CN110362457A - A kind of three value optical computer method of evaluating performance and system based on Vacation queue - Google Patents
A kind of three value optical computer method of evaluating performance and system based on Vacation queue Download PDFInfo
- Publication number
- CN110362457A CN110362457A CN201910378982.0A CN201910378982A CN110362457A CN 110362457 A CN110362457 A CN 110362457A CN 201910378982 A CN201910378982 A CN 201910378982A CN 110362457 A CN110362457 A CN 110362457A
- Authority
- CN
- China
- Prior art keywords
- average
- stage
- queue
- request
- sas
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3409—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment for performance assessment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3447—Performance evaluation by modeling
-
- 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)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Computer And Data Communications (AREA)
Abstract
The invention discloses a kind of three value optical computer method of evaluating performance and system based on Vacation queue.This method comprises: obtaining the service model based on serial and Vacation queue three value optical computers to be evaluated;The service model is divided into four-stage;Calculate the average response time and average request number during each stage running;Establish Evaluating Models;The Evaluating Models are used to calculate the average response time and average request number of three value optical computers;The average response time and average request number of the model to be serviced are calculated by the average response time and average request number during each stage running according to the Evaluating Models;According to the average response time of the service model and average request number, performance evaluation is carried out to the three value optical computers to be evaluated.This method or system can accurately evaluate the performance of three value optical computers.
Description
Technical field
The present invention relates to computer performance evaluation fields, based on a kind of three value optics by Vacation queue
Calculation machine method of evaluating performance and system.
Background technique
Due to the particularity of TOC optical processor, such as huge position property, reconfigurability, step-by-step allocability, these performances with
Assay model is all not directly adaptable to use TOC especially for cloud computing system.
TOC performance evaluation is modeled based on M/M/1 queuing system in the prior art, it is found that calculation amount and network pass
Defeated speed is the bottleneck for influencing system response time.But the model still cannot more really reflect the calculating ecology of TOC.Such as it should
It will not break down after model hypothesis TOC self-starting, that is, not need maintenance and breakdown maintenance, run always.Obviously, this is
A kind of ideal state.
Summary of the invention
In view of the above-mentioned problems, the present invention provides a kind of three value optical computer performance evaluation sides based on Vacation queue
Method and system.
To achieve the above object, the present invention provides following schemes:
A kind of three value optical computer method of evaluating performance based on Vacation queue, which comprises
Obtain the service model based on serial and Vacation queue three value optical computers to be evaluated;
The service model is divided into four-stage;
Calculate the average response time and average request number during each stage running;
Establish Evaluating Models;When the Evaluating Models are used to calculate the average response of three value optical computers
Between and average request number;
According to the Evaluating Models, pass through the average response time and average request during each stage running
Number calculates the average response time and average request number of the model to be serviced;
According to the average response time of the service model and average request number, to the three value optics to be evaluated
Computer carries out performance evaluation.
Optionally, the four-stage includes first stage, second stage, phase III and fourth stage;
In the first stage, server is received by not received operation request and presses FCFS strategy the first layer queue of deposit,
And a request is taken out from non empty request queue, the request taken out from non empty request queue is sent to data prediction
The operation received request is converted into TOC task to be processed by server;
In the second stage, the decimal data that user inputs is converted into MSD data by data prediction server,
The control signal that optical processor is generated according to operation, sends control signals to SAS;
In the phase III, SAS will receive task first and be inserted into third time queue by FCFS strategy, then first
It checks OP whether on holiday state, if OP is in state of having a holiday, waits;Otherwise check whether that there are idle processors;If no
In the presence of then continuing waiting for;Otherwise, according to scheduling strategy scheduler task;The RU of OP is according to distribution information and restructuring directive step-by-step weight
Structure arithmetic unit, then obtains data, and OP carries out concurrent operation using control signal, the operation result that DC indicates optical signal into
Row decoding, and decoded data are sent to operation result and send server TS;While OP carries out operation, TOC judgement
Whether the data of unfinished operation are had, if so, then continuing data;Otherwise, TOC sends " operation completion signal " to SAS;
SAS checks whether unscheduled task after receiving signal;If so, then SAS again scheduler task, distribution resource;Otherwise,
SAS sends " no request " signal to TOC, and TOC starts having a holiday for a random-length;At the end of having a holiday, TOC is sent " not to SAS
False knot beam " signal, SAS checks whether unscheduled task after receiving signal again, and handles it;
In the fourth stage, the operation result indicated with MSD data that TS sends over TOC is added by FCFS strategy
Enter the 4th layer queue, obtains result queue, and successively take out user's operation result from non-empty result queue, or according to need
Operation result is converted into decimal data, is sent to relative users.
Optionally, the average response time and average request number in first stage operational process are calculated, is specifically included:
Obtain the average arrival rate, RS in the unit time that operation calculation is requested in the unit time in first stage operational process
Average receptance;
According to the average arrival rate and the average receptance, the average request number of the first stage is calculated;
According to the average request number of the first stage and the average arrival rate, the flat of the first stage is calculated
The equal response time.
A kind of three value optical computer Performance Appraisal Systems based on Vacation queue, the system comprises:
Model obtains module, for obtaining the clothes based on serial and Vacation queue three value optical computers to be evaluated
Business model;
Division module, for the service model to be divided into four-stage;
First computing module, for calculating average response time and average request number during each stage running;
Modeling module, for establishing Evaluating Models;The Evaluating Models are for calculating three value optical computers
Average response time and average request number;
Second computing module, for passing through the average response during each stage running according to the Evaluating Models
Time and average request number calculate the average response time and average request number of the model to be serviced;
Evaluation module, for the average response time and average request number according to the service model, to described to be evaluated
Three value optical computers of valence carry out performance evaluation.
Optionally, the four-stage includes first stage, second stage, phase III and fourth stage;
In the first stage, server is received by not received operation request and presses FCFS strategy the first layer queue of deposit,
And a request is taken out from non empty request queue, the request taken out from non empty request queue is sent to data prediction
The operation received request is converted into TOC task to be processed by server;
In the second stage, the decimal data that user inputs is converted into MSD data by data prediction server,
The control signal that optical processor is generated according to operation, sends control signals to SAS;
In the phase III, SAS will receive task first and be inserted into third time queue by FCFS strategy, then first
It checks OP whether on holiday state, if OP is in state of having a holiday, waits;Otherwise check whether that there are idle processors;If no
In the presence of then continuing waiting for;Otherwise, according to scheduling strategy scheduler task;The RU of OP is according to distribution information and restructuring directive step-by-step weight
Structure arithmetic unit, then obtains data, and OP carries out concurrent operation using control signal, the operation result that DC indicates optical signal into
Row decoding, and decoded data are sent to operation result and send server TS;While OP carries out operation, TOC judgement
Whether the data of unfinished operation are had, if so, then continuing data;Otherwise, TOC sends " operation completion signal " to SAS;
SAS checks whether unscheduled task after receiving signal;If so, then SAS again scheduler task, distribution resource;Otherwise,
SAS sends " no request " signal to TOC, and TOC starts having a holiday for a random-length;At the end of having a holiday, TOC is sent " not to SAS
False knot beam " signal, SAS checks whether unscheduled task after receiving signal again, and handles it;
In the fourth stage, the operation result indicated with MSD data that TS sends over TOC is added by FCFS strategy
Enter the 4th layer queue, obtains result queue, and successively take out user's operation result from non-empty result queue, or according to need
Operation result is converted into decimal data, is sent to relative users.
Optionally, the average response time and average request number in first stage operational process are calculated, is specifically included:
Acquiring unit, calculated for obtaining in first stage operational process operation in the unit time request average arrival rate,
The average receptance of RS in unit time;
First computing unit, for calculating first rank according to the average arrival rate and the average receptance
The average request number of section;
Second computing unit, for according to the first stage average request number and the average arrival rate, calculate
The average response time of the first stage.
Compared with prior art, the present invention has following technical effect that the present invention calculated for three value optical computer each stages
Average response time and average request number in operational process;According to the Evaluating Models, pass through each stage running mistake
Average response time and average request number in journey calculate the average response time of the model to be serviced and averagely ask
Ask several;According to the average response time of the service model and average request number, to the three value optical computings to be evaluated
Machine carries out performance evaluation.The performance of three value optical computers can be accurately evaluated by the above method.
Detailed description of the invention
It in order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will be in embodiment
Required attached drawing is briefly described, it should be apparent that, the accompanying drawings in the following description is only some realities of the invention
Example is applied, it for those of ordinary skill in the art, without any creative labor, can also be according to these
Attached drawing obtains other attached drawings.
Fig. 1 is the flow chart of three value optical computer method of evaluating performance of the embodiment of the present invention based on Vacation queue;
Fig. 2 is the phase III work flow diagram that band of the embodiment of the present invention is had a holiday;
Fig. 3 is the state transition diagram of RS team leader of the embodiment of the present invention based on M/M/1 queuing system;
Fig. 4 is that processor of the embodiment of the present invention divides equally the lower simulated AC curve state transition diagram of strategy;
Fig. 5 is the structural block diagram of three value optical computer Performance Appraisal Systems of the embodiment of the present invention based on Vacation queue.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts it is all its
His embodiment, shall fall within the protection scope of the present invention.
The object of the present invention is to provide a kind of three value optical computer method of evaluating performance based on Vacation queue and it is
System, the performance to accurate evaluation computer.
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
As shown in Figure 1, a kind of three value optical computer method of evaluating performance based on Vacation queue include:
Step 101: obtaining the service model based on serial and Vacation queue three value optical computers to be evaluated.
Step 102: the service model is divided into four-stage.
The four-stage includes first stage, second stage, phase III and fourth stage;
In the first stage, server is received by not received operation request and presses FCFS strategy the first layer queue of deposit,
And a request is taken out from non empty request queue, the request taken out from non empty request queue is sent to data prediction
The operation received request is converted into TOC task to be processed by server;
In the second stage, the decimal data that user inputs is converted into MSD data by data prediction server,
The control signal that optical processor is generated according to operation, sends control signals to SAS;
As shown in Fig. 2, SAS will receive task first and be inserted into third time team by FCFS strategy in the phase III
Whether on holiday column, then first look at OP state, if OP is in state of having a holiday, waits;Otherwise check whether there is the free time
Processor;If it does not exist, then it continues waiting for;Otherwise, according to scheduling strategy scheduler task;The RU of OP is according to distribution information and again
Structure instructs step-by-step restructing operation device, then obtains data, and OP carries out concurrent operation using control signal, and DC indicates optical signal
Operation result be decoded, and by decoded data be sent to operation result send server TS;Operation is carried out in OP
Meanwhile TOC judges whether there is the data of unfinished operation, if so, then continuing data;Otherwise, TOC sends " operation to SAS
Complete signal ";SAS checks whether unscheduled task after receiving signal;If so, then SAS again scheduler task, distribution
Resource;Otherwise, SAS sends " no request " signal to TOC, and TOC starts having a holiday for a random-length;At the end of having a holiday, TOC to
SAS sends " end of having a holiday " signal, and SAS checks whether unscheduled task after receiving signal again, and handles it;
In the fourth stage, the operation result indicated with MSD data that TS sends over TOC is added by FCFS strategy
Enter the 4th layer queue, obtains result queue, and successively take out user's operation result from non-empty result queue, or according to need
Operation result is converted into decimal data, is sent to relative users.
Step 103: calculating the average response time and average request number during each stage running.
The average response time and average request number in first stage operational process are calculated, is specifically included: obtaining first
During stage running in the unit time operation calculate the average arrival rate of request, in the unit time RS average receptance;According to
The average arrival rate and the average receptance, calculate the average request number of the first stage;According to first rank
The average request number of section and the average arrival rate, calculate the average response time of the first stage.
The present invention chooses average response time, average number of tasks, handling capacity and TOC utilization rate etc. and refers to as performance evaluation
Mark.Initially set up the mathematical model for calculating average number of tasks R and response time T
Wherein RiAnd Ti(i=1,2,3,4) the average request number and average clothes in service model in each stage are respectively indicated
It is engaged in the time.
The performance evaluation of Stage 1
The Stage 1 of serial queuing model has a server for receiving the operation request of user's transmission, therefore optional
It is modeled with M/M/1 queuing system, i.e. independent same distribution is obeyed at the arrival time interval of hypothesis operation request and parameter is 1/
The exponential distribution of λ, receiving time obeys independent same distribution and parameter is 1/ μ1Exponential distribution.RS receives operation in Stage 1
Request the state transition diagram of Shi Qi team leader as shown in Figure 3.
Work as ρ1=λ/μ1When < 1, RS is up to stable state, and there are Stationary Distributions.Remember that m-th of shape probability of state is(m
=0,1,2 ...), by KShi algebraic equation, following equation group can be obtained:
It can obtain
In conjunction with regularityObtain the idle probability of RS
Then, stage average request number R1
By Little formula, the average response time T of request can must be received1
Wherein: λ is the average arrival rate of operation request in the unit time, μ1For the average received rate of RS in the unit time.
Assuming again that average amount to be transmitted is requested in operation is D, and the average transmission speed of network is ξ, then μ1=ξ/D.It is substituted into
(1) it is obtained with (2)
The performance evaluation of Stage 2
According to Burke theorem, the output of Stage1 is the Poisson mistake that a mean value is λ when system reaches equilibrium state
Journey.Namely the average arrival rate that request reaches Stage 2 in the unit time is also λ.The equally available M/M/1 of Stage 2 is lined up system
System expression.PPS is similar with Fig. 4 by the when neat CTMC model state transfer figure of FCFS policy service.Assuming that PPS progress data are pre-
The rate of processing is τ, then its service speed μ2=τ/D.WhenWhen, Stage 2 has identical as Stage 1
Probability of stability equation group.Therefore, the performance evaluation conclusion of Stage 1 can be directly applied to Stage 2 and obtain its average request
Number R2With average response time T2
The performance evaluation of Stage 3
Operation request is sent to TOC to after operation request scheduling by FCFS strategy by SAS, while being scheduled each fortune
Request distribution optical processor is calculated, and the reconstruct code of allocation result and institute's allocation processing device is sent to TOC.TOC optical treatment
After the reconstruction means of device OP complete reconstruct in a manner of full parellel, the data that encoder indicates control Internal Code are encoded, i.e., will
Electric signal is converted into optical signal, and then arithmetic unit just carries out optical oomputing to it, and last decoder D converts operation result with logical
Believe the data that Internal Code indicates.
Because optical processor OP has huge position property, such as the data bit for the TOC computing platform built for 2018 has reached
It 1152, can be with the multiple operation requests of parallel processing.The present invention is quasi- to be considered as having Multiple Vacations strategy and exhaustive service
M/M/n queuing system carries out performance analysis and evaluation to the stage.Obviously, request reaches being averaged for Stage 3 in the unit time
Arrival rate is also λ.Assuming that reach the time interval of Stage3, TOC vacation time, TOC busy period mutually indepedent for task.
Divide equally in processor using under strategy, optical processor is divided into n small optical processors.Obviously, this n
Small optical processor is isomorphism, i.e., the restructural optical processor of hardware configuration having the same --- identical digit, thus
Computing capability having the same.Accordingly it is contemplated that express Stage3 with classics M/M/n queuing system, successive vacation twice
Between time be the TOC busy period.After each small optical processor completion task, if not having pending task in SAS, and
Still there are other small optical processors in processing task, it cannot directly enter the state of having a holiday, and can only switch to idle state.Once
Queue Q in SAS becomes empty, and TOC, that is, n small optical processors are just synchronous to start a random-length having a holiday for v; TOC
Reaching during having a holiday for task will be sequentially inserted into the tail of the queue of Q;TOC certain have a holiday at the end of, all to SAS send " knot of having a holiday
Beam " signal such as receives " having a holiday " signal, then starts independent same distribution next time and have a holiday, otherwise will end a holiday, into the busy period;
If the queue Q of SAS has LQ (< n) a task at this time, according to all tasks in FCFS strategy scheduling Q, LQ small optical processors
It starts to work, remaining n-LQ small optical processors are in idle condition;If LQ >=n, equally according in FCFS strategy scheduling Q
N task, remaining n-LQ task continue to wait in line.
The vacation time v of TOC obeys the exponential distribution that parameter is 1/ δ, the processing speed of all optical processors of TOC
For σ, then its service speed μ3=σ/D.Because the reconstruct of TOC optical processor be it is parallel, required time be a very little it is normal
Number, therefore ignored.The service speed μ of each small optical processor3E=σ/nD.
R3 and T under systematic steady state is sought with quasi- raw (QBD, Quasi-birth-and-death) process model that goes out below3.It enables
Q (t) indicates number of tasks in t moment SAS, and is defined as follows V (t) function.
{ (Q (t), V (t)) } constitutes a two-dimensional Markov process, and state space is
Ω={ (0,1) } ∪ (k, j) | k >=1 and k ∈ N, j=0,1 }.
At the end of the operation or optical processor that have task arrival SAS, TOC one task of completion and decoder are had a holiday,
State changes.By in level, that is, SAS number of tasks and optical processor and decoder whether have a holiday the shape of the process
State sequence, it is as shown in Figure 4 can to obtain its state metastasis.Fig. 4 respectively indicated at the middle and upper levels with each node next time optical processor and
Decoder it is in running order with have a holiday state.For example, state (2,1) indicates that optical processor and decoder are had a holiday in not
False state, and have 2 unscheduled tasks in SAS.It has a holiday before end, if the 3rd task reaches, it is turned with λ rate
Move to state (3,1);If just ended a holiday, then it is transferred to state (2,0) with δ rate before the 3rd task reaches.Shape
State (2,0) indicates that optical processor and decoder are had a holiday in running order, and have 2 tasks to occupy 2 small optical treatments
Device, with 2 μ3EService speed operation, remaining n-2 small optical processors are in idle condition, therefore it is with 2 μ3ERate turns
Move to state (1,0).Before this 2 tasks terminate, the 3rd task is reached, then SAS will immediately be scheduled it, be it
A small optical processor is distributed, i.e. state (2,0) is transferred to state (3,0) with λ rate.Particularly, to state (1,0), one
After a small optical processor completion task, if reached without task, optical processor and decoder will enter shape of having a holiday
State, i.e., with μ3ERate is transferred to state (0,1) by state (1,0).
State is arranged by lexcographical order, Fig. 4 can be obtained and generate variable matrix G as follows.
WhereinG can be written as follow block triangular battle array.
Wherein A0=-λ, C0=(0, λ),1≤i≤n,
I is second order unit matrix, and is met
(A0+C0) I=(A1+B1+C1) I=...=(An+Bn+Cn) I=(A+B+C) I=0
Obviously, G has piecemeal tridiagonal structure, shows that { (Q (t), V (t)) } is a simulated AC curve.It enables(Q, V) indicates the stable state limit of process (Q (t), V (t)).Work as ρ3When < 1, P is rememberediIndicate that TOC is in work
There is the probability of i task when state in system;pijIndicate that (Q, V) is in (i, j) shape probability of state, (i, j) ∈ Ω;piIt indicates
TOC is in the probability for having i task when having a holiday state in system, i.e.,
Pi=P Q=i, J=0), i >=0,
pij=P { Q=i, J=j }, (i, j) ∈ Ω,
pi=pI1,i≥0。
The distribution of (Q, V) can be obtained
Wherein
Then, when can obtain system and reaching steady state, average number of tasks R in Stage 33Probability distribution P (R3=0 }=
C, P { R3=j }=Pj+pj, j >=1.
Wherein, α=ρ (ρ+n-n ρ).It can be obtained by Little formula
If q0And q1The probability that TOC under stable state is in the busy period He has a holiday is respectively indicated, then
q0=P { J=0 }=1-q1,
The average utilization of TOC
The performance evaluation of Stage 4
The output of Stage 3 is also the Poisson process that a mean value is λ when system reaches equilibrium state.In other words,
The Mean Speed that task reaches Stage 4 is similarly λ.We equally use M/M/1 queuing system to the data conversion in the stage and
Data transmission is modeled.
Because the result of M binary three-valued logic operations is M.In other words, the binary for being D to average amount
Three-valued logic operation, it is D/2 that TS, which needs the data volume of processor,.Assuming that the rate of TS processing data carries out data with PPS and locates in advance
The rate of reason is identical, i.e., is also τ, then the data service rates μ of TS41=2 τ/D, data send service speed μ42=2 ξ/D.WhenAndWhen, the data processing of Stage 4 and data, which are sent all, to be had and Stage 1
Identical probability of stability equation group.Therefore, the performance evaluation conclusion of Stage 1 can be directly applied to Stage 4 its be averaged
Number of request R4With average response time T4
(2)~(6) are respectively substituted into (1), average number of tasks in Mean Time of Systemic Response and system can be obtained.
Step 104: establishing Evaluating Models;The Evaluating Models are used to calculate the flat of three value optical computers
Equal response time and average request number.
Step 105: according to the Evaluating Models, by average response time during each stage running and
Average request number calculates the average response time and average request number of the model to be serviced.
Step 106: according to the average response time of the service model and average request number, to described to be evaluated
Three value optical computers carry out performance evaluation.
The specific embodiment provided according to the present invention, the invention discloses following technical effects: the present invention calculates three value light
Learn the average response time and average request number during each stage running of computer;According to the Evaluating Models, lead to
The average response time and average request number during each stage running are crossed, the average response of the model to be serviced is calculated
Time and average request number;According to the average response time of the service model and average request number, to described to be evaluated
Three value optical computers carry out performance evaluation.The performance of three value optical computers can be accurately evaluated by the above method.
As shown in figure 5, the present invention also provides a kind of three value optical computer performance evaluation systems based on Vacation queue
System, the system comprises:
Model obtains module 501, for obtaining based on serial and Vacation queue three value optical computers to be evaluated
Service model.
Division module 502, for the service model to be divided into four-stage.
The four-stage includes first stage, second stage, phase III and fourth stage;
In the first stage, server is received by not received operation request and presses FCFS strategy the first layer queue of deposit,
And a request is taken out from non empty request queue, the request taken out from non empty request queue is sent to data prediction
The operation received request is converted into TOC task to be processed by server;
In the second stage, the decimal data that user inputs is converted into MSD data by data prediction server,
The control signal that optical processor is generated according to operation, sends control signals to SAS;
In the phase III, SAS will receive task first and be inserted into third time queue by FCFS strategy, then first
It checks OP whether on holiday state, if OP is in state of having a holiday, waits;Otherwise check whether that there are idle processors;If no
In the presence of then continuing waiting for;Otherwise, according to scheduling strategy scheduler task;The RU of OP is according to distribution information and restructuring directive step-by-step weight
Structure arithmetic unit, then obtains data, and OP carries out concurrent operation using control signal, the operation result that DC indicates optical signal into
Row decoding, and decoded data are sent to operation result and send server TS;While OP carries out operation, TOC judgement
Whether the data of unfinished operation are had, if so, then continuing data;Otherwise, TOC sends " operation completion signal " to SAS;
SAS checks whether unscheduled task after receiving signal;If so, then SAS again scheduler task, distribution resource;Otherwise,
SAS sends " no request " signal to TOC, and TOC starts having a holiday for a random-length;At the end of having a holiday, TOC is sent " not to SAS
False knot beam " signal, SAS checks whether unscheduled task after receiving signal again, and handles it;
In the fourth stage, the operation result indicated with MSD data that TS sends over TOC is added by FCFS strategy
Enter the 4th layer queue, obtains result queue, and successively take out user's operation result from non-empty result queue, or according to need
Operation result is converted into decimal data, is sent to relative users.
First computing module 503, for calculating average response time and average request during each stage running
Number.
Modeling module 504, for establishing Evaluating Models;The Evaluating Models are by calculating based on three value optics
The average response time and average request number of calculation machine.
Second computing module 505, for passing through being averaged during each stage running according to the Evaluating Models
Response time and average request number calculate the average response time and average request number of the model to be serviced.
Evaluation module 506, for the average response time and average request number according to the service model, to described
Three value optical computers to be evaluated carry out performance evaluation.
The average response time and average request number in first stage operational process are calculated, is specifically included:
Acquiring unit, calculated for obtaining in first stage operational process operation in the unit time request average arrival rate,
The average receptance of RS in unit time;
First computing unit, for calculating first rank according to the average arrival rate and the average receptance
The average request number of section;
Second computing unit, for according to the first stage average request number and the average arrival rate, calculate
The average response time of the first stage.
Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with its
The difference of his embodiment, the same or similar parts in each embodiment may refer to each other.For being disclosed in embodiment
For system, since it is corresponded to the methods disclosed in the examples, so being described relatively simple, related place is referring to method portion
It defends oneself bright.
Used herein a specific example illustrates the principle and implementation of the invention, above embodiments
Illustrate to be merely used to help understand method and its core concept of the invention;At the same time, for those skilled in the art,
According to the thought of the present invention, there will be changes in the specific implementation manner and application range.In conclusion this specification
Content should not be construed as limiting the invention.
Claims (6)
1. a kind of three value optical computer method of evaluating performance based on Vacation queue, which is characterized in that the described method includes:
Obtain the service model based on serial and Vacation queue three value optical computers to be evaluated;
The service model is divided into four-stage;
Calculate the average response time and average request number during each stage running;
Establish Evaluating Models;The Evaluating Models be used for calculate three value optical computers average response time and
Average request number;
According to the Evaluating Models, pass through the average response time and average request number during each stage running, meter
Calculate the average response time and average request number of the model to be serviced;
According to the average response time of the service model and average request number, to the three value optical computers to be evaluated
Carry out performance evaluation.
2. the three value optical computer method of evaluating performance according to claim 1 based on Vacation queue, which is characterized in that
The four-stage includes first stage, second stage, phase III and fourth stage;
In the first stage, receives server and request not received operation to be stored in the first layer queue by FCFS strategy, and from
A request is taken out in non empty request queue, and the request taken out from non empty request queue is sent to data prediction service
The operation received request is converted into TOC task to be processed by device;
In the second stage, the decimal data that user inputs is converted into MSD data by data prediction server, according to fortune
The control signal for generating optical processor is calculated, SAS is sent control signals to;
In the phase III, SAS will receive task first and be inserted into third time queue by FCFS strategy, then first look at OP
Whether on holiday state waits if OP is in state of having a holiday;Otherwise check whether that there are idle processors;If it does not exist, then
It continues waiting for;Otherwise, according to scheduling strategy scheduler task;The RU of OP is according to distribution information and restructuring directive step-by-step restructing operation
Then device obtains data, OP carries out concurrent operation using control signal, and the operation result that DC indicates optical signal is decoded,
And decoded data are sent to operation result and send server TS;While OP carries out operation, TOC is judged whether there is not
The data for completing operation, if so, then continuing data;Otherwise, TOC sends " operation completion signal " to SAS;SAS receives letter
Unscheduled task is checked whether after number;If so, then SAS again scheduler task, distribution resource;Otherwise, SAS is sent to TOC
" no request " signal, TOC start having a holiday for a random-length;At the end of having a holiday, TOC sends " end of having a holiday " signal to SAS,
SAS checks whether unscheduled task after receiving signal again, and handles it;
In the fourth stage, the operation result indicated with MSD data that TS sends over TOC is added the 4th by FCFS strategy
Layer queue obtains result queue, and successively takes out user's operation result from non-empty result queue, or as needed by operation
As a result it is converted into decimal data, is sent to relative users.
3. the three value optical computer method of evaluating performance according to claim 2 based on Vacation queue, which is characterized in that
The average response time and average request number in first stage operational process are calculated, is specifically included:
It obtains in first stage operational process operation in the unit time and calculates that the average arrival rate of request, RS's is averaged in the unit time
Receptance;
According to the average arrival rate and the average receptance, the average request number of the first stage is calculated;
According to the average request number of the first stage and the average arrival rate, the average response of the first stage is calculated
Time.
4. a kind of three value optical computer Performance Appraisal Systems based on Vacation queue, which is characterized in that the system comprises:
Model obtains module, for obtaining the service mould based on serial and Vacation queue three value optical computers to be evaluated
Type;
Division module, for the service model to be divided into four-stage;
First computing module, for calculating average response time and average request number during each stage running;
Modeling module, for establishing Evaluating Models;The Evaluating Models are used to calculate the flat of three value optical computers
Equal response time and average request number;
Second computing module, for passing through the average response time during each stage running according to the Evaluating Models
And average request number, calculate the average response time and average request number of the model to be serviced;
Evaluation module, for the average response time and average request number according to the service model, to described to be evaluated
Three value optical computers carry out performance evaluation.
5. the three value optical computer Performance Appraisal Systems according to claim 4 based on Vacation queue, which is characterized in that
The four-stage includes first stage, second stage, phase III and fourth stage;
In the first stage, receives server and request not received operation to be stored in the first layer queue by FCFS strategy, and from
A request is taken out in non empty request queue, and the request taken out from non empty request queue is sent to data prediction service
The operation received request is converted into TOC task to be processed by device;
In the second stage, the decimal data that user inputs is converted into MSD data by data prediction server, according to fortune
The control signal for generating optical processor is calculated, SAS is sent control signals to;
In the phase III, SAS will receive task first and be inserted into third time queue by FCFS strategy, then first look at OP
Whether on holiday state waits if OP is in state of having a holiday;Otherwise check whether that there are idle processors;If it does not exist, then
It continues waiting for;Otherwise, according to scheduling strategy scheduler task;The RU of OP is according to distribution information and restructuring directive step-by-step restructing operation
Then device obtains data, OP carries out concurrent operation using control signal, and the operation result that DC indicates optical signal is decoded,
And decoded data are sent to operation result and send server TS;While OP carries out operation, TOC is judged whether there is not
The data for completing operation, if so, then continuing data;Otherwise, TOC sends " operation completion signal " to SAS;SAS receives letter
Unscheduled task is checked whether after number;If so, then SAS again scheduler task, distribution resource;Otherwise, SAS is sent to TOC
" no request " signal, TOC start having a holiday for a random-length;At the end of having a holiday, TOC sends " end of having a holiday " signal to SAS,
SAS checks whether unscheduled task after receiving signal again, and handles it;
In the fourth stage, the operation result indicated with MSD data that TS sends over TOC is added the 4th by FCFS strategy
Layer queue obtains result queue, and successively takes out user's operation result from non-empty result queue, or as needed by operation
As a result it is converted into decimal data, is sent to relative users.
6. the three value optical computer Performance Appraisal Systems according to claim 5 based on Vacation queue, which is characterized in that
The average response time and average request number in first stage operational process are calculated, is specifically included:
Acquiring unit, for obtaining, the average arrival rate of request, unit are calculated in operation in the unit time in first stage operational process
The average receptance of RS in time;
First computing unit, for calculating the first stage according to the average arrival rate and the average receptance
Average request number;
Second computing unit, for according to the first stage average request number and the average arrival rate, described in calculating
The average response time of first stage.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910378982.0A CN110362457B (en) | 2019-05-08 | 2019-05-08 | Three-value optical computer performance evaluation method and system based on vacation queuing |
ZA2019/04644A ZA201904644B (en) | 2019-05-08 | 2019-07-16 | Method and system for performance evaluation of ternary optical computer based on vacation queue |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910378982.0A CN110362457B (en) | 2019-05-08 | 2019-05-08 | Three-value optical computer performance evaluation method and system based on vacation queuing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110362457A true CN110362457A (en) | 2019-10-22 |
CN110362457B CN110362457B (en) | 2023-03-28 |
Family
ID=68215400
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910378982.0A Active CN110362457B (en) | 2019-05-08 | 2019-05-08 | Three-value optical computer performance evaluation method and system based on vacation queuing |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN110362457B (en) |
ZA (1) | ZA201904644B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117632084A (en) * | 2024-01-26 | 2024-03-01 | 华东交通大学 | Large integer power modular rapid computing method and system based on three-value optical processor |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS5428503A (en) * | 1977-08-06 | 1979-03-03 | Sumitomo Electric Ind Ltd | Transmission system for ternary optical data |
CN104331477A (en) * | 2014-11-04 | 2015-02-04 | 哈尔滨工业大学 | Method for testing concurrency property of cloud platform based on federated research |
JP2015118434A (en) * | 2013-12-17 | 2015-06-25 | インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Machines Corporation | Calculation method, calculation device, computer system, and program for evaluating response performance in computer system in which multiple arithmetic processing units can operate in arithmetic core |
CN105610897A (en) * | 2015-11-20 | 2016-05-25 | 阜阳师范学院 | M/M/1 queuing based TOC (Ternary Optical Computer) service model and calculation method for service response time of model |
CN106572159A (en) * | 2016-11-01 | 2017-04-19 | 阜阳师范学院 | Operation method of complex queuing system-based three-valued optical computer service response time |
CN107066029A (en) * | 2017-01-27 | 2017-08-18 | 丽水市飞天人机械设计有限公司 | Three value optical computer limitation incoming symbol single step MSD adders |
CN107416400A (en) * | 2017-07-04 | 2017-12-01 | 山东大学 | Model and its optimization method based on the intensive automatic storage system of cross-layer shuttle |
-
2019
- 2019-05-08 CN CN201910378982.0A patent/CN110362457B/en active Active
- 2019-07-16 ZA ZA2019/04644A patent/ZA201904644B/en unknown
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS5428503A (en) * | 1977-08-06 | 1979-03-03 | Sumitomo Electric Ind Ltd | Transmission system for ternary optical data |
JP2015118434A (en) * | 2013-12-17 | 2015-06-25 | インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Machines Corporation | Calculation method, calculation device, computer system, and program for evaluating response performance in computer system in which multiple arithmetic processing units can operate in arithmetic core |
CN104331477A (en) * | 2014-11-04 | 2015-02-04 | 哈尔滨工业大学 | Method for testing concurrency property of cloud platform based on federated research |
CN105610897A (en) * | 2015-11-20 | 2016-05-25 | 阜阳师范学院 | M/M/1 queuing based TOC (Ternary Optical Computer) service model and calculation method for service response time of model |
CN106572159A (en) * | 2016-11-01 | 2017-04-19 | 阜阳师范学院 | Operation method of complex queuing system-based three-valued optical computer service response time |
CN107066029A (en) * | 2017-01-27 | 2017-08-18 | 丽水市飞天人机械设计有限公司 | Three value optical computer limitation incoming symbol single step MSD adders |
CN107416400A (en) * | 2017-07-04 | 2017-12-01 | 山东大学 | Model and its optimization method based on the intensive automatic storage system of cross-layer shuttle |
Non-Patent Citations (4)
Title |
---|
QUN XU等: "Application Research of Ternary Optical Computer on Gaussian Elimination", 《2015 IEEE 12TH INTL CONF ON UBIQUITOUS INTELLIGENCE AND COMPUTING AND 2015 IEEE 12TH INTL CONF ON AUTONOMIC AND TRUSTED COMPUTING AND 2015 IEEE 15TH INTL CONF ON SCALABLE COMPUTING AND COMMUNICATIONS AND ITS ASSOCIATED WORKSHOPS (UIC-ATC-SCALCOM)》 * |
徐群等: "三值光学计算机的MSD迭代除法算法和实现技术", 《中国科学:信息科学》 * |
徐群等: "基于复杂排队***的三值光学计算机服务模型与性能分析", 《国防科技大学学报》 * |
王先超等: "基于M/M/1排队***的三值光学计算机任务服务模型", 《阜阳师范学院学报(自然科学版)》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117632084A (en) * | 2024-01-26 | 2024-03-01 | 华东交通大学 | Large integer power modular rapid computing method and system based on three-value optical processor |
CN117632084B (en) * | 2024-01-26 | 2024-04-05 | 华东交通大学 | Large integer power modular rapid computing method and system based on three-value optical processor |
Also Published As
Publication number | Publication date |
---|---|
CN110362457B (en) | 2023-03-28 |
ZA201904644B (en) | 2020-05-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108351805B (en) | Flow-based accelerator processing of computational graphs | |
EP3705994B1 (en) | Processing computational graphs | |
CN108885571B (en) | Input of batch processing machine learning model | |
Lin et al. | Joint optimization of overlapping phases in MapReduce | |
Wang et al. | Maptask scheduling in mapreduce with data locality: Throughput and heavy-traffic optimality | |
US8706869B2 (en) | Distributed cloud placement software | |
US11429434B2 (en) | Elastic execution of machine learning workloads using application based profiling | |
US11055139B2 (en) | Smart accelerator allocation and reclamation for deep learning jobs in a computing cluster | |
CN106020933A (en) | Ultra-lightweight virtual machine-based cloud computing dynamic resource scheduling system and method | |
US11676013B2 (en) | Job-launch time reduction by node pre-configuration | |
Li et al. | An effective scheduling strategy based on hypergraph partition in geographically distributed datacenters | |
CN114647515A (en) | GPU cluster-oriented dynamic resource scheduling method | |
JP2024519267A (en) | Asynchronous distributed dataflow for machine learning workloads | |
CN114514536A (en) | Neural network training in distributed systems | |
CN110362457A (en) | A kind of three value optical computer method of evaluating performance and system based on Vacation queue | |
Stavrinides et al. | The impact of resource heterogeneity on the timeliness of hard real-time complex jobs | |
Xue et al. | BOLAS: bipartite-graph oriented locality-aware scheduling for MapReduce tasks | |
CN110401681A (en) | For data transmission, the method for data receiver and electronic equipment | |
CN114741161A (en) | HPC job cluster sensing method based on mixed cluster | |
CN106572159B (en) | The operating method of the three value optical computer service response time based on complicated queuing system | |
Bensaleh et al. | Optimal task scheduling for distributed cluster with active storage devices and accelerated nodes | |
Zhou et al. | Training and Serving System of Foundation Models: A Comprehensive Survey | |
Do et al. | Co-scheduling ensembles of in situ workflows | |
Teng et al. | Scheduling performance of real-time tasks on mapreduce cluster | |
Sunder et al. | Load balancing optimization based on enhanced genetic algorithm in cloud computing |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |