CN104184816A - Lookahead dynamic adjustment method based on simulation member event timestamp increment expectation - Google Patents
Lookahead dynamic adjustment method based on simulation member event timestamp increment expectation Download PDFInfo
- Publication number
- CN104184816A CN104184816A CN201410431715.2A CN201410431715A CN104184816A CN 104184816 A CN104184816 A CN 104184816A CN 201410431715 A CN201410431715 A CN 201410431715A CN 104184816 A CN104184816 A CN 104184816A
- Authority
- CN
- China
- Prior art keywords
- prediction amount
- time
- lookahead
- simulation
- adjust
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
The invention belongs to the field of distributed system simulation, and particularly relates to a lookahead dynamic adjustment method based on simulation member event timestamp increment expectation. The method includes the steps that (1) according to timestamp increment change situations that federal members generate events in a time interval, a lookahead adjustment scaling factor is determined, and whether lookahead needs to be adjusted or not is determined according to the scaling factor; (2) if it is determined that the lookahead needs to be adjusted, the lookahead is reasonably adjusted according to the HLA rule and the definition of the lookahead by an algorithm. Segmenting and grouping management is conducted on simulation time, the lookahead adjustment scaling factor is updated according to the simulation state, whether the lookahead needs to be adjusted or not is determined by means of the scaling factor, and finally the lookahead is reasonably updated according to constraint conditions for adjustment of the lookahead through the HLA. According to the method, artificial setting of the lookahead is abandoned, the lookahead can be better combined with the simulation state, simulation efficiency is improved, and the risk that deadlock occurs is lowered.
Description
Technical field
The invention belongs to Distribute d system simulation field, be specifically related to a kind of prediction amount dynamic adjusting method of expecting based on simulated members event time stamp increment.
Background technology
Prediction amount is to produce in to the research process of the conservative algorithm of parallel discrete event simulation.The basic thought of this algorithm is: analogue system has realizability and predictable, is meeting under local cause and effect constraints simultaneously, and system can complete correct simulation process.Because the binding character of conservative algorithm is very strong, need the moment to guarantee the front and back order of director's part of living in, so this algorithm is easy to produce of deadlock.
1978, Chandy and Misra used first prediction amount in the Chandy/Misra algorithm proposing.This algorithm is to utilize null message to solve Deadlock, logical process is by sending the only null message with time stamp, notify the minimum time stamp of its transmission message of other processes, other processes judge by this time stamp and cancel unnecessary wait, thereby remove the Deadlock causing because strictly observing local restriction condition.Null message method, by informing in advance the message time stamp lower limit of other these processes of process, is avoided the generation of Deadlock, and its essence is exactly a kind of thought of the amount of prediction.
In HLA specification, do not provide specific definition and the description of prediction amount.In specification, mention " the prediction amount of each federal member in analogue system, is determined according to own characteristic by each federal member program ".But Fujimoto has provided 5 suggestions that arrange for prediction amount size in the literature:
(1) can be set to prediction amount to the response time of incoming event by federal member.
(2) federal members send an event that can affect another federal member, and prediction amount can be set to the transmission time of this event in network.
(3), in the time cannot determining the time of origin of event, prediction amount can be set to the contingent time range of this event.
(4) the federal member prediction amount conventionally advancing based on step-length is set to time step.
(5) the non-property seized of event.A federal member calculates and advances according to the business model of self, and within the ensuing L time, not affecting other members is not also affected by other members, now can prediction amount be set to L.
The effect of prediction amount mainly contains 2 points: (1) improves the concurrency of analogue system; (2) avoid the generation of deadlock.
When prediction amount hour, the synchronism of analogue system is higher; In the time that prediction amount is larger, the concurrency of analogue system is higher.
The introducing of prediction amount is for avoiding deadlock to have important effect.Before introducing prediction amount,, federal member, because ask each other the restriction of promotion logic time, causes analogue system often to there will be deadlock.Prediction amount is a key concept in distributed time management, the event message timestamp of generation is informed in advance other members by federal member utilization prediction amount, other members, according to the promise in this pair of future, reasonably arrange the time of self to advance, and have accelerated like this operation of analogue system.
Conservative Time Advance Mechanism, ensure that analogue system adopts TSO order while processing federal member event, prediction amount is to guarantee the key of its trouble-free operation, each time-controllable type federal member is issued the prediction amount of self to RTI, RTI calculates the minimum time stamp of all events according to member's prediction amount and current logical time.
Summary of the invention
The object of the present invention is to provide a kind of prediction amount dynamic adjusting method of expecting based on simulated members event time stamp increment of the risk that reduces produce of deadlock.
The object of the present invention is achieved like this:
(1) the timestamp increment situation of change in time interval generation event according to federal member, determines prediction amount adjustment proportional factor, and determines whether to adjust prediction amount according to this scale factor;
(2) need to adjust prediction amount if determine, the definition according to this algorithm to prediction amount, how rationally to adjust prediction amount according to HLA rule:
(2.1) segmentation is carried out in simulation time interval according to the rules, point good after again since second time period by time period grouping, the time slice that comprises fixed number in each time grouping;
(2.2) set prediction amount initial value, this value can be used at the emulation initial stage;
(2.3) after simulation process starts, record the time time stamp increment of federal member in each time period, this increment is designated as prediction amount adjusted value, calculates the prediction amount adjustment proportional factor of this grouping in the first time period of each time grouping later simultaneously;
(2.4) grouping of each time, according to the scale factor in this group and member's event time stamp increment situation in each time period, judges whether to adjust the prediction amount size in the time period;
(2.5) if through judgement, need to adjust prediction amount, adjust strategy according to prediction amount in HLA, dynamically adjust the prediction amount size in the time period.
Beneficial effect of the present invention is: the method is by simulation time segmentation, grouping management, upgrade according to simulation status, prediction amount adjustment proportional factor, judges whether to adjust prediction amount in the proportion of utilization factor, finally the constraints to adjustment prediction amount according to HLA, reasonably upgrades prediction amount.The method has been abandoned the artificial setting to prediction amount, and prediction amount can be combined with simulation status better, and simulation efficiency is provided, and has reduced the risk of produce of deadlock.
Brief description of the drawings
Whole flow chart in Fig. 1 the present invention;
In Fig. 2 the present invention, prediction amount is adjusted dynamic adjusting method flow chart;
The process instance of in Fig. 3 the present invention, prediction amount being adjusted.
Embodiment
Below in conjunction with accompanying drawing, the present invention is described further.
The invention provides a kind of prediction amount dynamic adjusting method of expecting based on simulated members event time stamp increment.By analyzing correlation and the Time Advance Mechanism of federal member in analogue system, in conjunction with the related algorithm of maximum utilogic time GALT, the event time stamp producing from federal member is started with, and has proposed this kind of method of adjustment simultaneously.The method is passed through simulation time sectional management, and in each time period, calculate respectively prediction amount adjustment proportional factor, judge whether prediction amount to adjust according to this scale factor, in section computing time, the time stamp increment of member's event is expected simultaneously, and this expectation is prediction amount adjusted value.The method is along with the propelling of simulation process, prediction amount adjustment proportional factor, prediction amount size have been done to real-time update, make prediction amount more meet the real-time status of simulation process, shortened the computing time of GALT, the releasing of emulation deadlock is played an important role.
Prediction amount dynamic adjusting method provided by the invention, its step comprises:
Step 1. is simulation time space segmentation according to the rules, to the sectional management of simulation time, and the situation that the fluctuation of the effective balance time stamp of which increment is larger.After point good again since second time period by time period grouping, the time slice that comprises fixed number in each time grouping.Utilize the grouping management to the time period, can be according to the running status in simulation process, adjust dynamically the prediction amount adjustment proportional factor in each grouping, be that scale factor changes along with the variation of simulation status.
Step 2. is set prediction amount initial value, and this value can be used at the emulation initial stage.
Step 3., after simulation process starts, calculates the time time stamp increment of federal member in lower each time period.From analyzing, if event time stamp increment changing value is far longer than prediction amount set point, federal member need to repeatedly calculate and could process this event GALT; If event time stamp increment is less than prediction amount settings, there will be logic error, now need to readjust prediction amount, to ensure the logical correctness of federal operation.Therefore this increment is designated as prediction amount adjusted value, calculate the prediction amount adjustment proportional factor of this grouping later in the first time period of each time grouping simultaneously.Utilize the scale factor in time grouping, judge in group whether need to adjust prediction amount size in each time period.
Step 4. according to the relation of scale factor in this ratio and group, judges whether to adjust the prediction amount size in the time period compared with the event time stamp increment in the each time period in time grouping is expected to expect with event time stamp increment in the first time period.
If step 5. through judgement, needs to adjust prediction amount, adjust strategy according to prediction amount in HLA, dynamically adjust the prediction amount size in the time period.
In the present invention, most critical is that prediction is measured to big or small definition, impact according to prediction amount on the maximum utilogic time, prediction amount size using the desired value of event time stamp increment in each time period in the time period, in the time that needs are adjusted prediction amount, the prediction amount in the time period is updated to event time stamp increment and expects.
The invention provides a kind of prediction amount dynamic adjusting method of expecting based on simulated members event time stamp increment.The object of the invention is to solve the operation Deadlock that may exist in the distributing emulation system based on HLA.By analyzing the effect of prediction amount in maximum utilogic Time Calculation problem, and the constraint to adjustment prediction amount in conjunction with HLA time management, prediction amount dynamic adjusting method has been proposed.The method, by simulation time segmentation, grouping management, is upgraded according to simulation status, and prediction amount adjustment proportional factor, judges whether to adjust prediction amount in the proportion of utilization factor, and finally the constraints to adjustment prediction amount according to HLA, reasonably upgrades prediction amount.The method has been abandoned the artificial setting to prediction amount, and prediction amount can be combined with simulation status better, and simulation efficiency is provided, and has reduced the risk of produce of deadlock.
By analyzing the maximum utilogic time GALT algorithm of federal member, find reasonably to arrange number of times and time that prediction amount can effectively reduce federal member and calculates GALT, improve simulation efficiency, also reduced the risk that produces emulation deadlock simultaneously.The present invention proposes a kind of prediction amount dynamic adjusting method of expecting based on simulated members event time stamp increment, Fig. 1 has provided the realization flow figure of whole method, comprise from starting being fragmented into according to HLA specification of simulation time rationally adjusted prediction amount, Fig. 2 has provided the example of prediction amount adjustment method.
Below in conjunction with accompanying drawing 1 and embodiment, the present invention will be further described.
Step 1. is simulation time space segmentation according to the rules, point good after again since second time period by time period grouping, the time slice that comprises fixed number in each time grouping.
Step 2. is set prediction amount initial value, and this value can be used at the emulation initial stage.
Step 3., after simulation process starts, is recorded the time time stamp increment of federal member in each time period, and this increment is designated as prediction amount adjusted value, calculates the prediction amount adjustment proportional factor of this grouping in the first time period of each time grouping later simultaneously.
The grouping of each time of step 4., according to the scale factor in this group and member's event time stamp increment situation in each time period, judges whether to adjust the prediction amount size in the time period.In detail as shown in Figure 2, Rule of judgment is divided into three kinds of situations, and every kind of situation is all different for the processing of prediction amount.
If step 5. through judgement, needs to adjust prediction amount, adjust strategy according to prediction amount in HLA, dynamically adjust the prediction amount size in the time period.
Step 6. judges whether emulation finishes, the adjustment of prediction amount until emulation finish.
(1) for the sectional management of simulation time, the situation that the fluctuation of the effective balance time stamp of which increment is larger.Utilize the grouping management to the time period simultaneously, can be according to the running status in simulation process, adjust dynamically the prediction amount adjustment proportional factor in each grouping, be that scale factor changes along with the variation of simulation status.
(2) utilize the scale factor in time grouping, judge in group whether need to adjust prediction amount size in each time period.
(3) the most important thing is, the present invention is the impact on the maximum utilogic time according to prediction amount, prediction amount size using the desired value of event time stamp increment in each time period in the time period, in the time that needs are adjusted prediction amount, the prediction amount in the time period is updated to event time stamp increment and expects
Shown in Fig. 3, in the time that prediction quantitative change is large, RTI can agree to this request immediately, in the time that prediction amount need diminish, RTI can't upgrade prediction amount immediately, but dwindles gradually the value of prediction amount, has reached renewal value, this kind of update mode is in order to ensure that federal member can not receive out-of-date message, and causes emulation logic mistake.
Claims (1)
1. a prediction amount dynamic adjusting method of expecting based on simulated members event time stamp increment, is characterized in that:
(1) be the timestamp increment situation of change that produces event according to federal member at time interval, determine prediction amount adjustment proportional factor, and determine whether to adjust prediction amount according to this scale factor;
(2) need to adjust prediction amount if determine, the definition according to this algorithm to prediction amount, how rationally to adjust prediction amount according to HLA rule:
(2.1) segmentation is carried out in simulation time interval according to the rules, point good after again since second time period by time period grouping, the time slice that comprises fixed number in each time grouping;
(2.2) set prediction amount initial value, this value can be used at the emulation initial stage;
(2.3) after simulation process starts, record the time time stamp increment of federal member in each time period, this increment is designated as prediction amount adjusted value, calculates the prediction amount adjustment proportional factor of this grouping in the first time period of each time grouping later simultaneously;
(2.4) grouping of each time, according to the scale factor in this group and member's event time stamp increment situation in each time period, judges whether to adjust the prediction amount size in the time period;
(2.5) if through judgement, need to adjust prediction amount, adjust strategy according to prediction amount in HLA, dynamically adjust the prediction amount size in the time period.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410431715.2A CN104184816B (en) | 2014-08-28 | 2014-08-28 | One kind is based on the desired prediction amount dynamic adjusting method of simulated members event time stamp increment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410431715.2A CN104184816B (en) | 2014-08-28 | 2014-08-28 | One kind is based on the desired prediction amount dynamic adjusting method of simulated members event time stamp increment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104184816A true CN104184816A (en) | 2014-12-03 |
| CN104184816B CN104184816B (en) | 2018-04-24 |
Family
ID=51965545
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410431715.2A Active CN104184816B (en) | 2014-08-28 | 2014-08-28 | One kind is based on the desired prediction amount dynamic adjusting method of simulated members event time stamp increment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104184816B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108959098A (en) * | 2018-07-20 | 2018-12-07 | 大连理工大学 | A kind of test macro and method for distributed system program deadlock defect |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6044456A (en) * | 1998-01-05 | 2000-03-28 | Intel Corporation | Electronic system and method for maintaining synchronization of multiple front-end pipelines |
| US20080091928A1 (en) * | 2004-12-17 | 2008-04-17 | Eickemeyer Richard J | Branch lookahead prefetch for microprocessors |
| CN101442555A (en) * | 2008-12-19 | 2009-05-27 | 中国运载火箭技术研究院 | Artificial resource proxy service system facing HLA |
| CN102591759A (en) * | 2011-12-29 | 2012-07-18 | 中国科学技术大学苏州研究院 | Clock precision parallel simulation system for on-chip multi-core processor |
| CN103763377A (en) * | 2014-01-24 | 2014-04-30 | 北京仿真中心 | Synchronizing method for distributed simulation time advance and information transmission |
-
2014
- 2014-08-28 CN CN201410431715.2A patent/CN104184816B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6044456A (en) * | 1998-01-05 | 2000-03-28 | Intel Corporation | Electronic system and method for maintaining synchronization of multiple front-end pipelines |
| US20080091928A1 (en) * | 2004-12-17 | 2008-04-17 | Eickemeyer Richard J | Branch lookahead prefetch for microprocessors |
| CN101442555A (en) * | 2008-12-19 | 2009-05-27 | 中国运载火箭技术研究院 | Artificial resource proxy service system facing HLA |
| CN102591759A (en) * | 2011-12-29 | 2012-07-18 | 中国科学技术大学苏州研究院 | Clock precision parallel simulation system for on-chip multi-core processor |
| CN103763377A (en) * | 2014-01-24 | 2014-04-30 | 北京仿真中心 | Synchronizing method for distributed simulation time advance and information transmission |
Non-Patent Citations (1)
| Title |
|---|
| 王召福等: "《HLA 仿真系统中Lookahead 的分析与动态调整策略》", 《计算机仿真》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108959098A (en) * | 2018-07-20 | 2018-12-07 | 大连理工大学 | A kind of test macro and method for distributed system program deadlock defect |
| CN108959098B (en) * | 2018-07-20 | 2021-11-05 | 大连理工大学 | System and method for testing deadlock defects of distributed system program |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104184816B (en) | 2018-04-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104714852B (en) | A kind of parameter synchronization optimization method and its system suitable for distributed machines study | |
| CN106297774B (en) | A kind of the distributed parallel training method and system of neural network acoustic model | |
| CN111079921A (en) | Efficient neural network training and scheduling method based on heterogeneous distributed system | |
| CN109492753A (en) | A kind of method of the stochastic gradient descent of decentralization | |
| CN107330516A (en) | Model parameter training method, apparatus and system | |
| CN110990155B (en) | Parameter communication method for large-scale safety monitoring | |
| CN111191728B (en) | Deep reinforcement learning distributed training method and system based on asynchronization or synchronization | |
| CN110046048B (en) | Load balancing method based on workload self-adaptive fast redistribution | |
| US9619290B2 (en) | Hardware and runtime coordinated load balancing for parallel applications | |
| CN103561055B (en) | Web application automatic elastic extended method under conversation-based cloud computing environment | |
| CN111027708A (en) | Distributed machine learning-oriented parameter communication optimization method | |
| CN113157413B (en) | Deep learning task resource optimization configuration method and system based on service quality requirement | |
| WO2018010381A1 (en) | Multi-target elevator group control system and method | |
| CN104346214B (en) | Asynchronous task managing device and method for distributed environment | |
| CN104133857A (en) | New method for digging business process model on the basis of configuration constraint | |
| US20190155658A1 (en) | Managing processing system efficiency | |
| CN104184816A (en) | Lookahead dynamic adjustment method based on simulation member event timestamp increment expectation | |
| CN105824687B (en) | A kind of method and device of Java Virtual Machine performance automated tuning | |
| CN103957229A (en) | Active updating method, device and server for physical machines in IaaS cloud system | |
| CN105335537A (en) | Estimation method and system of network media information exposure in video album | |
| Yu et al. | Accelerating distributed training in heterogeneous clusters via a straggler-aware parameter server | |
| CN116542178A (en) | Constant flow numerical simulation method, device, equipment and medium | |
| CN103645952B (en) | A kind of non-precision tasks in parallel processing method based on MapReduce | |
| CN106506594B (en) | A kind of concurrent computation resource distribution method and device | |
| CN105721199B (en) | A kind of real-time cloud service bottleneck detection method based on Density Estimator and fuzzy inference system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |