CN109614038A - A kind of multi-speed disk-scheduling method of diversification QoS constraint - Google Patents
A kind of multi-speed disk-scheduling method of diversification QoS constraint Download PDFInfo
- Publication number
- CN109614038A CN109614038A CN201811403707.1A CN201811403707A CN109614038A CN 109614038 A CN109614038 A CN 109614038A CN 201811403707 A CN201811403707 A CN 201811403707A CN 109614038 A CN109614038 A CN 109614038A
- Authority
- CN
- China
- Prior art keywords
- disk
- cost
- time
- data
- user
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0602—Interfaces specially adapted for storage systems specifically adapted to achieve a particular effect
- G06F3/0625—Power saving in storage systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0628—Interfaces specially adapted for storage systems making use of a particular technique
- G06F3/0629—Configuration or reconfiguration of storage systems
- G06F3/0631—Configuration or reconfiguration of storage systems by allocating resources to storage systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0668—Interfaces specially adapted for storage systems adopting a particular infrastructure
- G06F3/067—Distributed or networked storage systems, e.g. storage area networks [SAN], network attached storage [NAS]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
- G06F9/505—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering the load
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5061—Partitioning or combining of resources
- G06F9/5072—Grid computing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Human Computer Interaction (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to a kind of multi-speed disk-scheduling methods of diversification QoS constraint to execute different dispatching algorithms according to different user's qos requirements;Different dispatching algorithms respectively correspond are as follows: disk scheduling based on time priority, based on the preferential disk scheduling of cost and based on the disk-scheduling method of benefit function.The multi-speed disk-scheduling method of diversification QoS constraint provided by the invention, contain three kinds of dispatching algorithms: the dispatching algorithm TPDS based on time priority, the dispatching algorithm BFDS based on the preferential dispatching algorithm CPDS of cost and based on benefit function, the dispatching for having diversified QoS constraint, it can be under the premise of meeting the quality of service requirement of user's multiplicity, disk by dispatching different operational modes carries out data storage, the energy consumption for farthest reducing system, can meet the needs of practical application well.
Description
Technical field
The invention belongs to technical field of the computer network, and in particular to a kind of multi-speed disk scheduling of diversification QoS constraint
Method.
Background technique
Why cloud computing mode can be applied more and more, and a big chunk reason is that enterprise provides for user
While transparent calculating and storage service, the diversified qos requirement of user is met.
QoS is the abbreviation of Quality of Service, the meaning of service quality.The vocabulary of QoS derives from network transmission
System refers to that a network can provide better service ability for the network communication of formulation, be network using various basic technologies
A kind of security mechanism, be for solve network delay and obstruction the problems such as a kind of technology: when network over loading or congestion,
QoS can ensure that important service amount is not postponed or abandoned, while guarantee the efficient operation of network.QoS is usual in network system
Three kinds of services: (1) Best-Effort Service (service model of doing one's best) are provided;(2)Integrated Service
(integrated service model, abbreviation Int-Serv);(3) Differentiated Service (differentiated service, abbreviation Diff-
Serv).In recent years, quality of service requirement is widely applied in scheduling of resource field, when computing resource, data resource
When belonging to a kind of emulative resource, how the different service quality of user met with the distribution of resource by reasonably scheduling
It studies a question it is required that being that scheduling of resource field is most important.In scheduling of resource field quality of service requirement be usually be directed to it is a certain
Mission requirements Resource Scheduler (Resource Scheduler) completes the task within the time of a certain restriction.This with sound
Form between seasonable as qualifications and as user's qos requirement, the single qos requirement as on ordinary meaning.
However, the computing system that nearly 15 coming year successively occurs, virtual computing system and cloud computing system, due to face
To user type it is numerous, simultaneously because successful business model in cloud computing technology, but also different users is using providing
With different requirements when source.Therefore, diversified qos requirement is also just come into being.Diversified qos requirement can be in system
Performance, cost overhead of user etc. obtains balanced.As previously mentioned, having requirement of real-time in diversified qos requirement very
High task, it is extremely sensitive to the response time, usually using the response time as the constraint condition of QoS;There are some non real-time natures
Task, to response time and insensitive, more concerned be generation for spending using the expense of payment required for a certain resource
Valence, this generic task is usually using cost as the constraint condition of QoS;And there are also some tasks to the response time and to charge costs
Sensibility is dynamic change with task progress situation, this usual generic task can be by the restriction item to response time and expense
Part is encapsulated in the expression that qos requirement is carried out in function as parameter.Above-mentioned different types of user is expressed by different modes
Requirement to the service quality of a certain particular system, corresponding resource manager pass through scheduling according to the qos requirement of the encapsulation
Mode reasonably distributes resource, in the form of meeting the different quality of service requirement of user, referred to as diversified qos requirement tune
Degree.
In cloud computing environment, the qos requirement of user is usually diversified: some request of data are because task is urgent
(video, be broadcast live and the scientific algorithm for needing quickly to return the result etc.), can pass through the side of budget increase with regardless of expense cost
Formula exchanges quick data access for, faces such request, and cloud storage provider needs to dispatch from the angle for maximizing interests
The back end of the request be can satisfy to cope with its request.And some request of data (return by the backup of similar file, data
Shelves) etc. from the angle for saving cost overhead, response time to request of data is simultaneously insensitive, and what is be more concerned about is that data are asked
Seek the cost that itself is spent.And there are also sensitivity of some request of data to the time and the sensitivity to the response time in its visit
Ask it is not changeless in the period, need according to request responded the case where dynamic adjust.Compared to previous distribution meter
Calculation system, cloud computing are adopted according to the different demands of user (to computing resource, the different demands of storage resource and Internet resources etc.)
Take different charging modes.It can yet be regarded as a kind of success and referential by economic means scheduling and manager backstage resource
Mode.
Resource involved in distributed system is provided services to the user in a manner of fixing a price, be from grid computing propose with
Attempt to solve the problems, such as, in cloud computing era, public cloud is externally provided service in a manner of charge collection pricing and is able to landing realization.
When resource not Dan Yiqi computing capability (computing resource), the mode of storage capacity (storage resource) and transmittability (Internet resources)
When being measured, user can be according to the characteristic of itself task, in price and meter by way of submitting different service quality
Calculation ability obtains optimal balance, and passes through economic theoretical maximum whole system between storage capacity and transmittability
Profit.
The task schedule based on QoS constraint or resource allocation methods of the prior art and in relation to QoS and task schedule it
Between relationship technical solution, be mainly used for constrain computing resource scheduling, constraint storage resource scheduling aspect more examine
Consider its storage capacity, the factor of specific energy consumption etc., does not have yet when the revolving speed and disk without considering disk handle data
Consider that the case where QoS of customer requires dynamic change and utilization benefit function portray the dynamic change of the service quality of user
Change situation, the disk scheduling of the prior art considers the diversification shortcoming that QoS of customer requires, these technological deficiencies
Cause the prior art that cannot meet the needs of practical application well.
Summary of the invention
For above-mentioned problems of the prior art, it can avoid above-mentioned skill occur the purpose of the present invention is to provide one kind
The multi-speed disk-scheduling method of the diversified QoS constraint of art defect.
The present invention provides a kind of multi-speed disk-scheduling method of diversification QoS constraint, it is intended to meet the service of user's multiplicity
Under the premise of quality requirement, by dispatch the disks of different operational modes (high speed higher power consumption mode and low speed low-power mode) into
The storage of row data, farthest reduces the energy consumption of system.
In order to achieve the above-mentioned object of the invention, technical solution provided by the invention is as follows:
A kind of multi-speed disk-scheduling method of diversification QoS constraint executes different according to different user's qos requirements
Dispatching algorithm, different dispatching algorithms respectively correspond are as follows: disk scheduling based on time priority, based on the preferential magnetic of cost
Disk dispatching algorithm and disk-scheduling method based on benefit function.
Further, based on the disk scheduling of time priority under the premise of meeting user's qos requirement, to optimize
The response time of energy consumption and suboptimization user are target.
Further, the disk scheduling based on time priority includes:
To data set (DU={ du in user's request1, du2... duu) each of data dui, repeat following
Step:
By data duiPiecemeal processing is carried out, the data block data_block of equal part is formedi, determine magnetic required for the data
The number pn of diski;
Data block data_block is completed to disk all in system using response time prediction deviceiRequired for processing
Time is estimated, and obtains that each disk is corresponding to estimate response time list: RT={ rt1, rt2... rtn};
According to the Real-Time Pricing mechanism of disk each in storage system, obtain each disk uses price list: DP=
{dp1, dp2... dpn};
The cost and time that user has spent are obtained, remaining budget and available time are calculated;
According to response time list RT and disk is estimated using the information in price list DP, the user time time limit will be met
Candidate disc-pack CN is added to the node of user's budget;
To the set CN of candidate disk, ascending sort is carried out according to energy consumption size;
To the identical disk of energy consumption size in the set CN of candidate disk, ascending sort is carried out according to the response time is estimated;
To the identical disk of energy consumption size in the set CN of candidate disk and response time identical disk is estimated, according to magnetic
Disk carries out ascending sort using price;
By the pn of the candidate foremost disc-pack CNiA disk distributes to the data du in subscriber data seti;
By data duiBlock parallel is assigned to corresponding pniA disk carries out storage processing;
Calculate the cost of each piecemeal the time it takes and cost;
The time of cost is equal to pniThe cost of the longest time spent in a data block, cost are equal to pniA data block
The summation of the cost spent;
Update the cost of time and cost that user has spent;
Next data in processes user data set;
Until having handled each of data acquisition system data, finishing scheduling.
Further, based on the preferential disk scheduling of cost under the premise of meeting user's qos requirement with optimize
The cost of energy consumption and suboptimization user are target.
Further, include: based on the preferential disk scheduling of cost
To data set (DU={ du in user's request1, du2... duu) each of data dui, repeat following
Step:
By data duiPiecemeal processing is carried out, the data block data_block of equal part is formedi, determine magnetic required for the data
The number pn of diski;
Data block data_block is completed to disk all in system using response time prediction deviceiRequired for processing
Time is calculated, and obtains that disk is corresponding to estimate response time list: RT={ rt1, rt2... rtn};
According to the Real-Time Pricing mechanism of disk each in storage system, obtain each disk uses price list: DP=
{dp1, dp2... dpn};
The cost and time that user has spent are obtained, remaining budget and available time are calculated;
According to response time list RT and disk is estimated using the information in price list DP, the user time time limit will be met
Candidate disc-pack CN is added to the node of user's budget;
To candidate disc-pack CN, ascending sort is carried out according to energy consumption size;
To the identical disk of energy consumption size in candidate disc-pack CN, ascending sort is carried out using price according to disk;
To the identical disk of energy consumption size in candidate disc-pack CN and using the identical disk of price, responded according to estimating
Time carries out ascending sort;
By the pn of the candidate foremost back end set CNiA disk distributes to the data du in subscriber data seti;
By data duiBlock parallel is assigned to corresponding pniA disk carries out storage processing;
Calculate the cost of each piecemeal the time it takes and cost;
The time of cost is equal to pniThe cost of the longest time spent in a data block, cost are equal to pniA data block
The summation of the cost spent;
Update the cost of time and cost that user has spent;
Next data in processes user data set;
Until having handled each of data acquisition system data, finishing scheduling.
Further, it expresses based on the disk-scheduling method of benefit function includes: building benefit function as scheduling is calculated
The progress of method, user's situation of Profit in the case of expense and cost expense in different times, to express user's qos requirement
Dynamic changes.
Further, the step of disk-scheduling method based on benefit function includes:
1) user inputs corresponding time-based benefit function DB_Function (Time_used) and based on cost
Benefit function BB_Function (Cost_used);
2) to data set (DU={ du in user's request1, du2... duu) each of data dui, repeat down
State step:
Calculate the time Time_used that the user has currently used;
Calculate the cost Cost_used that the user has currently spent;
Saving in time costs value DV=DB_Function (Time_used) is calculated according to the value of Time_used;
Cost benefit value CV=BB_Function (Cost_used) is calculated according to the value of Cost_used;
Compare the size of DV and CV;
The dispatching algorithm TPDS based on time priority is used if DV is greater than CV;
The dispatching algorithm CPDS preferential based on cost is used if DV is less than CV;
TPDS algorithm or CPDS algorithm are randomly choosed if DV is equal to CV;
3) disk of distribution is returned.
The multi-speed disk-scheduling method of diversification QoS constraint provided by the invention, contains three kinds of dispatching algorithms: when being based on
Between preferential dispatching algorithm TPDS, the dispatching algorithm BFDS based on the preferential dispatching algorithm CPDS of cost and based on benefit function,
Have the dispatching of diversified QoS constraint, scheduling can be passed through under the premise of meeting the quality of service requirement of user's multiplicity
The disk of different operational modes carries out data storage, farthest reduces the energy consumption of system, can meet well and actually answer
Needs.
Detailed description of the invention
Fig. 1 is the principle of the present invention figure;
Fig. 2 is the form figure of benefit function of the present invention;
Fig. 3 is the distribution situation figure that the wiki-workload used in simulated experiment reaches the time;
Fig. 4 is the distribution situation figure for the file size that different task Cloudlet is requested;
Fig. 5 is response time comparison diagram of four kinds of dispatching algorithms under the configuration of different hyperdisks;
Fig. 6 is expense cost comparison diagram of four kinds of dispatching algorithms under the configuration of different hyperdisks;
Fig. 7 is energy consumption comparison figure of four kinds of dispatching algorithms under the configuration of different hyperdisks.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawing and specific implementation
The present invention will be further described for example.It should be appreciated that described herein, specific examples are only used to explain the present invention, and does not have to
It is of the invention in limiting.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creative work premise
Under every other embodiment obtained, shall fall within the protection scope of the present invention.
As shown in Figure 1, a kind of multi-speed disk-scheduling method of diversification QoS constraint, comprising: user by user interface to
Cloud computing system submits quality of service requirement, and each back end is current in node load collector real-time collecting storage system
The use price of load and current data node disk;Cloud computing system according to the information of the storage system being collected into, according to
The different qos requirement of user (time priority, cost is preferential, or is based on benefit function), is dispatched by different data in magnetic disk and is calculated
Method, will be different farthest to reduce the energy consumption of system as target under the premise of meeting user's diversified qos requirement
User's request scheduling into the disk of different operational modes;According to the difference of user's qos requirement, cloud computing system executes difference
Dispatching algorithm, respectively correspond for based on time priority disk scheduling, based on the preferential disk scheduling of cost and
Disk-scheduling method based on benefit function.The detailed description of these three algorithms is shown in algorithm 1.1 (Algorithm 1.1) respectively,
Algorithm 1.2 (Algorithm 1.2, algorithm 1.3 (Algorithm 1.3).
The symbol being related in three algorithms is shown in Table 1.
The meaning of symbol involved in 1 three algorithms of table
Disk scheduling (TPDS) based on time priority is detailed in Algorithm 1.1.
Algorithm 1.2 is detailed in based on the preferential disk scheduling of cost (CPDS).
Disk scheduling (BFDS) based on benefit function is detailed in Algorithm 1.3.
(1) disk scheduling based on time priority (TPDS)
The main thought of disk scheduling based on time priority is: under the premise of meeting user's qos requirement (including
The time limit of user and the budget of user), to optimize the response time of energy consumption and suboptimization user as target.
Basic step is as follows:
To data set (DU={ du in user's request1, du2... duu) each of data dui, repeat following
Step:
1) by data duiPiecemeal processing is carried out, the data block data_block of equal part is formedi, determine required for the data
The number pn of diski;
2) data block data_block is completed to disk all in system using response time prediction deviceiRequired for processing
Time estimated, obtain that each disk is corresponding to estimate response time list: RT={ rt1, rt2... rtn};
3) according to the Real-Time Pricing mechanism of disk each in storage system, obtain each disk uses price list: DP
={ dp1, dp2... dpn};
4) cost and time that user has spent are obtained, remaining budget and available time are calculated;
5) according to response time list RT and disk is estimated using the information in price list DP, the user time phase will be met
Limit and the node of user's budget are added to candidate disc-pack CN;
6) to the set CN of candidate disk, ascending sort is carried out according to energy consumption size;
7) to the identical disk of energy consumption size in the set CN of candidate disk, ascending order row is carried out according to the response time is estimated
Sequence;
8) to the identical disk of energy consumption size in the set CN of candidate disk and response time identical disk is estimated, according to
Disk carries out ascending sort using price;
9) by the pn of the candidate foremost disc-pack CNiA disk distributes to the data du in subscriber data seti;
10) by data duiBlock parallel is assigned to corresponding pniA disk carries out storage processing;
11) cost of each piecemeal the time it takes and cost is calculated;
12) time spent is equal to pniThe cost of the longest time spent in a data block, cost are equal to pniA data
The summation for the cost that block is spent;
13) cost of time and cost that user has spent are updated;
14) next data in processes user data set;
15) until having handled each of data acquisition system data, finishing scheduling.
(2) based on the preferential disk scheduling of cost (CPDS)
Main thought based on the preferential disk scheduling of cost is: under the premise of meeting user's qos requirement (including
The time limit of user and the budget of user), to optimize the cost of energy consumption and suboptimization user as target.
Basic step is as follows:
To data set (DU={ du in user's request1, du2... duu) each of data dui, repeat following
Step:
1) by data duiPiecemeal processing is carried out, the data block data_block of equal part is formedi, determine required for the data
The number pn of diski;
2) data block data_block is completed to disk all in system using response time prediction deviceiRequired for processing
Time calculated, obtain that disk is corresponding to estimate response time list: RT={ rt1, rt2... rtn};
3) according to the Real-Time Pricing mechanism of disk each in storage system, obtain each disk uses price list: DP
={ dp1, dp2... dpn};
4) cost and time that user has spent are obtained, remaining budget and available time are calculated;
5) according to response time list RT and disk is estimated using the information in price list DP, the user time phase will be met
Limit and the node of user's budget are added to candidate disc-pack CN;
6) to candidate disc-pack CN, ascending sort is carried out according to energy consumption size;
7) to the identical disk of energy consumption size in candidate disc-pack CN, ascending sort is carried out using price according to disk;
8) to the identical disk of energy consumption size in candidate disc-pack CN and using the identical disk of price, according to estimating sound
Ascending sort is carried out between seasonable;
9) by the pn of the candidate foremost back end set CNiA disk distributes to the data in subscriber data set
dui;
10) by data duiBlock parallel is assigned to corresponding pniA disk carries out storage processing;
11) cost of each piecemeal the time it takes and cost is calculated;
12) time spent is equal to pniThe cost of the longest time spent in a data block, cost are equal to pniA data
The summation for the cost that block is spent;
13) cost of time and cost that user has spent are updated;
14) next data in processes user data set;
15) until having handled each of data acquisition system data, finishing scheduling.
(3) disk scheduling based on benefit function (BFDS)
Disk scheduling based on benefit function has a premise, be need to construct a benefit function express with
The situation of Profit of the progress user of the dispatching algorithm user in the case of expense and cost expense in different times, in other words
The satisfaction situation of user, to be used to express the dynamic changes of user's qos requirement.
It constructs the main purpose of benefit function: dynamically portraying the consumption user institute with time or budget using the function
The variation of " benefit " that can obtain.So that system has a dynamic foundation or standard in scheduler task: maximizing user's
" benefit ".User defines a variety of different benefit functions according to their own needs, and simplest situation is: time-based benefit
Function is inversely proportional with task completion time, and the benefit function based on cost is then inversely proportional with the consumption of budget.The present invention uses
Benefit function figure form as shown in Figure 2, formula (1) is its mathematic(al) representation, wherein a, b, c be it is user-defined often
Number.
The basic thought of disk scheduling based on benefit function: the time-based benefit function submitted according to user
The energy of system is farthest reduced under the premise of meeting user time time limit and budget with the benefit function based on cost
Consumption is target, in sometime node, according to the cost of currently used time and cost, for the purpose of maximizing benefit, phase
Select dispatching algorithm or the dispatching algorithm preferential based on cost based on time priority with answering.
Therefore, the basic step of the disk-scheduling method based on benefit function is specific as follows:
1) user inputs corresponding time-based benefit function DB_Function (Time_used) and based on cost
Benefit function BB_Function (Cost_used);
2) to data set (DU={ du in user's request1, du2... duu) each of data dui, repeat down
State step:
A) the time Time_used that the user has currently used is calculated;
B) the cost Cost_used that the user has currently spent is calculated;
C) Saving in time costs value DV=DB_Function (Time_used) is calculated according to the value of Time_used;
D) cost benefit value CV=BB_Function (Cost_used) is calculated according to the value of Cost_used;
E) compare the size of DV and CV;
F) the dispatching algorithm TPDS based on time priority is used if DV is greater than CV;
G) the dispatching algorithm CPDS preferential based on cost is used if DV is less than CV;
H) TPDS algorithm or CPDS algorithm are randomly choosed if DV is equal to CV;
3) disk of distribution is returned.
In order to assess and verify the correlated performance of multi-speed disk-scheduling method of the invention (abbreviation MQDS method) (when response
Between, energy consumption, cost expense etc.), the simulator software CloudSimDisk of selection low cost, implementable repeated experiment.
CloudSimDisk is expanded by CloudSim, for verifying and the simulation of the perceptually relevant algorithm of the energy consumption of mock disc
Device, therefore there is good adaptability to the application background of MQDS method.
(1) the software and hardware condition tested
The software and hardware condition that table 2 is tested
(2) load simultaneous tested
Three kinds of algorithms in disk scheduling policy MQDS in order to evaluate and test the diversified QoS constraint of proposition are based on time priority
Disk scheduling TPDS, based on the preferential disk scheduling CPDS of cost and based on the disk scheduling of benefit function
BFDS.During simulation test, road real load (wiki-workload) of a period in wikipedia has been extracted
Diameter is synthesized the load of simulated experiment by the true access path: each loads a task requests for representing user,
It in CloudSimDisk simulator, is replaced with a Cloudlet, user submits 5000 tasks in 2 seconds or so time
(Cloudlet), the arrival time of each task is as shown in Figure 3.
Meanwhile the file size of each task requests storage is different, point of the file size of different Cloudlet requests
Cloth between file size 1M~10M of request as shown in figure 4, be distributed.
Under the configuration of above-mentioned parameter, the evaluation and test parameter that (two major classes) are different under isomorphism disk and Heterogeneous disk influences
Under, loop scheduling algorithm (RRDS) task carried in the three kinds of disk schedulings and CloudSimDisk in MQDS strategy exists
Performance indicator situation in terms of response time, charge costs and energy consumption.
Experiment one: the performance of the test lower four kinds of disk schedulings of isomorphism magnetic disk
This experiment is using Hitachi's HGSTUltrastarHUC109090CSS600 disk, the wherein relevant parameter of disk
It is extracted from its storage model.It is 6 that total disk number is arranged in experiment, and there are two types of the disks of speed in 6 disks: high speed
Disk and low speed disk, wherein the number of hyperdisk changes between 0~6, and the number of corresponding low speed disk is from 6~0
Change between a.The detailed parameter of the type disk is shown in Table 3.
The parameter list of 3 Hitachi's HUC109090CSS600 disk of table
Wherein the benefit function based on Saving in time costs function and based on expense is respectively as expressed by formula (2) and (3):
The parameter of the related benefit function used in experiment is as shown in table 4
Parameter list of the table 4 in relation to benefit function and its value used in an experiment
Under above-mentioned parameter setting, in experiment, the loop scheduling algorithm carried in CloudSimDisk is tested respectively
Round-Robin (RRDS) and the dispatching algorithm TPDS based on time priority that we design, based on the preferential scheduling of cost
Algorithm CPDS and dispatching algorithm BFDS based on benefit function, tests it in response time, energy consumption and charge costs etc.
Performance indicator.Resulting experimental result is as shown in table 5.
5 four kinds of disk schedulings of table experimental result under different disk configuring conditions
Three kinds of algorithms shown in the table 5 under different high speed, the configuring condition of low speed disk in MQDS: TPDS,
CPDS, BFDS, which are listed under three kinds of performance indicators under having, following feature: the dispatching algorithm TPDS tool based on time priority
There is the smallest response time, the smallest charge costs are had based on the preferential dispatching algorithm CPDS of cost, and is based on benefit function
It falls between in time, energy consumption and cost expense.And wherein when the configuration of hyperdisk is at 2~5, it is based on the time
Preferential disk scheduling policy TPDS is compared and circulation tune included in CloudSimDisk system on response time and energy consumption
Degree algorithm RRDS all has certain advantage.The three kinds of algorithms proposed simultaneously are completed in the Deadline of limitation and Budget
The request of user.Fig. 5~Fig. 7 further compare four kinds of dispatching algorithms in terms of response time, energy consumption and charge costs
Performance.
As shown in figure 5, the dispatching algorithm TPDS based on time priority all has excellent under the configuring condition of different disks
Gesture, wherein when the configuration of hyperdisk and low speed disk quite (when being respectively 3) performance advantage is the most obvious.And it is based on generation
Performance capabilities of the dispatching algorithm CPDS of valence expense then in four kinds of dispatching algorithms in terms of the response time is worst, in order to seek most
Small cost expense, because low speed disk has more cheap storage and processing expense, therefore CPDS is usually by file task tune
It spends in low speed disk and is stored and processed, therefore bring longer storage and processing delay.Therefore it is in the response time
In performance also within expection, but be their ability to the quality of service requirement for meeting user to cost sensitive.And it is based on benefit letter
Performance of several dispatching algorithm BFDS in terms of the response time is based between TPDS and CPDS.
As shown in fig. 6, based on the preferential dispatching algorithm CPDS of cost on charge costs in addition to it is all be all high speed magnetic
Disk or it is all be all low speed disk configuring condition under as the performance of other dispatching algorithms except, in other disks
Configuring condition under, the expense spent is least.And in general, the dispatching algorithm TPDS based on time priority is taking
It is then that performance is worst in four kinds of dispatching algorithms with the performance in terms of expense, because hyperdisk has higher transmission rate, but
Be there is higher storage and processing expense to generally select hyperdisk to optimize time performance as much as possible, thus with
Sacrifice charge costs are cost.And the dispatching algorithm BFDS based on benefit function then can be carried out preferably in CPDS and TPDS
Compromise.
As shown in fig. 7, the dispatching algorithm TPDS based on time priority is under the configuration with friction speed disk, consumption
Energy is minimum, therefore it is with energy-efficient characteristic.And in general, it is then four kinds of calculations based on the preferential dispatching algorithm CPDS of cost
Showed in terms of energy consumption in method it is worst, although it always selects low speed disk to save money expense, the then place of low speed disk
The increase for managing delay, also increases the consumption of energy to a certain extent.Likewise, being based on for the TPDS and CPDS that compares
Performance of the dispatching algorithm BFDS of benefit function in terms of energy consumption falls between, and can accomplish preferably to compromise.
Because in general if diversified QoS is constrained when the ratio of the hyperdisk of data center and low speed disk is suitable
Disk-scheduling method MQDS in three kinds of dispatching algorithms TPDS, CPDS and BFDS compare following of carrying in CloudSimDisk
Ring dispatching algorithm RRDS is showed with better performance.When being therefore 3 to hyperdisk, four kinds of algorithms are further deeply observed
Response time needed for each Cloudlets in wiki workload, the expense of cost and the energy of consumption.
Although the time that each Cloudlet is completed is variant, in general, it is substantially based on the scheduling of time priority
The deadline of each Cloudlet is shortest in algorithm TPDS, and loop scheduling algorithm RRDS takes second place, excellent based on cost
The deadline longest of Cloudlet in first dispatching algorithm CPDS, and the dispatching algorithm BFDS based on benefit function is then still situated between
Between TPDS and CPDS.
In the performance indicator of energy consumption, wiki-workload, the still table of the dispatching algorithm TPDS based on time priority are completed
It is now the most prominent, and consumed energy is most in general based on the preferential dispatching algorithm CPDS of cost, and it is based on benefit letter
Several dispatching algorithm BFDS is still between TPDS and CPDS.And each Cloudlet of dispatching algorithm RRDS recycled is consumed
Energy consumption fluctuation it is very big, some consumption energy it is very big, some are then seldom.Reason is that it, in 6 disk round-robin schedulings, is not examined
Consider the performance (processing speed and energy expenditure rate) of disk.
Each Cloudlet is optimal based on the preferential scheduling strategy CPDS performance of cost in terms of charge costs, is spent
Expense it is minimum.And the dispatching algorithm TPDS based on time priority then takes most expenses, the scheduling based on benefit function
Algorithm BFDS is based between CPDS and TPDS.And the charge costs of each Cloudlet of loop scheduling algorithm RRDS and its in energy
Behaving like in consumption, fluctuation greatly, the sometimes cost of corresponding Cloudlet and CPDS quite (minimum), sometimes then with
TPDS is quite (most).This does not also consider the difference of disk with it in scheduling, and it is related to circuit sequentially scheduling.
Experiment two: test Heterogeneous disk is laid out the performance of lower four kinds of disk schedulings
This part, the composition of the disk at test data center are isomeries, by the disk structure of different manufacturer's different performances
At total disk number is 6, and wherein Hitachi HUC109090CSS600 is 2, and 1 is set as hyperdisk, and 1 is set as
Low speed disk.The disk number of Seagate ST6000VN0001 is 2,1 and is set as hyperdisk, and 1 is set as low speed disk.Toshiba
The disk number of MG04SCA500E is 2, and setting hyperdisk and low speed disk are respectively 1.Three classes disk (Hitachi
HUC109090CSS600, Seagate ST6000VN0001 and Toshiba MG04SCA500E) detail parameters description be shown in Table 3 respectively, table
6, table 7.
The parameter list of 6 Seagate ST6000VN0001 disk of table
The parameter list of 7 Toshiba's MG04SCA500E disk of table
Under the configuration of above-mentioned disk, four kinds of dispatching algorithms are tested, in completing wiki workload 5000 in 2 seconds
Performance when a task requests in terms of response time, energy consumption and charge costs.It is wherein based on being related in benefit function
Parameter and experiment it is one identical, detail parameters are shown in Table 4.Resulting experimental result is as shown in table 8:
Experimental result under the configuration of the disk of 8 three kinds of different models of table
Dispatching algorithm | Response time (second) | Energy consumption (coke) | It spends (cent) |
RRDS | 84.90188538 | 2565.270668 | 205865.663 |
TPDS | 66.6532607 | 1933.050016 | 301667.6175 |
CPDS | 500.1972857 | 2067.603483 | 94825.1965 |
BFDS | 355.6826107 | 2022.752327 | 163772.6702 |
As shown in table 8, under the disk model of isomery and the different disks with speed, the dispatching algorithm based on time priority
TPDS has optimal performance on response time and energy consumption, and then carries on as usual ground based on the preferential dispatching algorithm CPDS of cost
Showed on charge costs optimal, the expense of cost is the one third or so of other dispatching algorithms.And based on benefit function
Performance of the dispatching algorithm BFDS in the performance indicator on the response time, energy consumption and charge costs between TPDS and CPDS it
Between, can preferably it compromise therebetween.And the loop scheduling algorithm RRDS carried in CloudSimDisk is being responded
The energy for being better than CPDS and BFDS on time, but being inferior to TPDS, but consume is most in four kinds of dispatching algorithms.And it is taking
It is better than TPDS with expense aspect, but is inferior to CPDS and BFDS.Therefore isomery disk configuration under, at a high speed with low speed disk
In the case of proportion is suitable, the present invention is proposed and three kinds of dispatching algorithms of design are guaranteeing the requirement of the diversified QoS of user
Under the premise of, there is certain consumption reduction effect compared to RRDS.
The present invention is directed in cloud computing environment the characteristics of the diversification of user's qos requirement, mobilism, is based on multi-speed disk
Efficiency framework, in the level of disk, using meet the diversified qos requirement of user and reduce system energy consumption as target, if
Count and realize the dispatching method of the data in multi-speed disk.
The present invention is on the basis of illustrating the related definition of qos requirement and diversified qos requirement, to current related QoS
Under the premise of the correlative study of task schedule is investigated, the multi-speed data in magnetic disk dispatching method of diversified QoS constraint is proposed
(abbreviation MQDS) finally generates load using the true access path of wiki workload, simulates in the CloudSimDisk of extension
The adaptability and the energy efficiency under certain condition of the diversified QoS of three kinds of dispatching algorithms in MQDS are verified and had evaluated on device,
MQDS become towards in the middle-level progressive high energy efficiency technology of cloud storage system third level (disk level) reduce be
The method for energy consumption of uniting.
The diversification required for existing disk scheduling QoS of customer lacks ill-conceived status, and the present invention mentions
Go out in disk-scheduling method MQDS, the MQDS method of diversified QoS constraint and contained three kinds of dispatching algorithms: based on time priority
Dispatching algorithm TPDS, the dispatching algorithm BFDS based on the preferential dispatching algorithm CPDS of cost and based on benefit function.Based on when
Between preferential dispatching algorithm (TPDS) for time sensitivity the high task of requirement of real-time and design, and it is preferential based on cost
Dispatching algorithm (CPDS) it is then sensitive for charge costs and the lower task of requirement of real-time is designed, be based on benefit function
Dispatching algorithm (BFDS) then for time-sensitive and cost-sensitive may dynamic change the case where task and design.Exist simultaneously
In CloudSimDisk by four kinds of dispatching algorithms of data center's simulation test of the framework of multi-speed disk (including
The loop scheduling algorithm RRDS that CloudSimDisk itself is carried) to containing the wiki workload of 5000 task requests
Performance when being scheduled, in the response time, in three performance indicators such as energy consumption and charge costs.The result table of simulated experiment
Bright: in the case where hyperdisk and low speed disk match comparable situation, three kinds of algorithms in the MQDS of proposition are generally with the obvious advantage,
And specific in terms of time performance, then the dispatching algorithm TPDS performance based on time priority is optimal, and is then in terms of charge costs
There is optimal performance based on the preferential dispatching algorithm CPDS of cost, and in terms of energy consumption index, three kinds of dispatching algorithms of proposition
It is superior to the included loop scheduling algorithm RRDS of CloudSimDisk, it is shown that the consumption reduction ability of MQDS strategy.It is additionally based on effect
The dispatching algorithm BFDS of beneficial function in three kinds of performance indicators between TPDS and CPDS, can preferably therebetween
Compromise is obtained, mostly a kind of selection of user is given, presents the dispatching that MQDS has diversified QoS constraint.
Embodiments of the present invention above described embodiment only expresses, the description thereof is more specific and detailed, but can not
Therefore limitations on the scope of the patent of the present invention are interpreted as.It should be pointed out that for those of ordinary skill in the art,
Without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to protection model of the invention
It encloses.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (8)
1. a kind of multi-speed disk-scheduling method of diversification QoS constraint, which is characterized in that according to different user's qos requirements, hold
The different dispatching algorithm of row.
2. the disk-scheduling method of diversification QoS constraint according to claim 1, which is characterized in that different scheduling is calculated
Method respectively corresponds are as follows: disk scheduling based on time priority based on the preferential disk scheduling of cost and is based on benefit
The disk-scheduling method of function.
3. the disk-scheduling method of diversification QoS constraint according to claim 1 to 2, which is characterized in that excellent based on the time
First disk scheduling is under the premise of meeting user's qos requirement, to optimize the response time of energy consumption and suboptimization user
For target.
4. the disk-scheduling method of diversification QoS constraint according to claim 1 to 3, which is characterized in that excellent based on the time
First disk scheduling includes:
To data set (DU={ du in user's request1, du2... duu) each of data dui, repeat following step:
By data duiPiecemeal processing is carried out, the data block data_block of equal part is formedi, determine disk required for the data
Number pni;
Data block data_block is completed to disk all in system using response time prediction deviceiTime required for handling
It is estimated, obtains that each disk is corresponding to estimate response time list: RT={ rt1, rt2... rtn};
According to the Real-Time Pricing mechanism of disk each in storage system, obtain each disk uses price list: DP={ dp1,
dp2... dpn};
The cost and time that user has spent are obtained, remaining budget and available time are calculated;
According to response time list RT and disk is estimated using the information in price list DP, user time time limit and use will be met
The node of family budget is added to candidate disc-pack CN;
To the set CN of candidate disk, ascending sort is carried out according to energy consumption size;
To the identical disk of energy consumption size in the set CN of candidate disk, ascending sort is carried out according to the response time is estimated;
To the identical disk of energy consumption size in the set CN of candidate disk and response time identical disk is estimated, is made according to disk
Ascending sort is carried out with price;
By the pn of the candidate foremost disc-pack CNiA disk distributes to the data du in subscriber data seti;By data dui
Block parallel is assigned to corresponding pniA disk carries out storage processing;
Calculate the cost of each piecemeal the time it takes and cost;
The time of cost is equal to pniThe cost of the longest time spent in a data block, cost are equal to pniA data block is spent
The summation for the cost taken;
Update the cost of time and cost that user has spent;
Next data in processes user data set;
Until having handled each of data acquisition system data, finishing scheduling.
5. the disk-scheduling method of diversified QoS constraint described in -4 according to claim 1, which is characterized in that excellent based on cost
First disk scheduling is under the premise of meeting user's qos requirement to optimize the cost of energy consumption and suboptimization user as mesh
Mark.
6. the disk-scheduling method of diversified QoS constraint described in -5 according to claim 1, which is characterized in that excellent based on cost
First disk scheduling includes:
To data set (DU={ du in user's request1, du2... duu) each of data dui, repeat following step:
By data duiPiecemeal processing is carried out, the data block data_block of equal part is formedi, determine disk required for the data
Number pni;
Data block data_block is completed to disk all in system using response time prediction deviceiTime required for handling
It is calculated, obtains that disk is corresponding to estimate response time list: RT={ rt1, rt2... rtn};
According to the Real-Time Pricing mechanism of disk each in storage system, obtain each disk uses price list: DP={ dp1,
dp2... dpn};
The cost and time that user has spent are obtained, remaining budget and available time are calculated;
According to response time list RT and disk is estimated using the information in price list DP, user time time limit and use will be met
The node of family budget is added to candidate disc-pack CN;
To candidate disc-pack CN, ascending sort is carried out according to energy consumption size;
To the identical disk of energy consumption size in candidate disc-pack CN, ascending sort is carried out using price according to disk;
To the identical disk of energy consumption size in candidate disc-pack CN and using the identical disk of price, according to estimating the response time
Carry out ascending sort;
By the pn of the candidate foremost back end set CNiA disk distributes to the data du in subscriber data seti;
By data duiBlock parallel is assigned to corresponding pniA disk carries out storage processing;
Calculate the cost of each piecemeal the time it takes and cost;
The time of cost is equal to pniThe cost of the longest time spent in a data block, cost are equal to pniA data block is spent
The summation for the cost taken;
Update the cost of time and cost that user has spent;
Next data in processes user data set;
Until having handled each of data acquisition system data, finishing scheduling.
7. the disk-scheduling method of diversified QoS constraint described in -6 according to claim 1, which is characterized in that be based on benefit letter
Several disk-scheduling methods includes: building benefit function to express the progress with dispatching algorithm, and user is when different
Between situation of Profit in the case of expense and cost expense, to express the dynamic changes of user's qos requirement.
8. the disk-scheduling method of diversified QoS constraint described in -7 according to claim 1, which is characterized in that be based on benefit letter
The step of several disk-scheduling methods includes:
1) user inputs corresponding time-based benefit function DB_Function (Time_used) and the benefit based on cost
Function BB_Function (Cost_used);
2) to data set (DU={ du in user's request1, du2... duu) each of data dui, repeat following steps
It is rapid:
Calculate the time Time_used that the user has currently used;
Calculate the cost Cost_used that the user has currently spent;
Saving in time costs value DV=DB_Function (Time_used) is calculated according to the value of Time_used;
Cost benefit value CV=BB_Function (Cost_used) is calculated according to the value of Cost_used;
Compare the size of DV and CV;
The dispatching algorithm TPDS based on time priority is used if DV is greater than CV;
The dispatching algorithm CPDS preferential based on cost is used if DV is less than CV;
TPDS algorithm or CPDS algorithm are randomly choosed if DV is equal to CV;
3) disk of distribution is returned.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811403707.1A CN109614038A (en) | 2018-11-23 | 2018-11-23 | A kind of multi-speed disk-scheduling method of diversification QoS constraint |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811403707.1A CN109614038A (en) | 2018-11-23 | 2018-11-23 | A kind of multi-speed disk-scheduling method of diversification QoS constraint |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109614038A true CN109614038A (en) | 2019-04-12 |
Family
ID=66004847
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811403707.1A Pending CN109614038A (en) | 2018-11-23 | 2018-11-23 | A kind of multi-speed disk-scheduling method of diversification QoS constraint |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109614038A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113254222A (en) * | 2021-07-13 | 2021-08-13 | 苏州浪潮智能科技有限公司 | Task allocation method and system for solid state disk, electronic device and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101271407A (en) * | 2008-05-13 | 2008-09-24 | 武汉理工大学 | Gridding scheduling method based on energy optimization |
CN103024048A (en) * | 2012-12-17 | 2013-04-03 | 南京邮电大学 | Resources scheduling method under cloud environment |
CN104657220A (en) * | 2015-03-12 | 2015-05-27 | 广东石油化工学院 | Model and method for scheduling for mixed cloud based on deadline and cost constraints |
CN108768703A (en) * | 2018-05-15 | 2018-11-06 | 长沙理工大学 | A kind of energy consumption optimization method, the cloud computing system of cloud workflow schedule |
-
2018
- 2018-11-23 CN CN201811403707.1A patent/CN109614038A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101271407A (en) * | 2008-05-13 | 2008-09-24 | 武汉理工大学 | Gridding scheduling method based on energy optimization |
CN103024048A (en) * | 2012-12-17 | 2013-04-03 | 南京邮电大学 | Resources scheduling method under cloud environment |
CN104657220A (en) * | 2015-03-12 | 2015-05-27 | 广东石油化工学院 | Model and method for scheduling for mixed cloud based on deadline and cost constraints |
CN108768703A (en) * | 2018-05-15 | 2018-11-06 | 长沙理工大学 | A kind of energy consumption optimization method, the cloud computing system of cloud workflow schedule |
Non-Patent Citations (1)
Title |
---|
游新冬等: "基于效益函数的网格任务调度算法", 《计算机科学》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113254222A (en) * | 2021-07-13 | 2021-08-13 | 苏州浪潮智能科技有限公司 | Task allocation method and system for solid state disk, electronic device and storage medium |
CN113254222B (en) * | 2021-07-13 | 2021-09-17 | 苏州浪潮智能科技有限公司 | Task allocation method and system for solid state disk, electronic device and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110096349B (en) | Job scheduling method based on cluster node load state prediction | |
Mao et al. | Scaling and scheduling to maximize application performance within budget constraints in cloud workflows | |
Singh et al. | QRSF: QoS-aware resource scheduling framework in cloud computing | |
CN105302630B (en) | A kind of dynamic adjusting method and its system of virtual machine | |
CN103761147B (en) | The management method and system of calculated examples in a kind of cloud platform | |
CN106844051A (en) | The loading commissions migration algorithm of optimised power consumption in a kind of edge calculations environment | |
CN103064744B (en) | The method for optimizing resources that a kind of oriented multilayer Web based on SLA applies | |
CN108804227A (en) | The method of the unloading of computation-intensive task and best resource configuration based on mobile cloud computing | |
Zhang et al. | Energy-efficient tasks scheduling heuristics with multi-constraints in virtualized clouds | |
Mekala et al. | A DRL-based service offloading approach using DAG for edge computational orchestration | |
Chakravarthi et al. | TOPSIS inspired budget and deadline aware multi-workflow scheduling for cloud computing | |
CN103488539A (en) | Data center energy saving method based on central processing unit (CPU) dynamic frequency modulation technology | |
CN110308967A (en) | A kind of workflow cost based on mixed cloud-delay optimization method for allocating tasks | |
Zhou et al. | Deep reinforcement learning-based methods for resource scheduling in cloud computing: A review and future directions | |
Alshathri et al. | A New Reliable System For Managing Virtual Cloud Network. | |
CN112363827A (en) | Multi-resource index Kubernetes scheduling method based on delay factors | |
Addya et al. | A game theoretic approach to estimate fair cost of VM placement in cloud data center | |
Sun et al. | ET2FA: A hybrid heuristic algorithm for deadline-constrained workflow scheduling in cloud | |
Li et al. | A dynamic I/O sensing scheduling scheme in Kubernetes | |
CN108768703A (en) | A kind of energy consumption optimization method, the cloud computing system of cloud workflow schedule | |
Li et al. | A QoS-based scheduling algorithm for instance-intensive workflows in cloud environment | |
Taghinezhad-Niar et al. | Reliability, rental-cost and energy-aware multi-workflow scheduling on multi-cloud systems | |
Yang et al. | Design of kubernetes scheduling strategy based on LSTM and grey model | |
CN109614038A (en) | A kind of multi-speed disk-scheduling method of diversification QoS constraint | |
Zhu et al. | A multi-resource scheduling scheme of Kubernetes for IIoT |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190412 |
|
WD01 | Invention patent application deemed withdrawn after publication |