CN106201847A - Consider method for allocating tasks, device and the system of the decay of cloud platform host performance - Google Patents

Abstract

The invention discloses a kind of method for allocating tasks considering that cloud platform host performance is decayed, relate to cloud computing system control field, comprise the steps of: first, monitor each main frame and gather its performance information；Secondly, assessment main frame current performance and load, assess each host performance attenuation degree；Then, new task allocative decision is generated according to the host performance dough softening；Finally, main frame is automatically distributed to new arrival task；In addition the invention also discloses and based on said method a kind of consider the task allocation apparatus that cloud platform host performance is decayed and the system comprising this device.The present invention, when carrying out new task allocation schedule, have evaluated performance degradation degree and the expection backlog quantity of each main frame, it is to avoid new task is distributed to the main frame that performance degradation degree is high, backlog is many, makes task scheduling approach more reasonable.

Description

Task allocation method, device and system considering performance attenuation of cloud platform host

Technical Field

The invention belongs to the field of cloud computing system control, and particularly relates to a task allocation method, device and system considering cloud platform host performance attenuation.

Background

Cloud computing is an internet-based computing approach by which shared software and hardware resources and information can be provided to computers and other devices on demand. Compared with traditional software and computing forms, cloud computing has the remarkable advantages of loose coupling, on-demand, controllable cost, virtual resources, heterogeneous cooperation and the like, and is more suitable for applications such as electronic commerce, flexible manufacturing, mobile internet and the like.

A cloud platform refers to a distributed computing system composed of multiple heterogeneous hosts connected together by a network and used for carrying enterprise-level applications providing online cloud services. In the cloud platform, a large number of hosts are centrally and uniformly managed, so that a stable power supply environment required by the operation of the hosts, and appropriate temperature and humidity control and network bandwidth conditions can be guaranteed.

Like other software and hardware systems, the load of the host in the cloud platform is also in real-time change. Because the cloud computing system is mostly used for high-load and high-complexity applications such as large-scale scientific computing, real-time finance, online transaction, streaming media multicast and the like, the host computer of the cloud computing system is often in an overload running state. After the cloud host runs under a high load for a long time, the overall performance of the cloud host is attenuated. If the task scheduling and load balancing strategy neglects performance degradation, the batch crash of tasks accumulated on the host with degraded performance is very easy to occur, and a great loss is caused. In the traditional task scheduling and allocating strategy, the allowable newly-added task capacity is calculated only according to the current load, the resource utilization rate and the reliability state of each host and each node, and the task allocation and scheduling scheme is determined according to the static data, so that the real-time performance fluctuation and attenuation of each host in the cloud computing system are ignored, and the situation that a new task is allocated to the host which is in a performance reduction situation although the performance is not low at present is not actively avoided.

Under the background, how to dynamically track the operation situation of the cloud platform, track and predict the performance change of the cloud host in real time, perform early warning on the performance attenuation condition, and realize a reasonable task allocation strategy becomes a research hotspot and difficulty.

Disclosure of Invention

In view of the above-mentioned drawbacks of the prior art, the technical problems to be solved by the present invention are: the problem that task allocation is misappropriate due to the fact that host performance attenuation is not considered in cloud system task allocation in the prior art is solved. Aiming at the problem, the invention provides a task allocation method, a device and a system considering the performance attenuation of a cloud platform host.

In order to achieve the above object, the present invention provides a task allocation method considering performance attenuation of a cloud platform host, comprising the following steps:

monitoring each host and collecting performance information of each host; the performance information includes the total running time KT of the host_iTotal number of completed tasks KWC_iAnd the number of tasks completed WC within the last t time_iThe number XD of newly arrived tasks of each host_iNumber of completed tasks for host error or re-submission after suspension CTJ_iNumber of faulty tasks CW_iAnd the number of migrated tasks QC_iRecording the downtime of the host; wherein i is a subscript value to indicate the host number to which the variable belongs, and i is more than 0 and less than or equal to n;

step two, evaluating the current performance and load of the host: calculating the equivalent survival rate CHL of each host in t time_iMultitask failure rate RWSXL_iEquivalent execution rate DXZX_iAnd average task input rate PJL of all hosts;

evaluating the performance attenuation degree of each host, evaluating the potential task stock of each host within the next t time, evaluating the load satisfaction degree of each host under the condition of considering the occurrence of the fault, and evaluating the performance attenuation degree of each host; wherein,

potential task inventory ZDL_iComprises the following steps:

$\begin{array}{c}{\mathrm{ZDL}}_i=t\&times;\left(\right(1-a)\&times;PJL+max\{\frac{{\mathrm{XD}}_i}{t}|0<i\&le;n\})+{\mathrm{QC}}_i+{\mathrm{CTJ}}_i\\ +\frac{{\mathrm{CW}}_i}{1-{\mathrm{RWSXL}}_i}\end{array}$

wherein a satisfies 0< a <1, and is a weight coefficient preset by a cloud system manager;

load satisfaction MZD_idXZX with host execution rate_iProportional to the amount of potential mission inventory ZDL_iInversely proportional to the host equivalent survival rate CHL_iAnd the number of error tasks CW_iCorrelation; degree of performance decay SJD_iWhen the first calculation is carried out, the default is set to 0, and the non-first calculation is carried out to satisfy the MZD_iJMZD_iJMZD of the difference and old satisfaction_iThe ratio of (A) to (B);

step four, generating a new task allocation scheme according to the performance attenuation of each host: degree of attenuation of performance SJD_iIs positive and has a performance attenuation degree SJD_iThe ratio of the average performance attenuation degree of each host is larger than b and the average equivalent execution rate of each host is DXZX_iTo process its potential task inventory ZDL_iThe host computer with the required time larger than c × t is marked as not accepting the new task, otherwise, the host computer is marked as accepting the new task, wherein, b (satisfies 1)<b<10) And c (satisfy 1)<c<10) Parameters are predefined for the system administrator.

Step five, automatically distributing the host computer to the newly arrived task continuously and updating the load satisfaction degree; when all hosts do not bear tasks, rejecting the newly arrived task; otherwise, the task is distributed to the host with the maximum load satisfaction value, and the ZDL is used_iAnd increasing the value by 1 and updating the load satisfaction degree.

Further, the equivalent survival rate CHL_iThe calculation formula is as follows:

wherein, HF_iXF the latest time spent in downtime within t_iThe interval time from the recovery moment to the current moment after the latest downtime within the time t;

further, the multitask failure rate is:

${\mathrm{RWSXL}}_i=1-{\left({\mathrm{CHL}}_i\right)}^{{\mathrm{CW}}_i}$

further, the equivalent execution rate is:

${\mathrm{DXZX}}_i=\left\{\begin{array}{cc}\frac{{\mathrm{WC}}_i-{\mathrm{CTJ}}_i}{{\mathrm{WC}}_i+{\mathrm{QC}}_i}\&times;{(\frac{t}{{\mathrm{WC}}_i}\&times;{\mathrm{CHL}}_i)}^{-1}& if{\mathrm{WC}}_i>0\\ \frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}& elseif{\mathrm{WC}}_i=0\end{array}\right.$

further, the average task input rate is:

$PJL=\frac{{\mathrm{\&Sigma;}}_{i=0}^n\frac{{\mathrm{XD}}_i}{t}}{n}$

wherein the number of tasks KWC_iThe initial value is 1; when the total running time of the system is less than t, KT_iIs set to t.

Further, the load satisfaction is:

${\mathrm{MZD}}_i=\left\{\begin{array}{cc}\frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{ZDL}}_i}& if{\mathrm{ZDL}}_i>0\\ \frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{KWC}}_i\&times;\frac{t}{{\mathrm{KT}}_i}}& else\end{array}\right.$

further, the fourth step is: marking the hosts with positive performance attenuation values, the ratio of the performance attenuation degrees to the average performance attenuation degrees of the hosts is greater than b, and the time required for processing the potential task inventory by the average equivalent execution rate of the hosts is greater than c x t as not accepting the new task, otherwise, marking the hosts as capable of accepting the new task; wherein b (1 < b <10 is satisfied) and c (1 < c <10 is satisfied) are parameters predetermined by a system administrator.

Further, in step five, the new task allocation scheme has a validity period, and the validity period of the new task allocation scheme is as follows:

$YXQ=\left\{\begin{array}{cc}\frac{{\mathrm{\&Sigma;}}_{i=0}^n{\mathrm{ZDL}}_i}{mean\left\{\begin{array}{cc}\frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}& |0<i\&le;n\end{array}\right\}}+t& ifmean\left\{{\mathrm{JS}}_i\right|0<i\&le;n\}=1\\ t& else\end{array}\right.$

and when the validity period of the new task allocation scheme is reached, executing a jump to execute the step one.

The invention also provides a task allocation device considering the performance attenuation of the cloud platform host, which comprises a cloud system performance monitoring module, a performance attenuation early warning module, a new task allocation scheme generation module and a new task allocation control module;

cloud system performance monitoring moduleEach host is monitored in block, performance information is collected, and the total running time KT of the host is obtained_iTotal number of completed tasks KWC_iAnd the number of tasks completed WC within the last t time_iThe number XD of newly arrived tasks of each host_iNumber of completed tasks for host error or re-submission after suspension CTJ_iNumber of faulty tasks CW_iAnd the number of migrated tasks QC_iThe host downtime record is sent to the performance attenuation early warning module; running the main machine for total time KT_iTotal number of completed tasks KWC_iSending the value to a new task allocation control module; wherein i is a subscript value to indicate the host number to which the variable belongs, and i is more than 0 and less than or equal to n;

the performance attenuation early warning module calculates the equivalent execution rate DXZX_iSending the value to a new task allocation scheme generation module, and calculating the performance attenuation degree SJD_iAnd potential task inventory ZDL_iSending the value to a new task allocation scheme generation module to store the potential task inventory ZDL_iSending the value to a new task allocation control module;

potential task inventory ZDL_iThe calculation formula is as follows:

$\begin{array}{c}{\mathrm{ZDL}}_i=t\&times;\left(\right(1-a)\&times;PJL+max\{\frac{{\mathrm{XD}}_i}{t}|0<i\&le;n\})+{\mathrm{QC}}_i+\\ {\mathrm{CTJ}}_i+\frac{{\mathrm{CW}}_i}{1-{\mathrm{RWSXL}}_i};\end{array}$

wherein a satisfies 0< a <1, and is a weight coefficient preset by a cloud system manager;

load satisfaction MZD_iProportional to the host execution rate, inversely proportional to the potential task stock, and related to the host equivalent survival rate and the number of error tasks; degree of performance decay SJD_iWhen the first calculation is carried out, the default is set to 0, and the non-first calculation is carried out to satisfy the MZD_iJMZD_iJMZD of the difference and old satisfaction_iThe ratio of (A) to (B);

and the new task allocation scheme generation module generates a new task allocation scheme and the new task allocation control module allocates the tasks.

Further, the equivalent survival rate CHL_iThe calculation formula is as follows:

wherein, HF_iXF the latest time spent in downtime within t_iThe interval time from the recovery moment to the current moment after the latest downtime within the time t;

further, multitask failure rate RWSXL_iThe calculation formula is as follows:

${\mathrm{RWSXL}}_i=1-{\left({\mathrm{CHL}}_i\right)}^{{\mathrm{CW}}_i};$

further, the equivalent execution rate DXZX_iThe calculation formula is as follows:

${\mathrm{DXZX}}_i=\left\{\begin{array}{cc}\frac{{\mathrm{WC}}_i-{\mathrm{CTJ}}_i}{{\mathrm{WC}}_i+{\mathrm{QC}}_i}\&times;{(\frac{t}{{\mathrm{WC}}_i}\&times;{\mathrm{CHL}}_i)}^{-1}& if{\mathrm{WC}}_i>0\\ \frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}& elseif{\mathrm{WC}}_i=0\end{array}\right.;$

further, the average task input rate PJL is calculated by the following formula:

$PJL=\frac{{\mathrm{\&Sigma;}}_{i=0}^n\frac{{\mathrm{XD}}_i}{t}}{n};$

wherein the number of tasks KWC_iThe initial value is 1; when the total running time of the system is less than t, KT_iIs set to t.

Further, the load satisfies the MZD_iComprises the following steps:

${\mathrm{MZD}}_i=\left\{\begin{array}{cc}\frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{ZDL}}_i}& if{\mathrm{ZDL}}_i>0\\ \frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{KWC}}_i\&times;\frac{t}{{\mathrm{KT}}_i}}& else\end{array}\right.;$

further, the new task allocation scheme has a validity period, and the calculation formula of the validity period of the new task allocation scheme is as follows:

$YXQ=\left\{\begin{array}{cc}\frac{{\mathrm{\&Sigma;}}_{i=0}^n{\mathrm{ZDL}}_i}{mean\left\{\begin{array}{cc}\frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}& |0<i\&le;n\end{array}\right\}}+t& ifmean\left\{{\mathrm{JS}}_i\right|0<i\&le;n\}=1\\ t& else\end{array}\right..$

in addition, the invention also provides a task allocation system considering the performance attenuation of the cloud platform host, which comprises a cloud platform server, wherein the cloud platform server is internally provided with a cloud platform host performance attenuation task allocation device; the cloud platform host performance attenuation task allocation device comprises a cloud system performance monitoring module, a performance attenuation early warning module, a new task allocation scheme generation module and a new task allocation control module;

the cloud system performance monitoring module monitors each host, collects performance information of each host, and runs the host for the total time KT_iTotal number of completed tasks KWC_iAnd the number of tasks completed WC within the last t time_iThe number XD of newly arrived tasks of each host_iNumber of completed tasks for host error or re-submission after suspension CTJ_iNumber of faulty tasks CW_iAnd the number of migrated tasks QC_iThe host downtime record is sent to the performance attenuation early warning module; running the main machine for total time KT_iTotal number of completed tasks KWC_iSending the value to a new task allocation control module; wherein i is a subscript value to indicate the host number to which the variable belongs, and i is more than 0 and less than or equal to n;

the performance attenuation early warning module calculates the equivalent execution rate DXZX_iSending the value to a new task allocation scheme generation module, and calculating the performance attenuation degree SJD_iAnd potential task inventory ZDL_iSending the value to a new task allocation scheme generation module to store the potential task inventory ZDL_iSending the value to a new task allocation control module;

potential task inventory ZDL_iThe calculation formula is as follows:

$\begin{array}{c}{\mathrm{ZDL}}_i=t\&times;\left(\right(1-a)\&times;PJL+max\{\frac{{\mathrm{XD}}_i}{t}|0<i\&le;n\})+{\mathrm{QC}}_i+\\ {\mathrm{CTJ}}_i+\frac{{\mathrm{CW}}_i}{1-{\mathrm{RWSXL}}_i};\end{array}$

wherein a satisfies 0< a <1, and is a weight coefficient preset by a cloud system manager;

load satisfaction MZD_iProportional to the host execution rate, and potential taskThe service stock is inversely proportional and is related to the equivalent survival rate of the host and the number of error tasks; degree of performance decay SJD_iWhen the first calculation is carried out, the default is set to 0, and the non-first calculation is carried out to satisfy the MZD_iJMZD_iJMZD of the difference and old satisfaction_iThe ratio of (A) to (B);

and the new task allocation scheme generation module generates a new task allocation scheme and the new task allocation control module allocates the tasks.

Further, the equivalent survival rate CHL_iThe calculation formula is as follows:

wherein, HF_iXF the latest time spent in downtime within t_iThe interval time from the recovery moment to the current moment after the latest downtime within the time t;

further, multitask failure rate RWSXL_iThe calculation formula is as follows:

${\mathrm{RWSXL}}_i=1-{\left({\mathrm{CHL}}_i\right)}^{{\mathrm{CW}}_i};$

further, the equivalent execution rate DXZX_iThe calculation formula is as follows:

${\mathrm{DXZX}}_i=\left\{\begin{array}{cc}\frac{{\mathrm{WC}}_i-{\mathrm{CTJ}}_i}{{\mathrm{WC}}_i+{\mathrm{QC}}_i}\&times;{(\frac{t}{{\mathrm{WC}}_i}\&times;{\mathrm{CHL}}_i)}^{-1}& if{\mathrm{WC}}_i>0\\ \frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}& elseif{\mathrm{WC}}_i=0\end{array}\right.;$

further, the average task input rate PJL is calculated by the following formula:

$PJL=\frac{{\mathrm{\&Sigma;}}_{i=0}^n\frac{{\mathrm{XD}}_i}{t}}{n};$

wherein the number of tasks KWC_iThe initial value is 1; when the total running time of the system is less than t, KT_iIs set to t.

Further, the load satisfies the MZD_iComprises the following steps:

${\mathrm{MZD}}_i=\left\{\begin{array}{cc}\frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{ZDL}}_i}& if{\mathrm{ZDL}}_i>0\\ \frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{KWC}}_i\&times;\frac{t}{{\mathrm{KT}}_i}}& else\end{array}\right.;$

further, the new task allocation scheme has a validity period, and the calculation formula of the validity period of the new task allocation scheme is as follows:

$YXQ=\left\{\begin{array}{cc}\frac{{\mathrm{\&Sigma;}}_{i=0}^n{\mathrm{ZDL}}_i}{mean\left\{\begin{array}{cc}\frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}& |0<i\&le;n\end{array}\right\}}+t& ifmean\left\{{\mathrm{JS}}_i\right|0<i\&le;n\}=1\\ t& else\end{array}\right..$

the invention has the beneficial effects that: the invention fully considers the dynamic fluctuation of the host performance in the cloud system, calculates the processing capacity satisfaction degree and the performance attenuation degree of each host closer to the actual task load, evaluates the performance attenuation degree and the expected backlog task quantity of each host, avoids distributing new tasks to the hosts with high performance attenuation degree and more backlog tasks, and ensures that the task scheduling scheme is more reasonable. Compared with a static task scheduling management strategy, the method realizes dynamic load balancing and can obtain better effect under the unreliable and high-load system operation environment.

Drawings

Fig. 1 is a flowchart illustrating a task allocation method according to an embodiment of the present invention, in which performance degradation of a cloud platform host is considered.

Fig. 2 is a schematic diagram of a task allocation apparatus according to an embodiment of the present invention, which considers performance degradation of a host of a cloud platform.

FIG. 3 is a schematic diagram of a specific embodiment of the task allocation system of the present invention that takes into account cloud platform host performance degradation.

Detailed Description

The invention is further illustrated by the following examples in conjunction with the accompanying drawings:

as shown in fig. 1, the present embodiment provides a task allocation method considering performance degradation of a cloud platform host, specifically including the following steps:

step one, collecting real-time performance information of a host

The cloud system performance monitoring module obtains each host in the cloud platform, and obtains the number of tasks KWC completed by the host when the host starts to run_iThe time KT elapsed from the start of the system operation to the present_i. Acquiring the number WC of tasks completed in the latest t time_iThe number XD of newly arrived tasks of each host_iNumber of completed tasks for host error or re-submission after suspension CTJ_iNumber of faulty tasks CW_iAnd the number of migrated tasks QC_iRecording the downtime of the host;

wherein, the system has not finished the task just in the beginning of the operation, at this moment the KWC_iThe initial value is set to 1. If the time elapsed from the start of the system to the present is less than t, KT_iIs set to t. i is a subscript value to indicate the host number of the variable, and i is more than 0 and less than or equal to n; the time interval t is given by the cloud system administrator and is set to 100 to 1000 milliseconds.

Obtaining the downtime record of the host, and setting the time HF consumed by the latest downtime of each host in t time_iAnd the interval XF from the recovery moment to the current moment after the latest downtime within the time t_i(ii) a If there is no downtime within t time, HF_iAnd XF_iAre recorded as-1.

Finally, the cloud system performance monitoring module acquires the WC_i、KT_i、KWC_i、XD_i、CTJ_i、CW_i、QC_i、XF_i、HF_iThe value is sent to a performance parameter preprocessing unit and an attenuation degree evaluation unit in a performance attenuation early warning module, and K is obtainedWC_iAnd KT_iThe value is sent to the new task assignment control module.

Step two, evaluating the current performance and load of the host:

a performance parameter preprocessing unit in a performance attenuation early warning module firstly calculates the equivalent survival rate CHL of each host within t time_i：

Wherein, HF_iXF the latest time spent in downtime within t_iThe interval time from the recovery moment to the current moment after the latest downtime within the time t;

then, calculating the task failure rate of each host considering the multi-task failure:

${\mathrm{RWSXL}}_i=1-{\left({\mathrm{CHL}}_i\right)}^{{\mathrm{CW}}_i}$

then, calculating the equivalent execution rate of each host within t time:

${\mathrm{DXZX}}_i=\left\{\begin{array}{cc}\frac{{\mathrm{WC}}_i-{\mathrm{CTJ}}_i}{{\mathrm{WC}}_i+{\mathrm{QC}}_i}\&times;{(\frac{t}{{\mathrm{WC}}_i}\&times;{\mathrm{CHL}}_i)}^{-1}& if{\mathrm{WC}}_i>0\\ \frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}& elseif{\mathrm{WC}}_i=0\end{array}\right.$

note that because CHL_iThe value cannot be 0 and the above calculation does not have the divisor 0.

Then, the average task input rate of each host is calculated:

$PJL=\frac{{\mathrm{\&Sigma;}}_{i=0}^n\frac{{\mathrm{XD}}_i}{t}}{n}$

finally, the performance parameter preprocessing unit calculates the calculated PJL and CHL_iAnd RWSXL_iThe value is sent to an attenuation degree evaluation unit in a performance attenuation early warning module, and DXZX is sent to_iThe value is sent to a new task allocation plan generation module.

Step three, evaluating the performance attenuation degree of each host

The attenuation degree evaluation unit in the performance attenuation early warning module firstly calculates the maximum possible potential task stock to be processed by each host within the next t time:

$\begin{array}{c}{\mathrm{ZDL}}_i=t\&times;\left(\right(1-a)\&times;PJL+max\{\frac{{\mathrm{XD}}_i}{t}|0<i\&le;n\})+{\mathrm{QC}}_i+{\mathrm{CTJ}}_i\\ +\frac{{\mathrm{CW}}_i}{1-{\mathrm{RWSXL}}_i}\end{array}$

wherein a is a weight coefficient preset by a cloud system manager and satisfies 0<a<1, and in general, any real number between 0.3 and 0.7 may be taken. The above mentionedThe intuitive meaning of the formula is that the estimated maximum stock task in the next t time can be estimated as the weighted task input rate multiplied by the time t and then the' number of migrated tasks QC_i"AND" Re-submit the number of tasks CTJ completed by execution after an error or suspension_i", multiply the expected number of rerun executions by" the number of tasks CW in error in the most recent t time_i"; max is the operation of maximizing the number of sets. Note that RWSXL_iThe value is more than or equal to 0 and less than 1, so that 1-RWSXL does not exist_iAs in the case of a divisor of 0.

Next, the variable MZD for the attenuation evaluating unit cell_iMarking the load satisfaction degree of each host, namely the degree of the current host execution rate which can still adapt to the task load under the condition of considering the occurrence of the fault:

${\mathrm{MZD}}_i=\left\{\begin{array}{cc}\frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{ZDL}}_i}& if{\mathrm{ZDL}}_i>0\\ \frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{KWC}}_i\&times;\frac{t}{{\mathrm{KT}}_i}}& else\end{array}\right.$

as can be seen, the satisfaction with ZDL_iThe value is inversely proportional to the equivalent execution rate DXZX of the host_iAnd is also related to the host equivalent survival rate and the number of error tasks. Wherein mean is the operation of averaging the number of sets. Notably, because of the KWC_iValue minimum 1, KT_iThe minimum value is also t, and the above calculation does not have the divisor 0.

Next, the attenuation degree evaluation unit calculates the performance attenuation degree of each host in consideration of load and reliability:

${\mathrm{SJD}}_i=\left\{\begin{array}{cc}\frac{{\mathrm{JMZD}}_i-{\mathrm{MZD}}_i}{{\mathrm{JMZD}}_i}& if{\mathrm{JMZD}}_i<\mathrm{\&infin;}\\ 0& else\end{array}\right.$

wherein JMZD_iRepresents the old satisfaction value of host i, whose initial value is set to positive infinity. The intuitive meaning of the formula is that the satisfaction degree is considered to be unchanged initially, so that the attenuation degree is 0, namely the default of the first calculation is set to be 0; whenever the old satisfaction value is less than positive infinity, i.e., not the first calculation, the attenuation is calculated as the ratio of the new and old satisfaction difference divided by the old attenuation. It is noted that the above-mentioned calculation structure may be negative, indicating that the performance of the host is not degraded, but rather improved.

Next, the attenuation degree evaluation unit subjects the MZD to_iValue to JMZD_iIt means that the satisfaction value generated in the current round is used as the old satisfaction value when the performance degradation degree evaluation is performed in the next round.

Finally, the attenuation degree evaluation unit calculates the SJD_iAnd ZDL_iSending the value to a new task allocation scheme generation module to convert the ZDL value to a value for the next task_iThe value is sent to the new task assignment control module.

Step four, generating a new task allocation scheme according to the performance attenuation of each host:

firstly, a new task allocation scheme generation module calculates a marking variable JS_iTo determine whether each host accepts new tasks:

${\mathrm{JS}}_i=\left\{\begin{array}{cc}0& if\frac{{\mathrm{SJD}}_i}{mean\left\{{\mathrm{SJD}}_i\right|0<i\&le;n\}}>b,{\mathrm{SJD}}_i>0,\frac{{\mathrm{ZDL}}_i}{t\&times;mean\left\{{\mathrm{DXZX}}_i\right|0<i\&le;n\}}>c\\ 1& else\end{array}\right.$

wherein b (1 < b <10 is satisfied) and c (1 < c <10 is satisfied) are predetermined parameters. The intuitive meaning of the above calculation is that if the performance attenuation of a certain host is greater than b times of the average performance attenuation of each host, the performance attenuation of the host is positive, and the maximum possible potential task inventory to be processed by the host needs t time which is more than c times of the average equivalent execution rate of each host, the host is considered to be unsuitable for carrying a new task from the viewpoint of performance attenuation and backlog task. The larger the values of b and c are taken, the fewer hosts are determined to be unsuitable for receiving new tasks.

Finally, the new task assignment scheme generation module generates JS_iThe value is sent to the new task assignment control module.

Step five, newly arrived task allocation

Firstly, the new task allocation control module calculates the validity period of the task scheduling policy in the current round:

$YXQ=\left\{\begin{array}{cc}\frac{{\mathrm{\&Sigma;}}_{i=0}^n{\mathrm{ZDL}}_i}{mean\left\{\begin{array}{cc}\frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}& |0<i\&le;n\end{array}\right\}}+t& ifmean\left\{{\mathrm{JS}}_i\right|0<i\&le;n\}=1\\ t& else\end{array}\right.$

the intuitive significance of the formula is that when all the hosts are judged to be suitable for receiving new tasks, the overall performance of the system is stable, so that the validity of the control in the current round can be increased moderately; otherwise, the smaller t time is taken as the effective period length.

At time YXQ, whenever a new task request arrives, the new task assignment control module calculates the update load satisfaction of each host:

${\mathrm{MZD}}_i=\left\{\begin{array}{cc}\frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{ZDL}}_i}& if{\mathrm{ZDL}}_i>0\\ \frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{KWC}}_i\&times;\frac{t}{{\mathrm{KT}}_i}}& else\end{array}\right.$

the new task assignment control module then assigns this newly arrived task to "all-host computation-tagged variable JS_iIs 1 and has a maximum MZD_iHost of value ", while the ZDL of this host getting the new task_iValue, self-increment by 1 (i.e. increase ZDL by one)_iUpdate to old ZDL_iThe value is increased by 1); if all the host computer's mark variable JS_iBoth are 0, this newly arrived task is rejected.

It is worth mentioning that the satisfaction can be directly updated after a task is allocated, and a new task is directly allocated when waiting for the new task. Meanwhile, when a new task comes, the satisfaction degree can be updated, and then the new task is redistributed.

Once the execution time of step five reaches YXQ, step one is entered and a new round of execution of the loop back and forth is started.

The intuitive significance of the operation is that the variable JS is marked_iFirstly, determining the range of alternative hosts which can accept the new task, then selecting the host with the maximum load satisfaction degree among the alternative hosts to accept the new task, and receiving the ZDL of the host after the task_iThe value is increased by 1 to ensure that other hosts have the potential to accept hosts as tasks after the next new task arrives.

The embodiment also provides a task allocation device considering the performance attenuation of the cloud platform host. As shown in fig. 2, the apparatus comprises: the system comprises a cloud system performance monitoring module, a performance attenuation early warning module, a new task allocation scheme generating module and a new task allocation control module.

As shown in fig. 3, the embodiment further discloses a task allocation system considering the performance attenuation of the cloud platform host, which includes a cloud platform server, and a task allocation device considering the performance attenuation of the cloud platform host is disposed in the cloud platform server; the task allocation device considering the performance attenuation of the cloud platform host comprises a cloud system performance monitoring module, a performance attenuation early warning module, a new task allocation scheme generation module and a new task allocation control module.

The cloud system performance monitoring module monitors each host, collects performance information of each host, and runs the host for the total time KT_iTotal number of completed tasks KWC_iAnd the number of tasks completed WC within the last t time_iThe number XD of newly arrived tasks of each host_iNumber of completed tasks for host error or re-submission after suspension CTJ_iNumber of faulty tasks CW_iAnd the number of migrated tasks QC_iThe host downtime record is sent to the performance attenuation early warning module; running the main machine for total time KT_iTotal number of completed tasks KWC_iSending the value to a new task allocation control module; wherein i is a subscript value to indicate the host number to which the variable belongs, and i is more than 0 and less than or equal to n;

the performance attenuation early warning module calculates the equivalent execution rate DXZX_iSending the value to a new task allocation scheme generation module, and calculating the performance attenuation degree SJD_iAnd potential task inventory ZDL_iSending the value to a new task allocation scheme generation module to store the potential task inventory ZDL_iSending the value to a new task allocation control module;

potential task inventory ZDL_iThe calculation formula is as follows:

$\begin{array}{c}{\mathrm{ZDL}}_i=t\&times;\left(\right(1-a)\&times;PJL+max\{\frac{{\mathrm{XD}}_i}{t}|0<i\&le;n\})+{\mathrm{QC}}_i+\\ {\mathrm{CTJ}}_i+\frac{{\mathrm{CW}}_i}{1-{\mathrm{RWSXL}}_i};\end{array}$

wherein a satisfies 0< a <1, and is a weight coefficient preset by a cloud system manager;

load satisfaction MZD_iProportional to the host execution rate, inversely proportional to the potential task stock, and related to the host equivalent survival rate and the number of error tasks; degree of performance decay SJD_iWhen the first calculation is carried out, the default is set to 0, and the non-first calculation is carried out to satisfy the MZD_iJMZD_iJMZD of the difference and old satisfaction_iThe ratio of (A) to (B);

and the new task allocation scheme generation module generates a new task allocation scheme and the new task allocation control module allocates the tasks.

In this embodiment, calculationTo obtain equivalent survival rate CHL_i(ii) a Wherein, HF_iXF the latest time spent in downtime within t_iThe interval time from the recovery moment to the current moment after the latest downtime within the time t;

in this embodiment, calculationObtaining multitask failure rate RWSXL_i；

In this embodiment, calculationGet the equivalent execution rate DXZX_i；

In this embodiment, calculationObtaining an average task input rate PJL;

wherein the number of tasks KWC_iThe initial value is 1; when the total running time of the system is less than t, KT_iIs set to t.

In this embodiment, calculation

And obtaining the load satisfaction degree MZDi.

In this embodiment, the new task allocation scheme has a validity period, and the calculation is performed

And obtaining the validity period of the new task allocation scheme.

In conclusion, the method, the device and the system provided by the invention fully consider the dynamic fluctuation of the host performance in the cloud system, calculate the processing capacity satisfaction degree and the performance attenuation degree of each host closer to the actual task load, evaluate the performance attenuation degree and the expected backlog task quantity of each host, avoid distributing a new task to the host with high performance attenuation degree and more backlog tasks, and enable the task scheduling scheme to be more reasonable. Compared with a static task scheduling management strategy, the method realizes dynamic load balancing and can obtain better effect under the unreliable and high-load system operation environment.

The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.

Claims (10)

1. A task allocation method considering performance attenuation of a cloud platform host is characterized by comprising the following steps:

monitoring each host and collecting performance information of each host; the performance information comprises the total running time KT of the host_iTotal number of completed tasks KWC_iAnd the number of tasks completed WC within the last t time_iThe number XD of newly arrived tasks of each host_iNumber of completed tasks for host error or re-submission after suspension CTJ_iNumber of faulty tasks CW_iAnd the number of migrated tasks QC_iRecording the downtime of the host; wherein i is a subscript value to indicate the host number to which the variable belongs, and i is more than 0 and less than or equal to n;

step two, evaluating the current performance and load of the host: calculating the equivalent survival rate CHL of each host in t time_iMultitask failure rate RWSXL_iEquivalent execution rate DXZX_iAnd average task input rate PJL of all hosts;

evaluating the performance attenuation degree of each host, and evaluating the potential task stock ZDL of each host in the next t time_iEvaluating the load satisfaction degree MZD of each host considering the occurrence of the fault_iEvaluating the degree of performance attenuation SJD of each host_i(ii) a Wherein,

the inventory of potential tasks ZDL_iComprises the following steps:

$\begin{array}{c}{\mathrm{ZDL}}_i=t\&times;\left(\right(1-a)\&times;PJL+max\{\frac{{\mathrm{XD}}_i}{t}|0<i\&le;n\})+{\mathrm{QC}}_i+\\ {\mathrm{CTJ}}_i+\frac{{\mathrm{CW}}_i}{1-{\mathrm{RWSXL}}_i}\end{array}$

wherein a satisfies 0< a <1, and is a weight coefficient preset by a cloud system manager;

the load satisfaction degree MZD_iProportional to the host execution rate, inversely proportional to the potential task stock, and related to the host equivalent survival rate and the number of error tasks; the degree of performance decay SJD_iWhen the first calculation is carried out, the default is set to 0, and the non-first calculation is carried out to satisfy the MZD_iJMZD_iJMZD of the difference and old satisfaction_iThe ratio of (A) to (B);

step four, generating a new task allocation scheme according to the performance attenuation of each host: attenuating the performance by SJD_iIs positive and the degree of performance attenuation SJD_iThe ratio of the average performance attenuation degree of each host is larger than b and the average equivalent execution rate of each host is DXZX_iTo process the potential task inventory ZDL thereof_iThe host computer with the required time larger than c × t is marked as not accepting the new task, otherwise, the host computer is marked as accepting the new task, wherein, b (satisfies 1)<b<10) And c (satisfy 1)<c<10) Presetting parameters for a system manager;

step five, continuously automatically distributing the host to the newly arrived task and updating the load satisfaction degree MZD_i(ii) a When all hosts do not bear tasks, rejecting the newly arrived task; otherwise, the task is executedLoad sharing satisfaction degree MZD_iLargest host, and the ZDL_iIncreasing the value by 1, and updating the load satisfaction degree MZD_i。

2. The method for task allocation according to claim 1, wherein the method comprises the following steps:

the equivalent survival rate CHL_iThe calculation formula is as follows:

wherein, HF_iXF the latest time spent in downtime within t_iThe interval time from the recovery moment to the current moment after the latest downtime within the time t;

the multitask failure rate RWSXL_iComprises the following steps:

${\mathrm{RWSXL}}_i=1-{\left({\mathrm{CHL}}_i\right)}^{{\mathrm{CW}}_i}$

the equivalent execution rate DXZX_iComprises the following steps:

${\mathrm{DXZX}}_i=\left\{\begin{array}{cc}\frac{{\mathrm{WC}}_i-{\mathrm{CTJ}}_i}{{\mathrm{WC}}_i+{\mathrm{QC}}_i}\&times;{(\frac{t}{{\mathrm{WC}}_i}\&times;{\mathrm{CHL}}_i)}^{-1}& if{\mathrm{WC}}_i>0\\ \frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}& elseif{\mathrm{WC}}_i=0\end{array}\right.$

the average task input rate PJL is:

$PJL=\frac{{\mathrm{\&Sigma;}}_{i=0}^n\frac{{\mathrm{XD}}_i}{t}}{n}$

wherein the number of tasks KWC_iThe initial value is 1; when the total running time of the system is less than t, KT_iIs set to t.

3. The method of claim 1, wherein the load satisfaction degree MZD is set to account for cloud platform host performance degradation_iComprises the following steps:

${\mathrm{MZD}}_i=\left\{\begin{array}{cc}\frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{ZDL}}_i}& if{\mathrm{ZDL}}_i>0\\ \frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{KWC}}_i\&times;\frac{t}{{\mathrm{KT}}_i}}& else\end{array}\right..$

4. the method according to claim 1, wherein in the step five, the new task allocation scheme has a validity period, and the validity period of the new task allocation scheme is:

$YXQ=\left\{\begin{array}{cc}\frac{{\mathrm{\&Sigma;}}_{i=1}^n{\mathrm{ZDL}}_i}{mean\left\{\frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}\right|0<i\&le;n\}}+t& ifmean\left\{{\mathrm{JS}}_i\right|0<i\&le;n\}=1\\ t& else\end{array}\right.$

and when the validity period of the new task allocation scheme is reached, executing a jump to execute the step one.

5. A task allocation device considering performance attenuation of a cloud platform host is characterized in that: the system comprises a cloud system performance monitoring module, a performance attenuation early warning module, a new task allocation scheme generation module and a new task allocation control module;

the cloud system performance monitoring module monitors each host, collects performance information of each host, and runs the host for total time KT_iTotal number of completed tasks KWC_iAnd the number of tasks completed WC within the last t time_iThe number XD of newly arrived tasks of each host_iNumber of completed tasks for host error or re-submission after suspension CTJ_iNumber of faulty tasks CW_iAnd the number of migrated tasks QC_iSending the host downtime record to the performance attenuation early warning module; running the main machine for total time KT_iTotal number of completed tasks KWC_iSending the value to the new task allocation control module; wherein i is a subscript value to indicate the host number to which the variable belongs, and i is more than 0 and less than or equal to n;

the performance attenuation early warning module calculates the equivalent execution rate DXZX_iSending the value to a new task allocation scheme generation module, and calculating load satisfaction degree MZD_iThe calculated performance attenuation degree SJD_iAnd potential task inventory ZDL_iSending the value to a new task allocation scheme generation module to store the potential task inventory ZDL_iSending the value to a new task allocation control module;

the inventory of potential tasks ZDL_iThe calculation formula is as follows:

$\begin{array}{c}{\mathrm{ZDL}}_i=t\&times;\left(\right(1-a)\&times;PJL+max\{\frac{{\mathrm{XD}}_i}{t}|0<i\&le;n\})+{\mathrm{QC}}_i+\\ {\mathrm{CTJ}}_i+\frac{{\mathrm{CW}}_i}{1-{\mathrm{RWSXL}}_i}\end{array};$

wherein a satisfies 0< a <1, and is a weight coefficient preset by a cloud system manager;

the load satisfaction degree MZD_iProportional to the host execution rate, inversely proportional to the potential task stock, and related to the host equivalent survival rate and the number of error tasks; the degree of performance decay SJD_iWhen the first calculation is carried out, the default is set to 0, and the non-first calculation is carried out to satisfy the MZD_iJMZD_iJMZD of the difference and old satisfaction_iThe ratio of (A) to (B);

and the new task allocation scheme generation module generates a new task allocation scheme and the new task allocation control module allocates the tasks.

6. The task allocation device according to claim 5, wherein the task allocation device takes into account the performance degradation of the cloud platform host computer:

the equivalent survival rate CHL_iThe calculation formula is as follows:

wherein, HF_iXF the latest time spent in downtime within t_iThe interval time from the recovery moment to the current moment after the latest downtime within the time t;

the multitask failure rate RWSXL_iThe calculation formula is as follows:

${\mathrm{RWSXL}}_i=1-{\left({\mathrm{CHL}}_i\right)}^{{\mathrm{CW}}_i};$

the equivalent execution rate DXZX_iThe calculation formula is as follows:

${\mathrm{DXZX}}_i=\left\{\begin{array}{cc}\frac{{\mathrm{WC}}_i-{\mathrm{CTJ}}_i}{{\mathrm{WC}}_i+{\mathrm{QC}}_i}\&times;{(\frac{t}{{\mathrm{WC}}_i}\&times;{\mathrm{CHL}}_i)}^{-1}& if{\mathrm{WC}}_i>0\\ \frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}& elseif{\mathrm{WC}}_i=0\end{array}\right.;$

the average task input rate PJL is calculated by the formula:

$PJL=\frac{{\mathrm{\&Sigma;}}_{i=0}^n\frac{{\mathrm{XD}}_i}{t}}{n};$

wherein the number of tasks KWC_iThe initial value is 1; when the total running time of the system is less than t, KT_iIs set to t.

7. The device of claim 5, wherein the load satisfaction degree MZD is set to account for performance degradation of the cloud platform host_iComprises the following steps:

${\mathrm{MZD}}_i=\left\{\begin{array}{cc}\frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{ZDL}}_i}& if{\mathrm{ZDL}}_i>0\\ \frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{KWC}}_i\&times;\frac{t}{{\mathrm{KT}}_i}}& else\end{array}\right.;$

the new task allocation scheme has a validity period, and the calculation formula of the validity period of the new task allocation scheme is as follows:

$YXQ=\left\{\begin{array}{cc}\frac{{\mathrm{\&Sigma;}}_{i=1}^n{\mathrm{ZDL}}_i}{mean\left\{\frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}\right|0<i\&le;n\}}+t& ifmean\left\{{\mathrm{JS}}_i\right|0<i\&le;n\}=1\\ t& else\end{array}\right.$

8. a task allocation system considering performance attenuation of a cloud platform host comprises a cloud platform server and is characterized in that: a cloud platform host performance attenuation task distribution device is arranged in the cloud platform server; the cloud platform host performance attenuation task allocation device comprises a cloud system performance monitoring module, a performance attenuation early warning module, a new task allocation scheme generation module and a new task allocation control module;

the cloud system performance monitoring module monitors each host, collects performance information of each host, and runs the host for total time KT_iTotal number of completed tasks KWC_iAnd the number of tasks completed WC within the last t time_iThe number XD of newly arrived tasks of each host_iNumber of completed tasks for host error or re-submission after suspension CTJ_iNumber of faulty tasks CW_iAnd the number of migrated tasks QC_iSending the host downtime record to the performance attenuation early warning module; running the main machine for total time KT_iTotal number of completed tasks KWC_iSending the value to the new task allocation control module; wherein i is a subscript value to indicate the host number to which the variable belongs, and i is more than 0 and less than or equal to n;

the performance attenuation early warning module calculates the equivalent execution rate DXZX_iSending the value to a new task allocation scheme generation module, and calculating the performance attenuation degree SJD_iAnd potential task inventory ZDL_iSending the value to a new task allocation scheme generation module to store the potential task inventory ZDL_iSending the value to a new task allocation control module;

the inventory of potential tasks ZDL_iThe calculation formula is as follows:

${\mathrm{ZDL}}_i=t\&times;\left(\right(1-a)\&times;PJL+max\{\frac{{\mathrm{XD}}_i}{t}|0<i\&le;n\})+{\mathrm{QC}}_i+$

${\mathrm{CTJ}}_i+\frac{{\mathrm{CW}}_i}{1-{\mathrm{RWSXL}}_i};$

wherein a satisfies 0< a <1, and is a weight coefficient preset by a cloud system manager;

the load satisfaction degree MZD_iProportional to the host execution rate, inversely proportional to the potential task stock, and related to the host equivalent survival rate and the number of error tasks; the degree of performance decay SJD_iWhen the first calculation is carried out, the default is set to 0, and the non-first calculation is carried out to satisfy the MZD_iJMZD_iJMZD of the difference and old satisfaction_iThe ratio of (A) to (B);

and the new task allocation scheme generation module generates a new task allocation scheme and the new task allocation control module allocates the tasks.

9. The system of claim 8, wherein the task allocation system is configured to account for performance degradation of the cloud platform hosts and is further configured to:

the equivalent survival rate CHL_iThe calculation formula is as follows:

wherein, HF_iXF the latest time spent in downtime within t_iThe interval time from the recovery moment to the current moment after the latest downtime within t time；

The multitask failure rate RWSXL_iThe calculation formula is as follows:

${\mathrm{RWSXL}}_i=1-{\left({\mathrm{CHL}}_i\right)}^{{\mathrm{CW}}_i};$

the equivalent execution rate DXZX_iThe calculation formula is as follows:

${\mathrm{DXZX}}_i=\left\{\begin{array}{cc}\frac{{\mathrm{WC}}_i-{\mathrm{CTJ}}_i}{{\mathrm{WC}}_i+{\mathrm{QC}}_i}\&times;{(\frac{t}{{\mathrm{WC}}_i}\&times;{\mathrm{CHL}}_i)}^{-1}& if{\mathrm{WC}}_i>0\\ \frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}& elseif{\mathrm{WC}}_i=0\end{array}\right.;$

the average task input rate PJL is calculated by the formula:

$PJL=\frac{{\mathrm{\&Sigma;}}_{i=0}^n\frac{{\mathrm{XD}}_i}{t}}{n};$

wherein the number of tasks KWC_iThe initial value is 1; when the total running time of the system is less than t, KT_iIs set to t.

10. The system of claim 8, wherein the load satisfaction metric MZD is defined as_iComprises the following steps:

${\mathrm{MZD}}_i=\left\{\begin{array}{cc}\frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{ZDL}}_i}& if{\mathrm{ZDL}}_i>0\\ \frac{(1-{\left({\mathrm{CHL}}_i\right)}^{mean\left\{{\mathrm{CW}}_i\right|0<i<<n\}})\&times;{\mathrm{DXZX}}_i}{{\mathrm{KWC}}_i\&times;\frac{t}{{\mathrm{KT}}_i}}& else\end{array}\right.;$

the new task allocation scheme has a validity period YXQ, and the new task allocation scheme validity period calculation formula is as follows:

$YXQ=\left\{\begin{array}{cc}\frac{{\mathrm{\&Sigma;}}_{i=1}^n{\mathrm{ZDL}}_i}{mean\left\{\frac{{\mathrm{KWC}}_i}{{\mathrm{KT}}_i}\right|0<i\&le;n\}}+t& ifmean\left\{{\mathrm{JS}}_i\right|0<i\&le;n\}=1\\ t& else\end{array}\right..$