determine playback unit A the first spatial content.Optionally, first spatial content can be weight A kind of transition form of the memory source overall availability in unit A is put, for example, it is assumed that playback unit A is in internal memory resource dimension Total resources utilization rate be 25%, then the first spatial content can be 25, certainly, the embodiment of the present invention is in a resource dimension Total resources utilization rate and gross space capacity between transition form be not limited, as long as ensureing total resources utilization rate here Transition form between gross space capacity and the below resource utilization for every kind of kernel function and the conversion shape of spatial content Formula is consistent.
After load reduction apparatus obtains playback unit A the first spatial content, load reduction apparatus is directed to internal memory core The memory source utilization rate of every kind of internal memory kernel function also carries out identical conversion in function set, so that it is determined that holding in the first space The maximum of every kind of internal memory kernel function uses number under amount.For example, it is assumed that playback unit A the first spatial content is 25, internal memory core The memory source utilization rate of internal memory kernel function 1 in function set is that 8% (space that the internal memory kernel function accounts for after transition form is 8), the memory source utilization rate of internal memory kernel function 2 is 9% (space that the internal memory kernel function accounts for after transition form is 9), internal memory core The memory source utilization rate of function 3 is 7% (space that the internal memory kernel function accounts for after transition form is 7), then loads reduction apparatus It is determined that it is 3 that the maximum of internal memory kernel function 1, which uses number, under the first spatial content, the maximum of internal memory kernel function 2 is using number 2, it is 3 that the maximum of internal memory kernel function 3, which uses number,.
Likewise, load reduction apparatus can also obtain playback unit A the of I/O resource dimension according to similar method The maximum of every kind of IO kernel functions uses number under one spatial content, that is, loads reduction apparatus and tieed up according to playback unit A in I/O resource Total resources utilization rate on degree, first spatial contents of the playback unit A in I/O resource dimension is determined, then according to first sky Between every kind of IO kernel functions in capacity and IO kernel function set I/O resource utilization rate, it is determined that in the every kind of IO of first spatial content The maximum of kernel function uses number.
S508：Number is used according to the maximum of every kind of space kernel function under first spatial content, it is determined that resetting single First A is in the second kernel function set corresponding to the space resources dimension.
Specifically, load reduction apparatus can be by every kind of internal memory kernel function under the first spatial content of memory source dimension Maximum is added using number, and the number for obtaining internal memory kernel function in the second kernel function set under memory source dimension (is set to N1, N1 Less than M)；Optionally, loading reduction apparatus can also be by every kind of IO kernel function under the first spatial content of I/O resource dimension most It is added using number greatly, the number for obtaining IO kernel functions in the second kernel function set under I/O resource dimension (is set to N2, N2 is less than P).For example, load reduction apparatus determines playback unit A internal memory kernel functions 1 under the first spatial content of internal memory resource dimension (k1) maximum is 3 using number, and it is 2 that the maximum of internal memory kernel function 2 (k2), which uses number, the maximum of internal memory kernel function 3 (k3) The use of number is 3, then loads reduction apparatus and determine in the second kernel function set of the playback unit A under internal memory resource dimension The total number N1 for depositing kernel function is 8, the second kernel function set (being set to the second kernel function set A) bag under the memory source dimension Include 3 k1,2 k2,3 k3.
Similarly, in the method, load reduction apparatus can also obtain second cores of the playback unit A under I/O resource dimension The number sum of function set (being set to the second kernel function set B) and the IO kernel functions in the second kernel function set.
S509：The second kernel function set determines playback unit A in sky according to corresponding to playback unit A in space resources dimension Between resource dimension kernel function sequence.
Specifically, load reduction apparatus needs the second kernel function collection according to corresponding to playback unit A in internal memory resource dimension Close determine playback unit A internal memory resource dimension kernel function sequence (be set to kernel function sequence a), and, it is also desirable to according to weight Put unit A and determine kernel function sequences of the playback unit A in I/O resource dimension in the second kernel function set corresponding to I/O resource dimension (it is set to kernel function sequence b).
Optionally, continue by taking above-mentioned example as an example, the playback unit that load reduction apparatus determines is in internal memory resource dimension pair The kernel function sequence answered can be (k1, k1, k1, k2, k2, k3, k3, k3).Optionally, loading reduction apparatus can also be according to the Two kernel function set B and certain sequence composition rule determine kernel function sequence bs of the playback unit A in I/O resource dimension, this In just differ and one have been illustrated.
S510：According to first object architectural feature values of the playback unit A in space resources dimension, advised using dynamic The method drawn determines playback unit A in space resources dimension from playback unit A in the kernel function sequence of space resources dimension Optimal kernel function combination.
Specifically, load reduction apparatus needs the first object system knot according to playback unit A in internal memory resource dimension Structure characteristic value Vtarget1, using Dynamic Programming method from playback unit A in the kernel function sequence a of internal memory resource dimension really Determine optimal kernel function combinations (be set to optimal kernel function combination X) of the playback unit A in internal memory resource dimension；It is also required to according to weight First object architectural feature value Vtarget2s of the unit A in I/O resource dimension is put, using the method for Dynamic Programming from again Put unit A and optimal kernel function groups of the playback unit A in internal memory resource dimension is determined in the kernel function sequence b of I/O resource dimension Close and (be set to optimal kernel function combination Y).
Used with loading reduction apparatus in kernel function sequence a corresponding to the method from memory source dimension of Dynamic Programming, really Exemplified by determining optimal kernel function combination X of the playback unit A in internal memory resource dimension, the schematic flow sheet with reference to shown in Fig. 3：
Let it be assumed, for the purpose of illustration, that load behavior scope of the load reduction apparatus in above-mentioned S501, and in S502 Analyze the load behavior of each kernel function in obtained kernel function storehouse, it is determined that the first kernel function set include three internal memory core letters Number kmem1, kmem2 and kmem3 (internal memory kernel function often has smaller cpu resource utilization rate and CPI values) and three CPU Kernel function kCPU1, kCPU2And kCPU3, CPU core function often has bigger CPU usage and CPI values, while does not have memory source Utilization rate and MAI memory access.The load behavior of three internal memory kernel functions and three CPU core functions may refer to shown in table 1：
Table 1
Then, reduction apparatus is loaded to obtain load according to the evaluation and test demand of user and reset daily record, it is assumed that load is reset The content of playback unit A in daily record is shown in table 2：
Table 2
With reference to Tables 1 and 2, the first spatial content of playback unit A that load reduction apparatus determines in internal memory resource dimension For 25, and it is 9 that the space that k1 is accounted in the first kernel function set, which is the space that 8, k2 is accounted for, and the space that k3 is accounted for is 7, and therefore, load is also Original equipment determines that playback unit A includes 3 k1,2 k2 and 3 k3 in the second kernel function set A of internal memory resource dimension, i.e., Second kernel function set A total number is 8, optionally, the core letter that load reduction apparatus determines according to the second kernel function set Number sequence a is (k1, k1, k1, k2, k2, k3, k3, k3).
Based on this, in Fig. 3, kiFor i-th kernel functions of the playback unit A in the kernel function sequence of internal memory resource dimension, N1 is the total number (N1 8) of internal memory kernel functions of the playback unit A in the kernel function sequence a of internal memory resource dimension, and C1 attaches most importance to First spatial contents (25) of the unit A under internal memory resource dimension is put, Vtarget1 is playback unit A in internal memory resource dimension First object architectural feature value (0.034), W [ki] it is internal memory kernel function kiThe space size of occupancy, V [ki] it is internal memory Kernel function kiIn playback unit A architectural feature value, V [i, x] refers to capacity when being x, preceding i internal memory kernel function composition The combination of optimal kernel function architectural feature caused by system operation and target architecture characteristic value error.
Fig. 3 detailed process is as follows：
Load reduction apparatus from kernel function sequence when determining optimal kernel function combination, since capacity x=1, progressively Judge capacity x from 1 rise to 25 when, the situation of which internal memory kernel function should be put into the combination of optimal kernel function.
S601：As x=1, V [0, x] is capacity when being 1, and the optimal kernel function combination of preceding 0 internal memory kernel function composition exists Architectural feature value caused by system operation is equal with first object architectural feature value Vtarget1 error, the mistake Difference is actually equal to Vtarget1 value.
S602：When x is less than C1, S603 is performed；When x is more than or equal to C1, x is incremented by+1, returns and performs S601.
S603：Judge W [ki] whether it is less than or equal to x；If it is not, perform S604；If so, perform 607.
S604：V [i, x]=V [i-1, x] is made, continues executing with S605.
Specifically, being directed to the 1st internal memory kernel function k1, the space (8) that internal memory kernel function k1 takes is more than 1, then illustrates V [1, x]=V [1-1, x]=V [0, x], i.e., preceding 1 internal memory kernel function architectural feature caused by system operation with The error of Vtarget1 error and preceding 0 internal memory kernel function in architectural feature caused by system operation and Vtarget1 Equally, because will not now be added to the 1st internal memory kernel function in optimal kernel function combination.
S605：Continue to judge whether current x is less than C1, if so, S606 is then performed, if it is not, then performing S608.
S606：By x plus 1, return and perform S603.
Specifically, continue to continue above-mentioned x=1 process, when judging that current x=1 is still less than C1 (25) again, then by x+ 1, return and perform circulation S603, when x=8, i is still 1, then judges the space that the 1st internal memory kernel function takes again (8) it is equal to the value of current x (current x value is 8), it is determined that be put into the 1st internal memory kernel function in optimal kernel function combination (i.e. It is determined that first internal memory kernel function is reset), and S607 is performed, to determine the 1st internal memory kernel function being put into optimal core During combination of function and Vtarget1 error.
S607：Determined according to formula V [i, x]=min { V [i-1, x], E (V [i], V [i-1, x-W [i]], Vtarget) } When preceding i internal memory kernel function is put into the combination of optimal kernel function and Vtarget1 error, continue executing with S605.
Wherein, V [i-1, x] is referred to when i-th of kernel function is not reset in the combination of optimal kernel function and Vtarget1 mistake Difference (combination of optimal kernel function and Vtarget1 error i.e. comprising preceding i-1 kernel function), E (V [i], V [i-1, x-W [i]], Vtarget1) refer to after putting i-th of internal memory kernel function into optimal kernel function combination, remaining spatial content i- before being put into During 1 internal memory kernel function and Vtarget1 error.
S608：Judge whether current i is less than N1；If so, then perform S609；If it is not, then perform S610.
S609：Reset by i+1, and by current x, return and perform S603.
S610：Export optimal kernel function combination and corresponding error.
According to the schematic flow sheet of the Dynamic Programming shown in above-mentioned Fig. 3, it is possible to obtain playback unit A and tieed up in memory source The optimal kernel function combination of degree.With continued reference to above-mentioned example, the optimal kernel function obtained here using dynamic programming method is combined For (k1, k2, k3), the optimal kernel function is combined to the playback script as playback unit A in internal memory resource dimension.
Combined for obtaining playback unit A using dynamic programming method in the optimal kernel function of I/O resource dimension, Ke Yican Carry out, will not be repeated here according to the process shown in above-mentioned Fig. 3.
It should be noted that playback unit A is obtained in internal memory resource dimension or in I/O resource using dynamic programming method The optimal kernel function combination of dimension, can greatly shorten the acquisition time of optimal kernel function combination, and then substantially increase system Evaluation and test efficiency.
S511：It is corresponding in cpu resource dimension to obtain optimal kernel function combinations of the playback unit A in space resources dimension Load behavioural information.
Wherein, the space resources dimension includes memory source dimension and/or I/O resource dimension, and the playback unit is in institute Stating the corresponding load behavioural information in cpu resource dimension of the combination of the optimal kernel function in space resources dimension includes the optimal core The CPU spatial contents and caused load estimated performance CPI value that combination of function need to take when being run in system.
Specifically, because internal memory kernel function or IO kernel functions can take certain CPU spatial contents in system operation And certain CPI can be produced, therefore, when load reduction apparatus determines that playback unit A is optimal in internal memory resource dimension , it is necessary to obtain CPU spatial contents and the production that the optimal kernel function combination X need to take when being run in system after kernel function combination X Raw CPI；Also, after load reduction apparatus determines optimal kernel function combination Y of the playback unit A in I/O resource dimension, Need to obtain CPU spatial contents and the caused CPI that the optimal kernel function combination Y need to take when running in system.
S512：Need to take when the optimal kernel function in space resources dimension is combined in system operation according to playback unit A Total resources utilization rate in cpu resource dimension of CPU spatial contents and playback unit A, determine playback unit A second space Capacity.
Specifically, load reduction apparatus is first converted to total resources utilization rates of the playback unit A in cpu resource dimension Total CPU spatial contents, it is optimal in space resources dimension then to subtract playback unit A from total CPU spatial contents The CPU spatial contents that kernel function combination need to take in system operation, i.e., subtract optimal core letter from total CPU spatial contents What the CPU spatial contents and optimal kernel function combination Y that array conjunction X need to take in system operation need to take in system operation CPU spatial contents, so as to obtain playback unit A second space capacity.
S513：Used according to the cpu resource of every kind of CPU core function in second space capacity and the CPU core function set Rate, it is determined that the maximum of every kind of CPU core function uses number under the second space capacity.
After load reduction apparatus obtains playback unit A second space capacity, load reduction apparatus is directed to CPU core letter The cpu resource utilization rate of every kind of CPU core function also carries out the conversion of spatial content during manifold is closed, so that it is determined that holding in second space The maximum of every kind of CPU core function uses number under amount.For example, it is assumed that playback unit A the first spatial content is 30, CPU core letter Manifold close in CPU core function 1 cpu resource utilization rate be 6% (space that the CPU core function accounts for after transition form is 6), The cpu resource utilization rate of CPU core function 2 is 10% (space that the CPU core function accounts for after transition form is 10), CPU core function 3 Cpu resource utilization rate be 15% (space that the CPU core function accounts for after transition form is 15), then load reduction apparatus determine exist The maximum of CPU core function 1 is 5 using number under second space capacity, and it is 3 that the maximum of CPU core function 2, which uses number, CPU core letter The maximum of number 3 is 2 using number.
S514：Number is used according to the maximum of every kind of CPU core function under the second space capacity, determines the playback Unit is in the second kernel function set corresponding to the cpu resource dimension.
Specifically, load reduction apparatus can be by every kind of CPU core function under the second space capacity of cpu resource dimension most It is added using number greatly, obtains the number (being set to N2) of CPU core function in the second kernel function set under cpu resource dimension, is continued By taking above-mentioned example as an example, load reduction apparatus determines playback unit A CPU core letters under the first spatial content of cpu resource dimension The maximum of number 1 (k1) is 5 using number, and it is 3 that the maximum of CPU core function 2 (k2), which uses number, the maximum of CPU core function 3 (k3) The use of number is 2, then load reduction apparatus determines the CPU in the second kernel function set of the playback unit A under cpu resource dimension The total number N2 of kernel function is 10, and the second kernel function set (being set to the second kernel function set C) under the cpu resource dimension includes 5 k1,3 k2,2 k3.
S515：According to caused by optimal kernel function combinations of the playback unit A in space resources dimension in system operation Second target architecture characteristic value of the CPI value and playback unit A in cpu resource dimension, determines that the playback unit exists The 3rd target architecture characteristic value in cpu resource dimension.
Specifically, because internal memory kernel function or IO kernel functions can not only take certain CPU spaces in system operation Capacity, certain CPI can be also produced, therefore, above-mentioned optimal kernel function can be combined X in system operation by load reduction apparatus Caused CPI value, above-mentioned optimal kernel function combination Y CPI caused by system operation value and playback unit A institute in itself Second target architecture characteristic values of the playback unit A of record in cpu resource dimension carries out mean value computation, obtains resetting list Threeth target architecture characteristic values of first A in cpu resource dimension, the 3rd target architecture characteristic value consider load Actual motion logic, itself be an accurate value, it is single that it may insure that the follow-up method using Dynamic Programming obtains resetting The degree of accuracy of optimal kernel function combinations of first A in cpu resource dimension.
S516：According to threeth target architecture characteristic values of the playback unit A in cpu resource dimension, advised using dynamic The method drawn determines the playback unit in cpu resource dimension from playback unit A in the kernel function sequence of cpu resource dimension Optimal kernel function combination.
Specifically, the process is specifically referred to the flow chart shown in Fig. 3, it is only necessary to which Vtarget1 therein is replaced with into Three target architecture characteristic values, and C1 is replaced with into second space capacity and internal memory kernel function is replaced with into CPU core function , it is possible to the optimal kernel function for obtaining playback unit A in cpu resource dimension combines T.
Based on this, load reduction apparatus just obtained playback unit A internal memory resource dimension optimal kernel function combination X, I/O resource dimension optimal kernel function combine Y and cpu resource dimension optimal kernel function combine T, back loading reduction Optimal kernel function is combined X and resets script as one by equipment, and optimal kernel function is combined into Y as another playback script, and will Optimal kernel function combines T as another playback script.
S517：Each playback unit in daily record is reset to load and performs S507-516 respectively, obtains each playback unit Optimal kernel function combination in each resource dimension, and the optimal core letter according to each playback unit in each resource dimension Array is closed, and obtains the mixed load for evaluating system.
Specifically, load reduction apparatus can be based on every according to above-mentioned S507 to S516 method to each playback unit Individual resource dimension is handled, and obtains optimal kernel function combination of each playback unit in each resource dimension, and then load Reduction apparatus has just obtained playback script of each playback unit in each resource dimension.Then, load reduction apparatus is based on These reset script, it is possible to load is reduced, obtains the mixed load for evaluating system.Using the embodiment of the present invention The method of offer, public cloud mixed load can be accurately reduced, further, it is possible to during by the operation that adjusts kernel function Between, quickly generating and evaluating and testing to mixed load can be supported, for example, by the way that the kernel function execution time is adjusted into original ten / mono-, can be shorten to original 1/10th, while it is further contemplated that multiple resource dimensions, ensure mixed load the evaluation and test time The high precision of reduction.
Load restoring method provided in an embodiment of the present invention, by considering the information of public cloud tenant, in associative function storehouse Kernel function take out different types of operation (or load), from the distribution situation of multiple resource dimensions analysis load, and tie The playback unit in load playback daily record is closed, the optimal kernel function that different resource dimension is obtained for each playback unit combines, Load reduction is carried out based on the optimal kernel function combination, the mixed load restored can reappear the true field of data center Scape, precision is higher, can quantify and evaluation and test foundation with a high credibility so as to provide neck for evaluating system；Meanwhile using dynamic The method of planning obtains each playback unit and combined in the optimal kernel function of each resource dimension, substantially increases optimal kernel function The acquisition efficiency of combination, and can also be by adjusting the run time of kernel function, to shorten the evaluation and test time of system, certain journey The evaluation and test efficiency of systematic function is improved on degree.
One of ordinary skill in the art will appreciate that：Realizing all or part of step of above-mentioned each method embodiment can lead to The related hardware of programmed instruction is crossed to complete.Foregoing program can be stored in a computer read/write memory medium.The journey Sequence upon execution, execution the step of including above-mentioned each method embodiment；And foregoing storage medium includes：ROM, RAM, magnetic disc or Person's CD etc. is various can be with the medium of store program codes.
Fig. 7 is the structural representation of load reduction apparatus embodiment one provided in an embodiment of the present invention, and the device can lead to Software realization is crossed, can also be realized by way of hardware or software and hardware combining.As shown in fig. 7, the device can include：Core Functional Analysis acquisition module 10, reset log acquisition module 11, kernel function matching module 12 and load recovery module 13.
Specifically, kernel function analyzes acquisition module 10, for according to the load behavioural information recorded in public cloud daily record, from The first kernel function set is obtained in kernel function storehouse；Wherein, the first kernel function set includes CPU core function set and space core Function set；The CPU core function set includes multiple CPU core functions, and the space kernel function set includes internal memory kernel function Set and/or IO kernel function set, the internal memory kernel function set include multiple internal memory kernel functions, the IO kernel functions set bag Include multiple IO kernel functions；
Log acquisition module 11 is reset, for the evaluation and test demand according to user and the public cloud daily record, obtains load weight Put daily record；Wherein, each playback unit that the load is reset in daily record includes all loads point in current playback unit Total resources utilization rate not in each resource dimension；
Kernel function matching module 12, for being used according to total resources of each playback unit in each resource dimension The load behavior of rate and each kernel function in the first kernel function set, obtained from the first kernel function set each The playback unit is in the second kernel function set corresponding to each resource dimension；Wherein, in the second kernel function set The total resources that are no more than in the resource dimension of resource utilization sum of the same kernel function in the resource dimension make With rate；
Recovery module 13 is loaded, for resetting all playback units in daily record according to the load in each resource dimension On the second kernel function set, obtain the mixed load for evaluating system performance.
Load reduction apparatus provided in an embodiment of the present invention, can perform above method embodiment, its realization principle and skill Art effect is similar, will not be repeated here.
Fig. 8 is the structural representation of load reduction apparatus embodiment two provided in an embodiment of the present invention, and the device can lead to Software realization is crossed, can also be realized by way of hardware or software and hardware combining.On the basis of above-mentioned embodiment illustrated in fig. 7 On, all loads in current playback unit are also included in the present embodiment, in each playback unit respectively in each resource Target architecture characteristic value in dimension；As shown in figure 8, above-mentioned load recovery module 13, is specifically included：Sequence determination unit 131st, Combinatorial Optimization unit 132 and load reduction unit 133.
Specifically, sequence determination unit 131, for corresponding in each resource dimension according to each playback unit The second kernel function set determine each kernel function sequence of the playback unit in each resource dimension；
Combinatorial Optimization unit 132, for according to each objective body of the playback unit in each resource dimension Architecture characteristic value, determine that each playback unit is optimal in each resource dimension from the kernel function sequence Kernel function combines；Wherein, all kernel functions in the optimal kernel function combination are in system knot caused by the system operation The error of structure characteristic value and the target architecture characteristic value in the resource dimension is minimum；
Reduction unit 133 is loaded, is tieed up for resetting all playback units in daily record according to the load in each resource Optimal kernel function combination on degree, obtains the mixed load for evaluating system performance.
Further, the Combinatorial Optimization unit 132, specifically for each playback unit of basis in each money Target architecture characteristic value in the dimension of source, determined using the method for Dynamic Programming from the kernel function sequence each described Optimal kernel function combination of the playback unit in each resource dimension.
Load reduction apparatus provided in an embodiment of the present invention, can perform above method embodiment, its realization principle and skill Art effect is similar, will not be repeated here.
Fig. 9 is the structural representation of load reduction apparatus embodiment three provided in an embodiment of the present invention, and the device can lead to Software realization is crossed, can also be realized by way of hardware or software and hardware combining.On the basis of above-mentioned embodiment illustrated in fig. 8 On, as shown in figure 9, the device can also include acquisition module 14；
The acquisition module 14, provided for obtaining each playback unit in the Combinatorial Optimization unit 132 in space After optimal kernel function combination in the dimension of source, each optimal core of the playback unit in the space resources dimension is obtained The corresponding load behavioural information in cpu resource dimension of combination of function；Wherein, the space resources dimension includes memory source dimension And/or I/O resource dimension, optimal kernel function combination of the playback unit in the space resources dimension are corresponding in cpu resource The load behavioural information of dimension includes CPU spatial contents and the production that the optimal kernel function combination need to take when being run in system Raw load estimated performance CPI value.
Further, with continued reference to shown in Fig. 9, the space kernel function collection is combined into the set of internal memory kernel function or IO kernel functions Set；The kernel function matching module 12, is specifically included：
Determining unit 121, for the total resources utilization rate according to the playback unit in space resources dimension, determine institute The first spatial content of playback unit is stated, and according to every kind of space in first spatial content and the space kernel function set The space resources utilization rate of kernel function, it is determined that the maximum of every kind of space kernel function uses number under first spatial content； And number is used according to the maximum of every kind of space kernel function under first spatial content, determine that the playback unit exists Second kernel function set corresponding to the space resources dimension, and according to the playback unit in the space resources dimension The CPU spatial contents and the playback unit that optimal kernel function combination need to take in system operation are in cpu resource dimension Total resources utilization rate, determine the second space capacity of the playback unit；And according to the second space capacity and described The cpu resource utilization rate of every kind of CPU core function in CPU core function set, it is determined that every kind of CPU core under the second space capacity The maximum of function uses number, and uses number according to the maximum of every kind of CPU core function under the second space capacity, it is determined that The playback unit is in the second kernel function set corresponding to the cpu resource dimension；
Cycling element 122, for indicating that the determining unit 121 obtains all playback lists in the load playback daily record Member is respectively in the second kernel function set corresponding to space resources dimension and cpu resource dimension.
Further, with continued reference to shown in Fig. 9, the Combinatorial Optimization unit 132, specifically include：
Dynamic optimization subelement 1321, for the first object system according to the playback unit in space resources dimension Structural Eigenvalue, institute in the kernel function sequence of space resources dimension is determined from the playback unit using the method for Dynamic Programming Optimal kernel function combination of the playback unit in space resources dimension is stated, and is tieed up according to the playback unit in the space resources Optimal kernel function on degree is combined in the value of CPI caused by system operation and the playback unit in cpu resource dimension Second target architecture characteristic value, determine threeth target architecture feature of the playback unit in cpu resource dimension Value；And the 3rd target architecture characteristic value according to the playback unit in cpu resource dimension, using Dynamic Programming Method determine the playback unit in cpu resource dimension in the kernel function sequence of cpu resource dimension from the playback unit On optimal kernel function combination；
Subelement 1322 is circulated, for indicating that the dynamic optimization unit obtains the institute in the load playback daily record respectively There is playback unit respectively in optimal kernel function combination corresponding to space resources dimension and cpu resource dimension.
Load reduction apparatus provided in an embodiment of the present invention, can perform above method embodiment, its realization principle and skill Art effect is similar, will not be repeated here.
Further, kernel function analysis acquisition module 10, specifically for according to being recorded in the public cloud daily record Load behavioural information determines load behavior scope, and load behavior and the institute of every kind of kernel function in the kernel function storehouse Load behavior scope is stated, the kernel function for meeting the load behavior scope is obtained from the kernel function storehouse, forms described first Kernel function set.
Optionally, the evaluation and test demand of the user includes recording in daily record machine type, the machine recorded in daily record The number of machine corresponding to device type and at least one in the period of daily record.
Figure 10 is the structural representation of load reduction apparatus example IV provided in an embodiment of the present invention, and the device can lead to Software realization is crossed, can also be realized by way of hardware or software and hardware combining.On the basis of above-mentioned embodiment illustrated in fig. 9 On, as shown in Figure 10, the playback log acquisition module 11, specifically include：
Daily record fragment chooses unit 111, is chosen for the evaluation and test demand according to the user from the public cloud daily record Daily record fragment, the daily record fragment include multiple first playback units, and first playback unit includes current first and reset The resource utilization being each supported in unit in each resource dimension；
Processing unit 112, for being each supported on each resource dimension in each first playback unit On resource utilization carry out sum operation, obtain total money of each first playback unit in each resource dimension Source utilization rate, and it is special to the architecture being each supported in each resource dimension in each first playback unit Value indicative carries out mean value computation, and it is special to obtain target architecture of each first playback unit in each resource dimension Value indicative；
Log acquisition unit 113 is reset, for each first playback unit of basis in each resource dimension Total resources utilization rate, each target architecture characteristic value of first playback unit in each resource dimension, Obtain the load and reset daily record.
Load reduction apparatus provided in an embodiment of the present invention, can perform above method embodiment, its realization principle and skill Art effect is similar, will not be repeated here.
Figure 11 is the structural representation of load reduction apparatus embodiment one provided in an embodiment of the present invention.As shown in figure 11, The load reduction apparatus can include memory 21, processor 22 and at least one communication bus 23.Communication bus 23 is used for real Communication connection between existing element.Memory 21 may include high-speed RAM memory, it is also possible to also including non-volatile memories NVM, for example, at least a magnetic disk storage, various programs can be stored in memory 21, for complete various processing functions with And realize the method and step of the present embodiment.The processor 22 for example can be central processing unit (Central Processing Unit, abbreviation CPU).
It should be noted that the communication bus 23 can be ISA (Industry Standard Architecture, work Industry standard architecture) bus, PCI (Peripheral Component Interconnect, external equipment interconnection) buses or EISA (Extended Industry Standard Architecture, extended industry-standard architecture) bus etc..It is described Communication bus 23 can be one or more physical circuit, and it is total that address bus, data can be divided into when being a plurality of physical circuit Line, controlling bus etc..
In embodiments of the present invention, processor 22, for according to the load behavioural information recorded in public cloud daily record, from core The first kernel function set, and the evaluation and test demand according to user and the public cloud daily record are obtained in function library, load is obtained and resets Daily record；And according to total resources utilization rate of each playback unit in each resource dimension and the first kernel function set In each kernel function load behavior, each playback unit is obtained from the first kernel function set each described Second kernel function set corresponding to resource dimension, and all playback units in the load playback daily record are in each resource The second kernel function set in dimension, obtains the mixed load for evaluating system performance；Wherein, the first kernel function set Including CPU core function set and space kernel function set；The CPU core function set includes multiple CPU core functions, the space Kernel function set includes the set of internal memory kernel function and/or IO kernel function set, and the internal memory kernel function set includes multiple internal memories Kernel function, the IO kernel functions set include multiple IO kernel functions；Each playback unit that the load is reset in daily record wraps Include all loads in current playback unit total resources utilization rate in each resource dimension respectively；The second kernel function collection Resource utilization sum of the same kernel function in the resource dimension in conjunction is no more than total money in the resource dimension Source utilization rate.
Optionally, all loads in current playback unit are also included in each playback unit respectively in each resource Target architecture characteristic value in dimension；
The processor 22, specifically for each playback unit of basis second corresponding to each resource dimension Kernel function set determines each kernel function sequence of the playback unit in each resource dimension, and according to each described Target architecture characteristic value of the playback unit in each resource dimension, determines each institute from the kernel function sequence State optimal kernel function combination of the playback unit in each resource dimension；And reset according to the load in daily record Optimal kernel function combination of all playback units in each resource dimension, obtains the mixed load for evaluating system performance； Wherein, all kernel functions in the optimal kernel function combination are in architectural feature value caused by the system operation and institute The error for stating the target architecture characteristic value in resource dimension is minimum.
Further, the processor 22, specifically for each playback unit of basis in each resource dimension On target architecture characteristic value, determined using the method for Dynamic Programming from the kernel function sequence it is each it is described reset it is single Optimal kernel function combination of the member in each resource dimension.
Optionally, the processor 22, it is additionally operable to obtaining each playback unit in space resources dimension After optimal kernel function combination, optimal kernel function combination pair of each playback unit in the space resources dimension is obtained Should be in the load behavioural information of cpu resource dimension；Wherein, the space resources dimension includes memory source dimension and/or IO is provided Source dimension, optimal kernel function combination of the playback unit in the space resources dimension are corresponding in the negative of cpu resource dimension Carrying behavioural information includes CPU spatial contents and the caused load that the optimal kernel function combination need to take when being run in system Estimated performance CPI value.
Optionally, the space kernel function collection is combined into the set of internal memory kernel function or IO kernel function set；
The processor 22, specifically for the total resources utilization rate according to the playback unit in space resources dimension, The first spatial content of the playback unit is determined, and according to every in first spatial content and the space kernel function set The space resources utilization rate of kind space kernel function, it is determined that the maximum of every kind of space kernel function uses under first spatial content Number, and number is used according to the maximum of every kind of space kernel function under first spatial content, determine the playback unit In the second kernel function set corresponding to the space resources dimension；And tieed up according to the playback unit in the space resources The CPU spatial contents and the playback unit that optimal kernel function combination on degree need to take in system operation are tieed up in cpu resource Total resources utilization rate on degree, the second space capacity of the playback unit is determined, and according to the second space capacity and institute The cpu resource utilization rate of every kind of CPU core function in CPU core function set is stated, it is determined that every kind of CPU under the second space capacity The maximum of kernel function uses number, and then uses number according to the maximum of every kind of CPU core function under the second space capacity, The playback unit is determined in the second kernel function set corresponding to the cpu resource dimension, and then obtains the load and resets day All playback units in will are respectively in the second kernel function set corresponding to space resources dimension and cpu resource dimension.
Optionally, the processor 22, specifically for the first mesh according to the playback unit in space resources dimension Mark system Structural Eigenvalue, using the method for Dynamic Programming from the playback unit in the kernel function sequence of space resources dimension Optimal kernel function combination of the playback unit in space resources dimension is determined, and according to the playback unit in the space Optimal kernel function combination in resource dimension is tieed up in the value of CPI caused by system operation and the playback unit in cpu resource The second target architecture characteristic value on degree, determine threeth objective body tying of the playback unit in cpu resource dimension Structure characteristic value, and the 3rd target architecture characteristic value according to the playback unit in cpu resource dimension, using dynamic The method of planning determines the playback unit in cpu resource from the playback unit in the kernel function sequence of cpu resource dimension Optimal kernel function combination in dimension, and then all playback units in the load playback daily record are obtained respectively in space resources Optimal kernel function combination corresponding to dimension and cpu resource dimension.
Further, the processor 22, specifically for according to the load behavioural information recorded in the public cloud daily record It is determined that load behavior scope, and the load behavior of every kind of kernel function in the kernel function storehouse and the load behavior model Enclose, the kernel function for meeting the load behavior scope is obtained from the kernel function storehouse, form the first kernel function set.
Optionally, the evaluation and test demand of the user includes recording in daily record machine type, the machine recorded in daily record The number of machine corresponding to device type and at least one in the period of daily record.
Further, the processor 22, specifically for the evaluation and test demand according to the user from the public cloud daily record Middle selection daily record fragment, and the resource being each supported in each resource dimension in each first playback unit is used Rate carries out sum operation, obtains total resources utilization rate of each first playback unit in each resource dimension；With And the architectural feature value being each supported in each resource dimension in each first playback unit is carried out Mean value computation, target architecture characteristic value of each first playback unit in each resource dimension is obtained, and Reset according to total resources utilization rate of each first playback unit in each resource dimension, each described first single Target architecture characteristic value of the member in each resource dimension, obtain the load and reset daily record；The daily record fragment Include multiple first playback units, first playback unit includes each being supported on each money in current first playback unit Resource utilization in the dimension of source.
Load reduction apparatus provided in an embodiment of the present invention, can perform above method embodiment, its realization principle and skill Art effect is similar, will not be repeated here.
Finally it should be noted that：Various embodiments above is merely illustrative of the technical solution of the present invention, rather than its limitations；To the greatest extent The present invention is described in detail with reference to foregoing embodiments for pipe, it will be understood by those within the art that：Its according to The technical scheme described in foregoing embodiments can so be modified, either which part or all technical characteristic are entered Row equivalent substitution；And these modifications or replacement, the essence of appropriate technical solution is departed from various embodiments of the present invention technology The scope of scheme.

Claims (27)

1. one kind load restoring method, it is characterised in that including：
According to the load behavioural information recorded in public cloud daily record, the first kernel function set is obtained from kernel function storehouse；Wherein, institute Stating the first kernel function set includes CPU core function set and space kernel function set；The CPU core function set includes multiple CPU core function, the space kernel function set include the set of internal memory kernel function and/or IO kernel function set, the internal memory core letter Manifold, which is closed, includes multiple internal memory kernel functions, and the IO kernel functions set includes multiple IO kernel functions；
According to the evaluation and test demand of user and the public cloud daily record, obtain load and reset daily record；Wherein, daily record is reset in the load In total resources of all loads respectively in each resource dimension that include in current playback unit of each playback unit make With rate；
According in total resources utilization rate and the first kernel function set of each playback unit in each resource dimension Each kernel function load behavior, each playback unit is obtained from the first kernel function set in each money Second kernel function set corresponding to the dimension of source；Wherein, the same kernel function in the second kernel function set is in the resource Resource utilization sum in dimension is no more than the total resources utilization rate in the resource dimension；
Second kernel function set of all playback units in each resource dimension in daily record is reset according to the load, is obtained Mixed load for evaluating system performance.
2. according to the method for claim 1, it is characterised in that also include current playback unit in each playback unit In all loads target architecture characteristic value in each resource dimension respectively；It is described that daily record is reset according to the load In second kernel function set of all playback units in each resource dimension, obtain and born for the mixing of evaluating system performance Carry, specifically include：
The second kernel function set according to corresponding to each playback unit in each resource dimension determines each described heavy Put kernel function sequence of the unit in each resource dimension；
According to target architecture characteristic value of each playback unit in each resource dimension, from the kernel function Optimal kernel function combination of each playback unit in each resource dimension is determined in sequence；Wherein, it is described optimal All kernel functions in kernel function combination are in architectural feature value caused by the system operation and the resource dimension Target architecture characteristic value error it is minimum；
Optimal kernel function of all playback units in each resource dimension reset according to the load in daily record combines, and obtains Mixed load for evaluating system performance.
3. according to the method for claim 2, it is characterised in that the basis each the playback unit in each money Target architecture characteristic value in the dimension of source, determine each playback unit each described from the kernel function sequence Optimal kernel function combination in resource dimension, is specifically included：
According to target architecture characteristic value of each playback unit in each resource dimension, using Dynamic Programming Method each optimal kernel function of the playback unit in each resource dimension is determined from the kernel function sequence Combination.
4. according to the method for claim 3, it is characterised in that tieed up when obtaining each playback unit in space resources After optimal kernel function combination on degree, methods described also includes：
It is corresponding in cpu resource dimension to obtain optimal kernel function combination of each playback unit in the space resources dimension Load behavioural information；Wherein, the space resources dimension includes memory source dimension and/or I/O resource dimension, the playback Optimal kernel function combination corresponding load behavioural information in cpu resource dimension of the unit in the space resources dimension includes institute State CPU spatial contents and the caused value for loading estimated performance CPI that optimal kernel function combination need to take when being run in system.
5. according to the method for claim 4, it is characterised in that the space kernel function collection be combined into the set of internal memory kernel function or IO kernel function set；
Each total resources utilization rate and the first kernel function collection of the playback unit in each resource dimension of the basis The load behavior of each kernel function in conjunction, each playback unit is obtained from the first kernel function set in each institute The second kernel function set corresponding to resource dimension is stated, is specifically included：
Step A：According to total resources utilization rate of the playback unit in space resources dimension, the of the playback unit is determined One spatial content, and provided according to the space of every kind of space kernel function in first spatial content and the space kernel function set Source utilization rate, it is determined that the maximum of every kind of space kernel function uses number under first spatial content；
Step B：Number is used according to the maximum of every kind of space kernel function under first spatial content, determines that the playback is single Member is in the second kernel function set corresponding to the space resources dimension；
Step C：According to the playback unit in the space resources dimension optimal kernel function combine in system operation when need Total resources utilization rate of the CPU spatial contents and the playback unit of occupancy in cpu resource dimension, determines the playback unit Second space capacity；
Step D：Used according to the cpu resource of every kind of CPU core function in the second space capacity and the CPU core function set Rate, it is determined that the maximum of every kind of CPU core function uses number under the second space capacity；
Step E：Number is used according to the maximum of every kind of CPU core function under the second space capacity, determines that the playback is single Member is in the second kernel function set corresponding to the cpu resource dimension；
Step F：Each playback unit in daily record is reset to the load and performs the step A respectively to the step E, is obtained All playback units in the load playback daily record are respectively in the second core letter corresponding to space resources dimension and cpu resource dimension Manifold is closed.
6. according to the method for claim 4, it is characterised in that the basis each the playback unit in each money Target architecture characteristic value in the dimension of source, determined using the method for Dynamic Programming from the kernel function sequence each described Optimal kernel function combination of the playback unit in each resource dimension, is specifically included：
Step G：According to first object architectural feature value of the playback unit in space resources dimension, advised using dynamic The method drawn determines the playback unit in space resources from the playback unit in the kernel function sequence of space resources dimension Optimal kernel function combination in dimension；
Step H：Produced according to the playback unit when the optimal kernel function in the space resources dimension is combined in system operation The the second target architecture characteristic value of raw CPI value and the playback unit in cpu resource dimension, determines the playback Threeth target architecture characteristic value of the unit in cpu resource dimension；
Step I：According to threeth target architecture characteristic value of the playback unit in cpu resource dimension, advised using dynamic The method drawn determines that the playback unit is tieed up in cpu resource from the playback unit in the kernel function sequence of cpu resource dimension Optimal kernel function combination on degree；
Step J：Each playback unit in daily record is reset to the load and performs the step G respectively to the step I, is obtained All playback units in the load playback daily record are respectively in optimal core letter corresponding to space resources dimension and cpu resource dimension Array is closed.
7. according to the method described in claim any one of 1-6, it is characterised in that what is recorded in the daily record according to public cloud is negative Behavioural information is carried, the first kernel function set is obtained from kernel function storehouse, is specifically included：
Load behavioural information according to being recorded in the public cloud daily record determines load behavior scope；
The load behavior of every kind of kernel function in the kernel function storehouse and the load behavior scope, from the kernel function The kernel function for meeting the load behavior scope is obtained in storehouse, forms the first kernel function set.
8. according to the method for claim 7, it is characterised in that the evaluation and test demand of the user includes the machine recorded in daily record The number of machine corresponding to the machine type recorded in device type, daily record and at least one in the period of daily record.
9. according to the method for claim 2, it is characterised in that the evaluation and test demand according to user and the public cloud day Will, obtain load and reset daily record, specifically include：
Daily record fragment is chosen from the public cloud daily record according to the evaluation and test demand of the user, the daily record fragment includes more Individual first playback unit, first playback unit include each being supported in each resource dimension in current first playback unit Resource utilization；
The resource utilization being each supported in each resource dimension in each first playback unit is asked And operation, obtain total resources utilization rate of each first playback unit in each resource dimension；
The architectural feature value being each supported in each resource dimension in each first playback unit is entered Row mean value computation, obtain target architecture characteristic value of each first playback unit in each resource dimension；
According to total resources utilization rate of each first playback unit in each resource dimension, each first weight Target architecture characteristic value of the unit in each resource dimension is put, the load is obtained and resets daily record.
10. one kind load reduction apparatus, it is characterised in that including：
Kernel function analyzes acquisition module, for according to the load behavioural information recorded in public cloud daily record, being obtained from kernel function storehouse Take the first kernel function set；Wherein, the first kernel function set includes CPU core function set and space kernel function set；Institute Stating CPU core function set includes multiple CPU core functions, and the space kernel function set includes the set of internal memory kernel function and/or IO Kernel function set, the internal memory kernel function set include multiple internal memory kernel functions, and the IO kernel functions set includes multiple IO cores Function；
Log acquisition module is reset, for the evaluation and test demand according to user and the public cloud daily record, load is obtained and resets daily record； Wherein, it is described load reset daily record in each playback unit include current playback unit in all loads respectively each Total resources utilization rate in resource dimension；
Kernel function matching module, for according to each total resources utilization rate and institute of the playback unit in each resource dimension The load behavior of each kernel function in the first kernel function set is stated, is obtained from the first kernel function set each described heavy Unit is put in the second kernel function set corresponding to each resource dimension；Wherein, it is same in the second kernel function set Resource utilization sum of the kind kernel function in the resource dimension is no more than the total resources utilization rate in the resource dimension；
Recovery module is loaded, for resetting all playback units in daily record the in each resource dimension according to the load Two kernel function set, obtain the mixed load for evaluating system performance.
11. device according to claim 10, it is characterised in that also include current playback list in each playback unit All loads in member target architecture characteristic value in each resource dimension respectively；The load recovery module, specifically Including：
Sequence determination unit, for each playback unit of basis in the second kernel function collection corresponding to each resource dimension Close and determine each kernel function sequence of the playback unit in each resource dimension；
Combinatorial Optimization unit, it is special for the target architecture according to each playback unit in each resource dimension Value indicative, each optimal kernel function group of the playback unit in each resource dimension is determined from the kernel function sequence Close；Wherein, all kernel functions in the optimal kernel function combination are in architectural feature value caused by the system operation It is minimum with the error of the target architecture characteristic value in the resource dimension；
Load reduction unit, for according to it is described load reset daily record in all playback units in each resource dimension most Excellent kernel function combination, obtains the mixed load for evaluating system performance.
12. device according to claim 11, it is characterised in that the Combinatorial Optimization unit, specifically for according to each Target architecture characteristic value of the playback unit in each resource dimension, using the method for Dynamic Programming from described Optimal kernel function combination of each playback unit in each resource dimension is determined in kernel function sequence.
13. device according to claim 12, it is characterised in that described device also includes：
Acquisition module, for obtaining each playback unit in space resources dimension most in the Combinatorial Optimization unit After excellent kernel function combination, it is corresponding to obtain optimal kernel function combination of each playback unit in the space resources dimension In the load behavioural information of cpu resource dimension；Wherein, the space resources dimension includes memory source dimension and/or I/O resource Dimension, optimal kernel function combination corresponding load in cpu resource dimension of the playback unit in the space resources dimension Behavioural information includes CPU spatial contents and the caused load meter that the optimal kernel function combination need to take when being run in system Calculate characteristic CPI value.
14. device according to claim 13, it is characterised in that the space kernel function collection is combined into internal memory kernel function set Or IO kernel function set；The kernel function matching module, is specifically included：
Determining unit, for the total resources utilization rate according to the playback unit in space resources dimension, determine the playback First spatial content of unit, and according to every kind of space kernel function in first spatial content and the space kernel function set Space resources utilization rate, it is determined that the maximum of every kind of space kernel function uses number under first spatial content；And root Number is used according to the maximum of every kind of space kernel function under first spatial content, determines the playback unit in the space Second kernel function set corresponding to resource dimension, and the optimal core letter according to the playback unit in the space resources dimension Array, which closes the total resources of the CPU spatial contents and the playback unit that need to be taken in system operation in cpu resource dimension, to be made With rate, the second space capacity of the playback unit is determined；And according to the second space capacity and the CPU core function The cpu resource utilization rate of every kind of CPU core function in set, it is determined that every kind of CPU core function is most under the second space capacity Number is used greatly, and number is used according to the maximum of every kind of CPU core function under the second space capacity, determines the playback Unit is in the second kernel function set corresponding to the cpu resource dimension；
Cycling element, for indicating that the determining unit obtains all playback units in the load playback daily record respectively in sky Between the second kernel function set corresponding to resource dimension and cpu resource dimension.
15. device according to claim 13, it is characterised in that the Combinatorial Optimization unit, specifically include：
Dynamic optimization subelement, for the first object architectural feature according to the playback unit in space resources dimension Value, the playback list is determined from the playback unit in the kernel function sequence of space resources dimension using the method for Dynamic Programming Member in space resources dimension optimal kernel function combination, and according to the playback unit in the space resources dimension most Excellent kernel function combines the second target in the value of CPI caused by system operation and the playback unit in cpu resource dimension Architectural feature value, determine threeth target architecture characteristic value of the playback unit in cpu resource dimension；And According to threeth target architecture characteristic value of the playback unit in cpu resource dimension, using Dynamic Programming method from The playback unit determines that the playback unit is optimal in cpu resource dimension in the kernel function sequence of cpu resource dimension Kernel function combines；
Subelement is circulated, for indicating that the dynamic optimization unit obtains all playback lists in the load playback daily record respectively Member is respectively in optimal kernel function combination corresponding to space resources dimension and cpu resource dimension.
16. according to the device described in claim any one of 10-15, it is characterised in that the kernel function analyzes acquisition module, tool Body is used to determine load behavior scope according to the load behavioural information recorded in the public cloud daily record, and according to the kernel function The load behavior of every kind of kernel function in storehouse and the load behavior scope, obtained from the kernel function storehouse and meet described bear The kernel function of load behavior scope, form the first kernel function set.
17. device according to claim 16, it is characterised in that
Corresponding to machine type that the evaluation and test demand of the user includes recording in daily record, the machine type recorded in daily record The number of machine and at least one in the period of daily record.
18. device according to claim 11, it is characterised in that the playback log acquisition module, specifically include：
Daily record fragment chooses unit, and day master chip is chosen from the public cloud daily record for the evaluation and test demand according to the user Section, the daily record fragment include multiple first playback units, and first playback unit is included in current first playback unit The resource utilization being each supported in each resource dimension；
Processing unit, for the resource being each supported in each resource dimension in each first playback unit Utilization rate carries out sum operation, obtains total resources of each first playback unit in each resource dimension and uses Rate, and the architectural feature value being each supported in each resource dimension in each first playback unit is entered Row mean value computation, obtain target architecture characteristic value of each first playback unit in each resource dimension；
Log acquisition unit is reset, for according to each total resources of first playback unit in each resource dimension Utilization rate, each target architecture characteristic value of first playback unit in each resource dimension, described in acquisition Daily record is reset in load.
19. one kind load reduction apparatus, it is characterised in that including：
Processor, for according to the load behavioural information recorded in public cloud daily record, the first kernel function to be obtained from kernel function storehouse Set, and the evaluation and test demand according to user and the public cloud daily record, obtain load and reset daily record；And according to each playback The load row of total resources utilization rate of the unit in each resource dimension and each kernel function in the first kernel function set To obtain each playback unit from the first kernel function set in the second core letter corresponding to each resource dimension Manifold is closed, and second kernel function collection of all playback units in each resource dimension in daily record is reset according to the load Close, obtain the mixed load for evaluating system performance；Wherein, the first kernel function set include CPU core function set and Space kernel function set；The CPU core function set includes multiple CPU core functions, and the space kernel function set includes internal memory Kernel function set and/or IO kernel function set, the internal memory kernel function set include multiple internal memory kernel functions, the IO kernel functions Set includes multiple IO kernel functions；Each playback unit that the load is reset in daily record includes the institute in current playback unit There is the load total resources utilization rate in each resource dimension respectively；Same kernel function in the second kernel function set exists Resource utilization sum in the resource dimension is no more than the total resources utilization rate in the resource dimension.
20. equipment according to claim 19, it is characterised in that also include current playback list in each playback unit All loads in member target architecture characteristic value in each resource dimension respectively；
The processor, specifically for each playback unit of basis in the second kernel function corresponding to each resource dimension Set determines each kernel function sequence of the playback unit in each resource dimension, and single according to each playback Target architecture characteristic value of the member in each resource dimension, each playback is determined from the kernel function sequence Optimal kernel function combination of the unit in each resource dimension；And reset according to the load all heavy in daily record Optimal kernel function combination of the unit in each resource dimension is put, obtains the mixed load for evaluating system performance；Wherein, institute All kernel functions in optimal kernel function combination are stated in architectural feature value caused by the system operation and the resource The error of target architecture characteristic value in dimension is minimum.
21. equipment according to claim 20, it is characterised in that the processor, specifically for according to each described heavy Target architecture characteristic value of the unit in each resource dimension is put, using the method for Dynamic Programming from the kernel function Optimal kernel function combination of each playback unit in each resource dimension is determined in sequence.
22. equipment according to claim 21, it is characterised in that the processor, be additionally operable to obtain it is each described After optimal kernel function combination of the playback unit in space resources dimension, obtain each playback unit and provided in the space The corresponding load behavioural information in cpu resource dimension of optimal kernel function combination in the dimension of source；Wherein, the space resources dimension Including memory source dimension and/or I/O resource dimension, optimal kernel function of the playback unit in the space resources dimension The corresponding load behavioural information in cpu resource dimension of combination includes the optimal kernel function combination to be needed to take when running in system CPU spatial contents and caused load estimated performance CPI value.
23. equipment according to claim 22, it is characterised in that the space kernel function collection is combined into internal memory kernel function set Or IO kernel function set；
The processor, specifically for the total resources utilization rate according to the playback unit in space resources dimension, determine institute The first spatial content of playback unit is stated, and according to every kind of space in first spatial content and the space kernel function set The space resources utilization rate of kernel function, it is determined that the maximum of every kind of space kernel function uses number under first spatial content, And number is used according to the maximum of every kind of space kernel function under first spatial content, determine the playback unit described Second kernel function set corresponding to space resources dimension；And according to the playback unit in the space resources dimension The CPU spatial contents and the playback unit that optimal kernel function combination need to take in system operation are in cpu resource dimension Total resources utilization rate, the second space capacity of the playback unit is determined, and according to the second space capacity and the CPU core The cpu resource utilization rate of every kind of CPU core function in function set, it is determined that every kind of CPU core function under the second space capacity Maximum use number, and then number is used according to the maximum of every kind of CPU core function under the second space capacity, determines institute Playback unit is stated in the second kernel function set corresponding to the cpu resource dimension, and then obtains the load and resets in daily record All playback units are respectively in the second kernel function set corresponding to space resources dimension and cpu resource dimension.
24. equipment according to claim 22, it is characterised in that
The processor, specifically for the first object architectural feature according to the playback unit in space resources dimension Value, the playback list is determined from the playback unit in the kernel function sequence of space resources dimension using the method for Dynamic Programming Member in space resources dimension optimal kernel function combination, and according to the playback unit in the space resources dimension most Excellent kernel function combines the second target in the value of CPI caused by system operation and the playback unit in cpu resource dimension Architectural feature value, determine threeth target architecture characteristic value of the playback unit in cpu resource dimension, Yi Jigen According to threeth target architecture characteristic value of the playback unit in cpu resource dimension, using the method for Dynamic Programming from institute State playback unit and optimal core of the playback unit in cpu resource dimension is determined in the kernel function sequence of cpu resource dimension Combination of function, and then all playback units in the load playback daily record are obtained respectively in space resources dimension and cpu resource Optimal kernel function combination corresponding to dimension.
25. according to the equipment described in claim any one of 19-24, it is characterised in that the processor, specifically for according to institute State the load behavioural information recorded in public cloud daily record and determine load behavior scope, and every seed nucleus in the kernel function storehouse The load behavior of function and the load behavior scope, obtained from the kernel function storehouse and meet the load behavior scope Kernel function, form the first kernel function set.
26. equipment according to claim 25, it is characterised in that the evaluation and test demand of the user includes what is recorded in daily record The number of machine corresponding to the machine type recorded in machine type, daily record and at least one in the period of daily record It is individual.
27. equipment according to claim 26, it is characterised in that
The processor, daily record fragment is chosen from the public cloud daily record specifically for the evaluation and test demand according to the user, And sum operation is carried out to the resource utilization being each supported in each resource dimension in each first playback unit, Obtain total resources utilization rate of each first playback unit in each resource dimension；And to each described first The architectural feature value being each supported in each resource dimension in playback unit carries out mean value computation, obtains each Target architecture characteristic value of first playback unit in each resource dimension, and according to each first weight Put total resources utilization rate of the unit in each resource dimension, each first playback unit is tieed up in each resource Target architecture characteristic value on degree, obtain the load and reset daily record；The daily record fragment includes multiple first and reset Unit, the resource that first playback unit includes each being supported in current first playback unit in each resource dimension use Rate.
CN201610377787.2A 2016-05-31 2016-05-31 Load reduction method, device and equipment Active CN107450968B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610377787.2A CN107450968B (en) 2016-05-31 2016-05-31 Load reduction method, device and equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610377787.2A CN107450968B (en) 2016-05-31 2016-05-31 Load reduction method, device and equipment

Publications (2)

Publication Number Publication Date
CN107450968A true CN107450968A (en) 2017-12-08
CN107450968B CN107450968B (en) 2020-09-08

Family

ID=60485968

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610377787.2A Active CN107450968B (en) 2016-05-31 2016-05-31 Load reduction method, device and equipment

Country Status (1)

Country Link
CN (1) CN107450968B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112486738A (en) * 2019-09-12 2021-03-12 阿里巴巴集团控股有限公司 Load testing method and device, electronic equipment and computer readable storage medium

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101800771A (en) * 2010-01-29 2010-08-11 北京理工大学 Copy selection method based on kernel density estimation
CN102576314A (en) * 2009-07-27 2012-07-11 先进微装置公司 Mapping processing logic having data parallel threads across processors
CN102681889A (en) * 2012-04-27 2012-09-19 电子科技大学 Scheduling method of cloud computing open platform
US20130279795A1 (en) * 2012-04-19 2013-10-24 Applied Materials Israel Ltd. Optimization of unknown defect rejection for automatic defect classification
CN104065663A (en) * 2014-07-01 2014-09-24 复旦大学 Auto-expanding/shrinking cost-optimized content distribution service method based on hybrid cloud scheduling model
CN105487930A (en) * 2015-12-01 2016-04-13 中国电子科技集团公司第二十八研究所 Task optimization scheduling method based on Hadoop
CN105487927A (en) * 2014-09-15 2016-04-13 华为技术有限公司 Resource management method and device

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102576314A (en) * 2009-07-27 2012-07-11 先进微装置公司 Mapping processing logic having data parallel threads across processors
CN101800771A (en) * 2010-01-29 2010-08-11 北京理工大学 Copy selection method based on kernel density estimation
US20130279795A1 (en) * 2012-04-19 2013-10-24 Applied Materials Israel Ltd. Optimization of unknown defect rejection for automatic defect classification
CN102681889A (en) * 2012-04-27 2012-09-19 电子科技大学 Scheduling method of cloud computing open platform
CN104065663A (en) * 2014-07-01 2014-09-24 复旦大学 Auto-expanding/shrinking cost-optimized content distribution service method based on hybrid cloud scheduling model
CN105487927A (en) * 2014-09-15 2016-04-13 华为技术有限公司 Resource management method and device
CN105487930A (en) * 2015-12-01 2016-04-13 中国电子科技集团公司第二十八研究所 Task optimization scheduling method based on Hadoop

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
XUPING GU等: "Investigation on extended black-start schemes of power system considering reasonable load restoration", 《2014 IEEE PES ASIA-PACIFIC POWER AND ENERGY ENGINEERING CONFERENCE (APPEEC)》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112486738A (en) * 2019-09-12 2021-03-12 阿里巴巴集团控股有限公司 Load testing method and device, electronic equipment and computer readable storage medium

Also Published As

Publication number Publication date
CN107450968B (en) 2020-09-08

Similar Documents

Publication Publication Date Title
CN108595157B (en) Block chain data processing method, device, equipment and storage medium
CN106294120A (en) The test method of code, equipment and computer program
CN112256720B (en) Data cost calculation method, system, computer device and storage medium
CN106940428A (en) Chip verification method, apparatus and system
US10685160B2 (en) Large cluster persistence during placement optimization of integrated circuit designs
TWI759539B (en) Computer equipment and device for determining critical value of risk control decision
CN102176200A (en) Software test case automatic generating method
CN109857984A (en) A kind of homing method and device of boiler load factor-efficacy curve
CN108335718A (en) A kind of test method and device
CN109189572A (en) A kind of resource predictor method and system, electronic equipment and storage medium
Canizares et al. An expert system for checking the correctness of memory systems using simulation and metamorphic testing
CN106201870A (en) A kind of method and device testing GPU
CN106095574A (en) Magnanimity calculates coarse granule Parallel Implementation and calculates task stochastic and dynamic distribution method
CN106415526A (en) FET processor and operation method
Macumber et al. A graphical tool for cloud-based building energy simulation
CN110008119A (en) Report test method, device, computer equipment and storage medium
CN109542785A (en) A kind of invalid bug determines method and apparatus
CN104700900B (en) A kind of detecting system and method for memory Single event upset effecf
CN107450968A (en) Load restoring method, device and equipment
CN107679766B (en) Dynamic redundant scheduling method and device for crowd-sourcing task
CN109783381A (en) A kind of test data generating method, apparatus and system
Bian et al. Concrete hyperheuristic framework for test case prioritization
CN109902001A (en) To the detection method and terminal device of uninitialized variable
Letras et al. Throughput and Memory Optimization for Parallel Implementations of Dataflow Networks Using Multi-Reader Buffers
CN106656684A (en) Grid resource reliability monitoring method and device

Legal Events

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