CN102207883A - Transaction scheduling method of heterogeneous distributed real-time system - Google Patents

Transaction scheduling method of heterogeneous distributed real-time system Download PDF

Info

Publication number
CN102207883A
CN102207883A CN2011101458047A CN201110145804A CN102207883A CN 102207883 A CN102207883 A CN 102207883A CN 2011101458047 A CN2011101458047 A CN 2011101458047A CN 201110145804 A CN201110145804 A CN 201110145804A CN 102207883 A CN102207883 A CN 102207883A
Authority
CN
China
Prior art keywords
affairs
time
priority
transaction
nuclear
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
Application number
CN2011101458047A
Other languages
Chinese (zh)
Inventor
王非
黄本雄
段炼
邓磊
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.)
Huazhong University of Science and Technology
Original Assignee
Huazhong University of Science and Technology
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 Huazhong University of Science and Technology filed Critical Huazhong University of Science and Technology
Priority to CN2011101458047A priority Critical patent/CN102207883A/en
Publication of CN102207883A publication Critical patent/CN102207883A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Computer And Data Communications (AREA)

Abstract

The invention relates to a transaction priority determination method in the field of heterogeneous distributed parallel processing. The method comprises the following steps of: determining the priority of a global transaction; distributing sub-transactions which are obtained by dividing the global transaction according to physical positions of data relevant to the sub-transactions; dynamically determining the priority of respective parts of local transactions; performing parallel processing on the local transactions; and returning an execution result. The invention provides a flow for dynamically determining the priority of the transaction based on a heterogeneous distributed system, so the real-time property and reliability of the system on processing of the global transaction are guaranteed.

Description

The transaction scheduling method of heterogeneous distributed real-time system
Technical field
The invention belongs to the distributed parallel process field, particularly be based on the dynamic transaction dispatching method of heterogeneous distributed real-time system.
Background technology
Many Information System configuration processes as traditional industries such as electric power, oil, petrochemical industry, traffic, logistics have been accelerated in the continuous evolution of current I T technology.But rapid growth on the data bulk that needs to handle in real time along with infosystem in the practical application disperses on the position to deposit, and traditional centralized processing system can't tackle, and therefore adopting distributed parallel to handle Real-time Transaction becomes development trend.The raising of this significant progress on dominant frequency and bandwidth of microprocessor simultaneously and polycaryon processor manufacture craft has further improved the ability of system's executed in parallel affairs.
But how to make full use of the performance of distributed structure/architecture and polycaryon processor, guarantee the efficient of system handles global transaction, the real-time that promptly guarantees global transaction is the problem that needs concern, therefore need study the priority that rational strategy is determined affairs at heterogeneous distributed multi-core environment, system is that standard is dispatched with this priority, the affairs that guarantee high priority are preferentially carried out, and have satisfied the real-time of affairs.Therefore, be necessary that research is applicable to and according to each processor cores running status and affairs forerunner implementation status, dynamically carries out the transaction scheduling method by heterogeneous distributed multi-core environment.
Summary of the invention
The objective of the invention is to satisfy the actual demand of distributed system, a kind of dynamic transaction scheduling method is provided in order to improve the parallel processing performance of distributing real-time data bank system handles complex transaction.
For achieving the above object, method provided by the invention may further comprise the steps:
Step 1, the global transaction scheduler of global transaction management system distribute priority for each global transaction;
Step 2, global transaction scheduler judge whether certain global transaction can decompose; If these affairs can be decomposed, then the plurality of sub affairs are analyzed and be decomposed into to the global transaction scheduler to this global transaction by the physical location of semantic and the data that relate to, the subtransaction that decomposition is obtained is assigned to the local affairs that become each node on the appropriate nodes, and the initial priority of each local affairs is the priority of former global transaction;
Otherwise, this global transaction is assigned on the appropriate nodes as local affairs;
Step 3, local real time's transaction management system makes up the directed acyclic graph model, and this directed acyclic graph model comprises the transaction layer of some forerunners of having, follow-up dependence;
Step 4 is assigned the node of subtransaction to each, carries out following steps respectively:
Step 4-1, determine the time that each kernel of the polycaryon processor of this intranodal can begin to carry out new affairs the earliest; Because each kernel etc. also exist in the pending formation some other affairs etc. pending, so transaction scheduling still needs to wait for after this process nuclear.Avail (m) shows that nuclear m will wait in the pending formation existing affairs to be finished and can begin to carry out the time of new affairs, this time can be weighed by pending affairs scale such as estimating in this nuclear, and the affairs scale is read and write number of times by affairs in the unit interval and judged.
But in fact owing to be based on the scheduling mode of the preemptive type of priority, the time ratio avail (m) that kernel really begins to carry out these affairs wants early.
Step 4-2 determines that all forerunner's affairs of each affairs of this node are finished the time; When affairs after successful execution finishes in process nuclear, local transaction scheduler can writing time for its subsequent transaction inquiry, so affairs determine that by inquiring about local transaction scheduler all forerunner's affairs are finished the time.
If forerunner's affairs of these affairs are inlet affairs then this time to be made as 0.
Step 4-3 determines the communication time loss between each affairs of this node; If father's affairs and subtransaction are carried out on same nuclear, then data transfer time between the two can be considered as 0.If carry out but be assigned on the different IPs, there is data transfer time between then, linear with the scale of affairs itself, be expressed as:
Trans_T mn=Trans_b mn+Trans_a i·t i (1)
Trans_T wherein MnExpression is delivered to the time that nuclear n needs with data at nuclear m.t iWhat represent is the scale of affairs, weighs Trans_a by the read-write number of times in the unit interval iExpression be the linear factor of scale and the affairs of affairs time of between different nodes, transmitting.Trans_b MnWhat represent is initiation parameter.
If forerunner's affairs of same these affairs are inlet affairs then this time to be made as 0.
Step 4-4, affairs are at the earliest start time EST of each kernel really.EST (a i, m) expression affairs a iNuclear on the m the earliest can the start time.EST (a i, m) can calculate by following formula:
EST (a i, m)=and Max{avail (m), Max{EFT (a x)+Trans_T Mn(a wherein x∈ pred (a i)) (2)
EFT(a i,m)=T i+EST(a i,m) (3)
Pred (a wherein i) expression affairs a iAll forerunner's affairs, avail (m) shows the idle time of can early start carrying out next affairs of nuclear m, weigh by the affairs scale of estimating wait scheduling in this nuclear, the affairs scale is judged by affairs read-write number of times in the unit interval equally.Trans_T MnThe loss of expression data transfer time obtains by formula (1), then is made as 0 if transmit in same nuclear.Affairs a iAt nuclear m start time EST (a i, m) add affairs execution time T itself iObtain affairs a iTime EFT (a the earliest is finished on nuclear m i, m).So affairs earliest start time in nuclear is determined jointly by three parts such as time loss that kernel can begin to carry out the time of affairs, time that all forerunner father's affairs of these affairs are finished and path transmission.
Step 4-5, when affairs when last layer is carried out this layer, recomputate the priority of each affairs in this layer, priority can calculate according to following formula:
rank(a i)=1/min{EST(a i,1),...,EST(a i,n)}+rank 0 (4)
Rank (a wherein i) expression affairs a iPriority, suppose that node is the n core processor, therefrom choose and guarantee affairs a iThe nuclear that early start is carried out is scheduling destination, rank 0The priority size of expression inlet affairs is the priority of global transaction;
Step 4-6 according to the priority that step 4-5 calculates, resequences to priority query;
Step 4-7, kernel is carried out affairs according to priority query, and local transaction scheduler monitors executing state and feeds back to the global transaction scheduler; If run succeeded, then write down execution result and execution concluding time; If carry out failure, then return error code and carry out necessary rolling back action.
Step 4-8 judges with each affairs of layer whether all be finished in each nuclear; If not then wait for;
Step 4-9 is finished with layer, then changes lower floor over to, returns step 4-5, up to carrying out the outlet affairs, promptly should the part affairs be finished;
Step 4-10 returns to the global transaction management system with local transaction execution results, carries out the result for it and gathers and return to application server.
Further improve,, make urgent affairs preferentially carry out by the method for setting priority, thereby guaranteed the real-time of affairs for Real-time Transaction.
The present invention proposes a kind of method of dynamically determining affairs priority, node is dynamically determined the priority of affairs according to the current implementation status of processor cores, and be standard with this priority, transaction scheduling is carried out to idle relatively kernel, reduced the execution time of local affairs, thereby improved response performance on the whole global transaction.The priority of affairs is dynamically to adjust according to the executing state of front, thereby has good adaptability, has avoided the estimation to the affairs execution time itself simultaneously in the priority deterministic process, thereby has reduced computational complexity, is convenient to practical operation.
Description of drawings
Above-mentioned explanation is the general introduction of technical solution of the present invention, for can clearer understanding technological means of the present invention, and can be implemented according to the content of instructions, and in order to allow above-mentioned purpose of the present invention, feature and advantage become apparent, below list accompanying drawing and in conjunction with detailed implementation step.
Fig. 1 is a heterogeneous distributed real-time system structural representation of the present invention.
Fig. 2 is the directed acyclic graph that the present invention relates to complex transaction partition and each local affairs.
Fig. 3 is the schematic flow sheet of definite affairs priority of the present invention.
Embodiment
For further setting forth the present invention,, realize the method based on definite affairs priority of heterogeneous distributed real-time system of the present invention's proposition below in conjunction with accompanying drawing and detailed implementation step for reaching expection goal of the invention and effect.Core concept of the present invention is, the priority of the affairs of node is dynamically determined to be assigned in deadline by considering the affairs forerunner, the state status of node itself and the aspects such as time loss that affairs are transmitted, the affairs priority scheduling that guarantees high priority is carried out, thereby has guaranteed the real-time of global transaction.Further describe embodiment of the present invention below in conjunction with accompanying drawing.
Fig. 1 be the present invention based on the structural drawing of heterogeneous distributed system.Heterogeneous distributed system comprises global transaction management system, some local real times transaction management system 1,2...n, the global transaction management system of application server deploy as root node, global transaction management system root node by network and strange land as the local transaction management system 1 of leaf node, the composition distributed system of 2...n.
Wherein the processing of the not responsible affairs of global transaction management system itself is dispatched on the appropriate nodes and just will decompose each subtransaction that obtains by global transaction by the global transaction scheduler.And monitor the implementation status of subtransaction in real time at each node.To merge by the result that each node returns at last, feed back to client.
Local transaction management system is made up of multiple-core server, responsible processing is stored in the local real time data of being obtained by acquisition layer that obtains, each local transaction management system has a local transaction scheduler to be responsible for dispatching affairs and monitors polycaryon processor executed in parallel affairs situation, and the result that will carry out feeds back to the global transaction management system.Native system is a kind of heterogeneous distributed system, so the system software of the number of the nuclear that comprises of the polycaryon processor of each node and operation can be inequality.
Fig. 2 is that the global transaction that the present invention relates to decomposes and local transaction model synoptic diagram.Global transaction T as shown in Figure 2 can be divided into global transaction T according to the node location at the data place of its operation: T={ST 0| ST 1∪ ST 2∪ ST 3.Subtransaction ST wherein 1Physical location according to its data that relate to is assigned to node 1, and its transaction model can be used directed acyclic graph (DAG) expression, subtransaction ST as can be seen 1The step a that comprises 2, a 3Can executed in parallel; Subtransaction ST in like manner 2Be assigned to node 2, by its transaction model comprise equally as can be seen can executed in parallel part; Subtransaction ST 3Be assigned to node 3, it can not further decompose for basic operation also just can not executed in parallel.ST 0Be responsible for coordinating each subtransaction execution and result's merging.
For subtransaction ST 1, ST 2, because its structure is comparatively complicated, the affairs that it can be comprised are assigned to process nuclear different in the polycaryon processor and carry out parallel processing.Need to prove that each kernel of said polycaryon processor has independently processing unit and Cache here, each kernel is by message communicating, can be with the Cache of kernel as pending formations such as affairs.If being assigned to the affairs that subtransaction comprised that same node belongs to different global transactions is assigned on the same process nuclear, then need according to priority and by the arrangement of time execution sequence, each pending affairs are waited in waiting pending formation according to this order, guarantee the preferential execution of high priority like this, thereby guaranteed the real-time of global transaction on the whole.
Fig. 3 is a process flow diagram of the present invention, the following steps that this flow process specifically comprises:
Step 301, the global transaction scheduler of global transaction management system distribute priority for each global transaction.The global transaction here refers to the affairs of also not carrying out resolution process imported into from client.Can directly adopt classical priority to determine method for the priority of global transaction, as the off period the earliest override, can arrive off period override, excess time the shortest override etc. the earliest.Distributed after the priority, entered affairs processing layer according to from high to low sequence arrangement.The order support of global transaction is seized.
Step 302, global transaction scheduler judge whether global transaction can decompose.Because affairs can be decomposed into:
T={ST 1∪ST 2∪K∪ST m}
Wherein m represents to constitute the number of the real-time subtransaction of distributed real-time affairs T.When m>1, show that these affairs can decompose; Then these affairs are as a whole when m=1 can not carry out operation splitting.The global transaction scheduler is analyzed by the physical location of semantic and the data that relate to this global transaction and is decomposed.With the subtransaction that decomposition obtains, for example the mode by hash or delivery is assigned on the appropriate nodes.The priority and the former global transaction of each subtransaction that decomposes are consistent simultaneously.
Step 303 is decomposed by global transaction and to be become local affairs after the plurality of sub affairs obtain are assigned to each node, and local transaction management system makes up directed acyclic graph (DAG) model by the operation steps of local affairs.
In the DAG model, there is dependence between each affairs, need to wait for after his father's affairs are finished the result can be begun to carry out as input parameter.General by forerunner pred (a i) and follow-up succ (a i) this data, the control dependence of expression between the affairs.As the local affairs ST of Fig. 2 2DAG shown in, affairs a wherein 1Without any forerunner's affairs, be called the inlet affairs, a 5Become the outlet affairs without any subtransaction.The affairs that will enter the mouth a 1All affairs with dependence are included into the DAG structure second layer, ST 2In be a 2And a 3, a then 1Be a 2, a 3Father's affairs, a 2, a 3Be a 1Subtransaction, a simultaneously 2, a 3Can be assigned to different disposal nuclear and go up executed in parallel, can make up the 3rd layer successively till arriving the outlet affairs.Therefore affairs can successively be dispatched on node, begin to carry out this layer affairs again after soon all forerunner's affairs of affairs all will be finished.
Need to prove the analysis of local affairs and monitor that its each several part all is to be finished by the local transaction scheduler of intranodal in the implementation status of each kernel.
Step 304 determines that each kernel of this node can begin to carry out the time of new affairs the earliest.Because each kernel etc. also exist in the pending formation some other affairs etc. pending, so transaction scheduling still needs to wait for after this process nuclear.Avail (m) shows that nuclear m will wait in the pending formation existing affairs to be finished and can begin to carry out the time of new affairs, this time can be weighed by pending affairs scale such as estimating in this nuclear, and the affairs scale is read and write number of times by affairs in the unit interval and judged.
But in fact owing to be based on the scheduling mode of the preemptive type of priority, the time ratio avail (m) that kernel really begins to carry out these affairs wants early.
Step 305 determines that all forerunner's affairs of these affairs are finished the time.When affairs after successful execution finishes in process nuclear, local transaction scheduler can writing time for its subsequent transaction inquiry, so affairs determine that by inquiring about local transaction scheduler all forerunner's affairs are finished the time.
If forerunner's affairs of these affairs are inlet affairs then this time to be made as 0.
Step 306 is determined the communication time loss between the affairs.If father's affairs and subtransaction are carried out on same nuclear, then data transfer time between the two can be considered as 0.If carry out but be assigned on the different IPs, there is data transfer time between then, linear with the scale of affairs itself, be expressed as:
Trans_T mn=Trans_b mn+Trans_a i·t i (1)
Trans_T wherein MnExpression is delivered to the time that nuclear n needs with data at nuclear m.t iWhat represent is the scale of affairs, weighs Trans_a by the read-write number of times in the unit interval iExpression be the linear factor of scale and the affairs of affairs time of between different nodes, transmitting.Trans_b MnWhat represent is initiation parameter.
If forerunner's affairs of same these affairs are inlet affairs then this time to be made as 0.
Step 307, affairs are at the earliest start time EST of each kernel really.EST (a i, m) expression affairs a iNuclear on the m the earliest can the start time.EST (a i, m) can calculate by following formula:
EST (a i, m)=and Max{avail (m), Max{EFT (a x)+Trans_T Mn(a wherein x∈ pred (a i)) (2)
EFT(a i,m)=T i+EST(a i,m) (3)
Pred (a wherein i) expression affairs a iAll forerunner's affairs, avail (m) shows the idle time of can early start carrying out next affairs of nuclear m, weigh by the affairs scale of estimating wait scheduling in this nuclear, the affairs scale is judged by affairs read-write number of times in the unit interval equally.Trans_T MnThe loss of expression data transfer time obtains by formula (1), then is made as 0 if transmit in same nuclear.Affairs a iAt nuclear m start time EST (a i, m) add affairs execution time T itself iObtain affairs a iTime EFT (a the earliest is finished on nuclear m i, m).So affairs earliest start time in nuclear is determined jointly by three parts such as time loss that kernel can begin to carry out the time of affairs, time that all forerunner father's affairs of these affairs are finished and path transmission.
When affairs when the last layer of logic is delivered to this layer, need redefine priority.Priority can calculate according to following formula:
rank(a i)=1/min{EST(a i,1),...,EST(a i,n)}+rank 0 (4)
Rank (a wherein i) expression affairs a iPriority, suppose that node is the n core processor, therefrom choose and guarantee affairs a iThe nuclear that early start is carried out is the scheduling destination.Rank 0The priority size of expression inlet affairs is the priority of global transaction.Can draw time that affairs can begin to carry out the earliest from this formula, priority is also corresponding the highest.
Step 308, each nuclear all has pending formations such as Cache deposits, after new affairs enter formation, medium pending each affairs of formation can according to separately priority and by the time be standard according to carry out at first override or off period the earliest override at inner queue minor sort again.Guarantee high priority the affairs of the shortest off period preferentially carry out.Kernel is carried out this affairs afterwards, and local transaction scheduler monitors executing state and feeds back to the global transaction scheduler.Record execution result and execution concluding time if run succeeded; If carry out failure, then return error code and carry out necessary rolling back action.
Step 309 judges with each affairs of layer whether all be finished in each nuclear.If not then wait for.
Step 310, layer is finished and then changes lower floor over to together, determines priority and scheduled for executing by same strategy, up to carrying out the outlet affairs, promptly should the part affairs be finished.Local transaction execution results is returned to the global transaction manager carries out the result for it and gathers and return to application server.
It should be noted last that, above embodiment is only unrestricted in order to technical scheme of the present invention to be described, although the present invention is had been described in detail with reference to preferred embodiment, those of ordinary skill in the art is to be understood that, can make amendment or be equal to replacement technical scheme of the present invention, and not breaking away from the spirit and scope of technical solution of the present invention, it all should be encompassed in the middle of the claim scope of the present invention.

Claims (2)

1. the transaction scheduling method of a heterogeneous distributed real-time system is characterized in that, may further comprise the steps:
Step 1, the global transaction scheduler of global transaction management system distribute priority for each global transaction;
Step 2, global transaction scheduler judge whether certain global transaction can decompose; If these affairs can be decomposed, then the plurality of sub affairs are analyzed and be decomposed into to the global transaction scheduler to this global transaction by the physical location of semantic and the data that relate to, the subtransaction that decomposition is obtained is assigned to and becomes local affairs on the appropriate nodes, and the initial priority of each local affairs is the priority of former global transaction;
Otherwise, this global transaction is assigned on the appropriate nodes as local affairs;
Step 3, the local real time's transaction management system on each node make up the directed acyclic graph model to the local affairs of this node respectively, and this directed acyclic graph model comprises the transaction layer of some forerunners of having, follow-up dependence;
Step 4 is assigned the node of subtransaction to each, carries out following steps respectively:
Step 4-1, determine the time that each kernel of the polycaryon processor of this intranodal can begin to carry out new affairs the earliest;
Step 4-2 determines that all forerunner's affairs of each affairs of this node are finished the time;
Step 4-3 determines the communication loss time between each affairs of this node; If father's affairs and subtransaction are carried out on same nuclear, then data transfer time between the two can be considered as 0; Carry out if be assigned on the different IPs, then the communication loss time between each affairs is:
Figure 2011101458047100001DEST_PATH_IMAGE002
(1)
Wherein Data are being examined in expression
Figure 2011101458047100001DEST_PATH_IMAGE006
Be delivered to nuclear
Figure 2011101458047100001DEST_PATH_IMAGE008
The time that needs,
Figure 2011101458047100001DEST_PATH_IMAGE010
What represent is the scale of affairs, weigh by the read-write number of times in the unit interval,
Figure 2011101458047100001DEST_PATH_IMAGE012
Expression be the linear factor of scale and the affairs of affairs time of between different nodes, transmitting,
Figure 2011101458047100001DEST_PATH_IMAGE014
What represent is initiation parameter;
Step 4-4, each affairs is in the earliest start time of each kernel really;
Figure 2011101458047100001DEST_PATH_IMAGE016
The expression affairs
Figure 2011101458047100001DEST_PATH_IMAGE018
At nuclear
Figure 170361DEST_PATH_IMAGE006
On the earliest can the start time,
Figure 25184DEST_PATH_IMAGE016
Can calculate by following formula:
Figure 2011101458047100001DEST_PATH_IMAGE020
(2)
Figure 2011101458047100001DEST_PATH_IMAGE022
(3)
Wherein
Figure 2011101458047100001DEST_PATH_IMAGE024
The expression affairs
Figure 680288DEST_PATH_IMAGE018
All forerunner's affairs, ,
Figure 2011101458047100001DEST_PATH_IMAGE028
Show nuclear
Figure 823300DEST_PATH_IMAGE006
The idle time of can early start carrying out next affairs, to weigh by estimating this nuclear in, to wait for the affairs scale of dispatching, the affairs scale is read and write number of times by affairs in the unit interval equally and is judged;
Figure 2011101458047100001DEST_PATH_IMAGE030
The loss of expression data transfer time obtains by formula (1), then is made as 0 if transmit in same nuclear; Affairs
Figure 2011101458047100001DEST_PATH_IMAGE032
At nuclear
Figure 388404DEST_PATH_IMAGE006
Start time
Figure 2011101458047100001DEST_PATH_IMAGE034
Add the affairs execution time itself Obtain affairs
Figure 647085DEST_PATH_IMAGE032
At nuclear
Figure 851801DEST_PATH_IMAGE006
On be finished the time the earliest
Figure 2011101458047100001DEST_PATH_IMAGE038
Step 4-5, when affairs when last layer is carried out this layer, recomputate the priority of each affairs in this layer, priority can calculate according to following formula:
Figure 2011101458047100001DEST_PATH_IMAGE040
(4)
Wherein
Figure 2011101458047100001DEST_PATH_IMAGE042
The expression affairs
Figure 923138DEST_PATH_IMAGE018
Priority, suppose that node is
Figure 280432DEST_PATH_IMAGE008
Core processor is therefrom chosen the assurance affairs
Figure 398430DEST_PATH_IMAGE018
The nuclear that early start is carried out is the scheduling destination,
Figure 2011101458047100001DEST_PATH_IMAGE044
The priority size of expression inlet affairs is the priority of global transaction;
Step 4-6 according to the priority that step 4-5 calculates, resequences to priority query;
Step 4-7, kernel is carried out affairs according to priority query, and local transaction scheduler monitors executing state and feeds back to the global transaction scheduler; If run succeeded, then write down execution result and execution concluding time; If carry out failure, then return error code and carry out necessary rolling back action;
Step 4-8 judges with each affairs of layer whether all be finished in each nuclear; If not then wait for;
Step 4-9 is finished with layer, then changes lower floor over to, returns step 4-5, up to carrying out the outlet affairs, promptly should the part affairs be finished;
Step 4-10 returns to the global transaction management system with local transaction execution results, carries out the result for it and gathers and return to application server.
2. the transaction scheduling method of heterogeneous distributed real-time system according to claim 1 is characterized in that, in described step 4-5, for the high Real-time Transaction of urgency level, can directly give its priority is rank 0
CN2011101458047A 2011-06-01 2011-06-01 Transaction scheduling method of heterogeneous distributed real-time system Pending CN102207883A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2011101458047A CN102207883A (en) 2011-06-01 2011-06-01 Transaction scheduling method of heterogeneous distributed real-time system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2011101458047A CN102207883A (en) 2011-06-01 2011-06-01 Transaction scheduling method of heterogeneous distributed real-time system

Publications (1)

Publication Number Publication Date
CN102207883A true CN102207883A (en) 2011-10-05

Family

ID=44696731

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2011101458047A Pending CN102207883A (en) 2011-06-01 2011-06-01 Transaction scheduling method of heterogeneous distributed real-time system

Country Status (1)

Country Link
CN (1) CN102207883A (en)

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103377032A (en) * 2012-04-11 2013-10-30 浙江大学 Fine granularity scientific computation parallel processing device on basis of heterogenous multi-core chip
CN103699455A (en) * 2013-12-27 2014-04-02 重庆大学 Method for acquiring accurate latest deadline in multi-core realtime fault-tolerant system
WO2015120603A1 (en) * 2014-02-13 2015-08-20 Sap Ag Database calculation using parallel-computation in directed acyclic graph
CN103647834B (en) * 2013-12-16 2017-03-22 上海证券交易所 System and method used for processing multi-phase distributed task scheduling
WO2017107810A1 (en) * 2015-12-21 2017-06-29 阿里巴巴集团控股有限公司 Database operating method and device
CN107241281A (en) * 2017-05-27 2017-10-10 上海东土远景工业科技有限公司 A kind of data processing method and its device
CN108255613A (en) * 2018-02-07 2018-07-06 北京航空航天大学 A kind of SOA system resource management methods based on graph coloring
CN108415757A (en) * 2018-02-02 2018-08-17 阿里巴巴集团控股有限公司 distributed transaction processing method and device
CN109388481A (en) * 2018-09-21 2019-02-26 网易(杭州)网络有限公司 A kind of transmission method of transaction information, device, calculates equipment and medium at system
WO2022150995A1 (en) * 2021-01-13 2022-07-21 王志平 Supercomputer architecture implementation method
CN118349319A (en) * 2024-06-18 2024-07-16 华能信息技术有限公司 Distributed transaction management method and system

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020194248A1 (en) * 2001-05-01 2002-12-19 The Regents Of The University Of California Dedicated heterogeneous node scheduling including backfill scheduling
CN101692208A (en) * 2009-10-15 2010-04-07 北京交通大学 Task scheduling method and task scheduling system for processing real-time traffic information
CN101770402A (en) * 2008-12-29 2010-07-07 中国移动通信集团公司 Map task scheduling method, equipment and system in MapReduce system
CN101339521B (en) * 2008-07-28 2011-04-20 华中科技大学 Tasks priority dynamic dispatching algorithm
CN102063336A (en) * 2011-01-12 2011-05-18 国网电力科学研究院 Distributed computing multiple application function asynchronous concurrent scheduling method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020194248A1 (en) * 2001-05-01 2002-12-19 The Regents Of The University Of California Dedicated heterogeneous node scheduling including backfill scheduling
CN101339521B (en) * 2008-07-28 2011-04-20 华中科技大学 Tasks priority dynamic dispatching algorithm
CN101770402A (en) * 2008-12-29 2010-07-07 中国移动通信集团公司 Map task scheduling method, equipment and system in MapReduce system
CN101692208A (en) * 2009-10-15 2010-04-07 北京交通大学 Task scheduling method and task scheduling system for processing real-time traffic information
CN102063336A (en) * 2011-01-12 2011-05-18 国网电力科学研究院 Distributed computing multiple application function asynchronous concurrent scheduling method

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103377032A (en) * 2012-04-11 2013-10-30 浙江大学 Fine granularity scientific computation parallel processing device on basis of heterogenous multi-core chip
CN103647834B (en) * 2013-12-16 2017-03-22 上海证券交易所 System and method used for processing multi-phase distributed task scheduling
CN103699455A (en) * 2013-12-27 2014-04-02 重庆大学 Method for acquiring accurate latest deadline in multi-core realtime fault-tolerant system
CN103699455B (en) * 2013-12-27 2016-05-25 重庆大学 In a kind of multinuclear real-time fault tolerance system, obtain accurately the method for deadline the latest
WO2015120603A1 (en) * 2014-02-13 2015-08-20 Sap Ag Database calculation using parallel-computation in directed acyclic graph
US9576072B2 (en) 2014-02-13 2017-02-21 Sap Se Database calculation using parallel-computation in a directed acyclic graph
WO2017107810A1 (en) * 2015-12-21 2017-06-29 阿里巴巴集团控股有限公司 Database operating method and device
CN107241281B (en) * 2017-05-27 2020-01-14 上海东土远景工业科技有限公司 Data processing method and device
CN107241281A (en) * 2017-05-27 2017-10-10 上海东土远景工业科技有限公司 A kind of data processing method and its device
CN108415757B (en) * 2018-02-02 2020-10-30 创新先进技术有限公司 Distributed transaction processing method and device
CN108415757A (en) * 2018-02-02 2018-08-17 阿里巴巴集团控股有限公司 distributed transaction processing method and device
CN108255613A (en) * 2018-02-07 2018-07-06 北京航空航天大学 A kind of SOA system resource management methods based on graph coloring
CN108255613B (en) * 2018-02-07 2021-01-01 北京航空航天大学 SOA system resource management method based on graph coloring
CN109388481A (en) * 2018-09-21 2019-02-26 网易(杭州)网络有限公司 A kind of transmission method of transaction information, device, calculates equipment and medium at system
CN109388481B (en) * 2018-09-21 2021-08-17 网易(杭州)网络有限公司 Transaction information transmission method, system, device, computing equipment and medium
WO2022150995A1 (en) * 2021-01-13 2022-07-21 王志平 Supercomputer architecture implementation method
CN118349319A (en) * 2024-06-18 2024-07-16 华能信息技术有限公司 Distributed transaction management method and system

Similar Documents

Publication Publication Date Title
CN102207883A (en) Transaction scheduling method of heterogeneous distributed real-time system
CN110737529B (en) Short-time multi-variable-size data job cluster scheduling adaptive configuration method
CN110297699B (en) Scheduling method, scheduler, storage medium and system
Liu et al. Resource preprocessing and optimal task scheduling in cloud computing environments
US8473951B2 (en) Method and system for traversing in reverse chronological order along a critical path of a plurality of jobs, and reducing time gaps between jobs until an estimated end time of the last job is less than or equal to a target end time
CN103309946B (en) Multimedia file processing method, Apparatus and system
US20190377601A1 (en) File Operation Task Optimization
US20200219028A1 (en) Systems, methods, and media for distributing database queries across a metered virtual network
CN103838621B (en) Method and system for scheduling routine work and scheduling nodes
CN102254246A (en) Workflow managing method and system
CN105373426B (en) A kind of car networking memory aware real time job dispatching method based on Hadoop
CN110928651B (en) Service workflow fault-tolerant scheduling method under mobile edge environment
CN103927225A (en) Multi-core framework Internet information processing and optimizing method
CN111782627B (en) Task and data cooperative scheduling method for wide-area high-performance computing environment
US20210390405A1 (en) Microservice-based training systems in heterogeneous graphic processor unit (gpu) cluster and operating method thereof
CN109960573A (en) A kind of cross-domain calculating task dispatching method and system based on Intellisense
CN106095582A (en) The task executing method of cloud platform
CN111913800B (en) Resource allocation method for optimizing cost of micro-service in cloud based on L-ACO
CN112306642B (en) Workflow scheduling method based on stable matching game theory
CN108833294B (en) Low-bandwidth-overhead flow scheduling method for data center wide area network
He et al. A two-stage scheduling method for deadline-constrained task in cloud computing
CN102760073A (en) Method, system and device for scheduling task
EP2541408B1 (en) Method and system for processing data for database modification
CN109976873A (en) The scheduling scheme acquisition methods and dispatching method of containerization distributed computing framework
CN116501828B (en) Non-perception vector query method and system for server based on unstructured data set

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20111005