JPH10198640A - Load control system for processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a load control system capable of improving the throughput of the whole system and guaranteeing the throughput of specific operation processing. SOLUTION: A parallel computer system 10 is composed of plural processors 10-11 to 10-MN. The aggregate of processors is defined as a node set to define plural node sets. An evaluation equation expressed by 'evaluation value E = [(threshold T - free guarantee value α) - load L] is set up in each node set, and at the time of allocating jobs, the evaluation value E is found out based on load L measured in each processor, jobs are allocated to processors successively from the processor having the highest evaluation value E, and when the evaluation value E of a processor shared by another node set is <=0, a new job 15 not allocated to the processor.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a load control method for a processor, and more particularly to a load control method for a processor suitable for application to a parallel computer system.

[0002]

2. Description of the Related Art In a conventional parallel computer system, as a method of controlling the load balance of a processor, a method of defining a set of processors to be used for a specific purpose is known. According to this method, in a parallel computer system, a set of processors used for each specific purpose is defined, and a shared processor group that can be used for a plurality of purposes is defined, thereby reducing the amount of other business processing. When the number of processors is small, the available processor resources can be used to improve the throughput of the job. However, in this method, when the processing amount of a certain task increases, the shared processor of another task cannot be used.

On the other hand, as a method for improving the throughput of the entire system, for example, Japanese Patent Laid-Open No. Hei 8-30
As described in Japanese Patent Application Laid-Open No. 560, there is known a method of obtaining a system load state using a load of each processor and a weight coefficient corresponding to each processor, limiting processing based on the load state, and controlling the load of the processor. Have been.

[0004]

However, in the method described in Japanese Patent Application Laid-Open No. Hei 8-30560, there is no means for securing a certain processing capacity for each node set, and the throughput for each business process cannot be guaranteed. was there.

An object of the present invention is to provide a processor load control method capable of improving the throughput of the entire system and guaranteeing the throughput of a specific business process.

[0006]

In order to achieve the above object, according to the present invention, a set of processors among a plurality of processors constituting a parallel computer system is defined as a node set. In the processor load control method that assigns a job to each processor in each node set based on the weight coefficient based on the weight coefficient, for each node set in a plurality of node sets, , A dedicated job guarantee value is set, and no new job assignment exceeding the above-mentioned exclusive job guarantee value is assigned to a processor shared with another node set. And the throughput of a specific business process can be guaranteed.

In the above processor load control method,
Preferably, an evaluation formula given by an evaluation value E = ((threshold value T−guarantee value α) −load L) × weight coefficient W is set for each node set. An evaluation value E is obtained based on the measured load L, and the job is allocated in the order of the processor having the higher evaluation value E, and the evaluation value E is assigned to a processor shared with another node set. When the value E is equal to or less than zero, new job assignment is not performed.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A processor load control method according to an embodiment of the present invention will be described below with reference to FIGS. First, an overall configuration of a parallel computer system using a processor load control method according to an embodiment of the present invention will be described with reference to FIG. FIG. 1 is a block diagram of a parallel computer system using a processor load control method according to an embodiment of the present invention.

The parallel computer 10 includes MN processors 1
0-11, 10-12, ..., 10-1N, 10-21,
..., 10-2N, ..., 10-M1, ..., 10-MN,
And a network 12 connecting them. The parallel computer 10 is connected to n external client terminals 30-1 and 30-n via a network 20.

The job execution request is input from the external client terminal 30 to the processor 10-MN in the parallel computer 10, and is assigned to the processors 10-11 to 10-MN and executed. The assignment method will be described later with reference to the flowchart shown in FIG.

Next, the software configuration of each processor will be described with reference to FIG. FIG. 2 is an explanatory diagram of the software configuration of each processor constituting the parallel computer system using the processor load control method according to one embodiment of the present invention.

Each processor 10-11, 10-12,
..., 10-1N, 10-21, ..., 10-2N, ..., 1
Each of 0-M1,..., 10-MN includes a processor load collection daemon 13. The processor load collection daemon 13 periodically measures the load of each processor, and compares the measured load of each processor with the processor 10.
-Send to MN.

Next, referring to FIG. 3, the processor 10 to which a job execution request is input from the external client terminal 30 will be described.
-The software configuration of the MN will be described. FIG.
FIG. 4 is an explanatory diagram of a software configuration of a processor to which one job execution request is input among processors configuring a parallel computer system using a processor load control method according to an embodiment of the present invention.

The external client terminals 30-1,..., 30
Processor 10 to which a job execution request is input from -n
The MN includes a scheduler 15, a node set definition table 16, a job queue 17, and an evaluation expression 18.

The scheduler 15 receives a job execution request from the external client terminals 30-1,..., 30-n, and uses the node set definition table 16 and the evaluation expression 18 to execute a job in the processors 10-11 to 10-MN. , A processor to which a job is assigned is determined, and the job is assigned.

As will be described later with reference to FIG. 4, the node set definition table 16 defines a node set composed of a plurality of processors and defines a weight coefficient to be assigned to a processor belonging to each node set in advance. It is.

In the job queue 17, job requests received from the external client terminals 30-1,..., 30-n are registered as queues.

The evaluation expression 18 is defined for each node set, and is given in advance by the following expression. How to use the evaluation expression 18 will be described later with reference to the flowchart shown in FIG.

Evaluation value E = ((T−α) −L) × W Here, T (%) is a threshold value. The threshold T is
It shows the maximum value of the CPU utilization of each of the processors 10-11 to 10-MN, and also shows the maximum load that can be assigned. The maximum value of the threshold value T is 100%. That is, a load of 100% or more of the processing capacity is not given to each processor. As the threshold T,
Any value can be set, for example, 90%
Is set. When the threshold value is set to 90%, a new job is not allocated when the CPU utilization of a certain processor exceeds 90%. However, for example, when the CPU utilization is 85%,
As a result of the assignment, the CPU utilization is allowed to exceed 90%. To the last, the threshold value is a numerical value used as a guide for judging whether or not to allocate a new job.

Α (%) is a vacant value. That is, when a certain processor x is used in common by the first node set and the second node set, the empty guarantee value α is used in the evaluation expression 18 defined for the first node set.
Is a value that guarantees free processing load on the second node set. (T-α) indicates a guaranteed value that can be exclusively used for the first node set. For example, in the evaluation formula 18 defined for the first node set, if the vacancy guarantee value α is 30%, the vacancy of the processing load for the second node set is guaranteed 30%. If the threshold value T is 90%, it indicates that the exclusive use of (T-α) = 60% is guaranteed for the first node set.

Further, L (%) is the value of each processor 10-
11 to 10-MN, and each processor 1
0-11 to 10-MN processor load collection daemon 1
3 is the processor load measured.

W is a weighting factor set for each processor for each node set.
It takes a value of 100 or more and 100 or less. Note that a processor having a weight coefficient of “0” is not included in the node set.

Next, the configuration of the node set definition table will be described with reference to FIG. FIG. 4 is an explanatory diagram of a node set definition table provided in a processor to which one job execution request is input among processors constituting a parallel computer system using a processor load control method according to an embodiment of the present invention.

The horizontal axis of the node set definition table 16 is
A processor is shown, and in the illustrated example, the processor is configured by MN processors 10-11 to 10-MN. The vertical axis of the node set definition table 16 indicates node sets 1 to L. And each node set 1
, W1-MN, W2-11,..., W2-M for each processor 10-11 to 10-MN.
N, WL-11, ..., WL-MN are set. Weight coefficient W
Takes a value of 0 or more and 100 or less as described above, but if it is 0, it indicates that it is not included in the node set.

Here, a specific example of the node set definition table will be described with reference to FIG. FIG. 5 is an explanatory diagram of an example of a node set definition table using a processor load control method according to an embodiment of the present invention.

It is assumed that the horizontal axis of the node set definition table 16 'includes 12 processors 10-11 to 10-34. It is assumed that the vertical axis of the node set definition table 16 'is configured by two node sets 1 and 2.

The processors 10-11 and 10-1
Weights of “100” are set for Node Set 1 for 2, 10-13, and 10-14. Processor 10-21, 10-22, 10-23, 10-2
4 is assigned a weight of “50” for the node set 1. Processors 10-31 and 10-3
Weights of “0” are set for node sets 1 for 2, 10-33, and 10-34. That is, the node set 1 includes the processors 10-11, 10-12,
10-13, 10-14 and processor 10-21, 1
It shows that the processor is composed of eight processors 0-22, 10-23, and 10-24. Processor 10-3
The weights for 1, 10-32, 10-33, and 10-34 are "0", indicating that these processors are not included in the node set 2.

The node set 2 includes the processor 10
-21, 10-22, 10-23, 10-24 and processors 10-31, 10-32, 10-33, 10-3
4 is composed of eight processors.
Processors 10-11, 10-12, 10-13, and 10
Since the weight for −14 is “0”, this indicates that these processors are not included in the node set 1.

Here, for convenience of explanation, the processor 10-
11, 10-12, 10-13, and 10-14 are referred to as a processor group G1, and processors 10-21 and 10-2.
2, 10-23, and 10-24 are referred to as a processor group G2, and the processors 10-31, 10-32, 10-33,
10-34 is referred to as a processor group G3. The processor group G1 is an exclusive processor of the node set 1, and
The processor group G3 is an exclusive processor of the node set 2. However, the processor group G2 has
This is a processor commonly used by the node set 1 and the node set 2.

For example, node set 1 may be used for interactive login purposes, and node set 2 may be used for batch purposes. That is, the processor group G1 can be used only for the purpose of interactive login, and the processor group G3 can be used only for the purpose of batch. The processor group G2 can be used for interactive login as well as for batch purposes.

Next, a specific example of the evaluation expression 18 described with reference to FIG. 3 will be described. Here, for the node set 1 and the node set 2 shown in FIG.
Each shall be given as follows.

That is, the evaluation value E = ((T−α) −L) × W
, The threshold value T for the node set 1 and the node set 2 is 90%, the guaranteed vacancy value α for the node set 1 is 30%, and the guaranteed vacancy value α for the node set 2 is 60%. Evaluation value E1 = ((90−30) −L) × W1 = (60−L) × W1 and evaluation value E2 for node set 2 = ((90−60) −L) × W2 = (30−L) ) × W2. Here, the weight coefficient W1 and the weight coefficient W2 are values set according to FIG. Therefore, when the load of each processor is measured by the processor load collection daemon 13 shown in FIG. 2, the evaluation value of each processor can be obtained based on the above equation.

Next, job information registered in the job queue 17 will be described with reference to FIG. FIG. 6 is an explanatory diagram of job information registered in a job queue using a processor load control method according to an embodiment of the present invention.

The job table 17 'having job information registered in the job queue 17 is composed of a "node set to be used", a "number of required processors", and a "path name of an execution module". Job table 1
7 'is created for each job requested to be executed.

A "node set to be used" is registered in advance according to the purpose of the job whose execution is requested.
In addition, the number of processors required to execute the job is registered in advance as “the required number of processors”.

Next, a job assignment process based on a processor load control method according to an embodiment of the present invention will be described with reference to FIGS. FIG. 7 is a flowchart illustrating a flow of a job assignment process based on a processor load control method according to an embodiment of the present invention. FIG. 8 is a diagram illustrating a load and evaluation based on a processor load control method according to an embodiment of the present invention. FIG. 4 is an explanatory diagram of values.

In step 100 of FIG. 7, when there is a job request in the job queue 17, the scheduler 15 of the processor 10-MN shown in FIG.

Here, the node set definition table is
For example, as shown in FIG. 5, it is assumed that the setting is made for the node set 1 and the node set 2. Also,
The node set used by the job request at the head of the job queue 17 is node set 1. And
Assume that the evaluation value E1 for the node set 1 is obtained by E1 = (60−L) × W1, as described above. Here, the processor 10-1
In the initial state, no job is executed at all, and the load L is “0”. The processor load L is measured using the processor load collection daemon 13 existing in each of the processors 10-11 to 10-34.

First, in step 102, the scheduler 15 refers to the node set definition table 16 and evaluates the evaluation expression 18 for the node set 1 used by the execution request job registered in the job queue 17. , An evaluation value E is obtained.

As shown in FIG. 5, each processor 10-11, 10-12, 10-13, 1 of the processor group G1
Since the weight coefficient W of 0-14 is “100”, the evaluation value EG1 for each processor of the processor group G1 is
It is “60 × 100”. Further, each processor 10-21, 10-22, 10-23, and 10 in the processor group G2.
Since the weight coefficient W of −18 is “50”, the evaluation value EG2 for each processor of the processor group G1 is “60”.
× 50 ”. This relationship is shown in FIG.

That is, in FIG. 8, in the first stage ST1 of the initial state, each processor 1 of the processor group G1
The load LG1 of 0-11, 10-12, 10-13, and 10-14 and each processor 10-21 of the processor group G2
Loads LG2 of 10-22, 10-23, and 10-24 are each set to “0”. The evaluation values EG1 and EG2 for the respective processor groups G1 and G2 are obtained as “60 × 100” and “60 × 50”, respectively.

In step 104, scheduler 1
5 indicates that the processors are counted in descending order of the value (evaluation value) of the evaluation expression, and the processor having the highest evaluation value is pointed. That is, as evaluated in step 102, the evaluation value of the processor group G1 is higher than the evaluation value of the processor group G2, and therefore, the processors 10-11, 10-12, 10-13, which constitute the processor group G1. Point to any processor in 10-14. Processor 1
Since the evaluation values of 0-11, 10-12, 10-13, and 10-14 are all the same, any processor may be pointed. For example, the processor may be pointed in the order in which the processors are evaluated. For example, the processor 10-11 is pointed.

Next, at step 106, the scheduler 15 determines whether the evaluation value of the processor pointed at at step 104 is "0" or less. When the evaluation value is equal to or less than "0", the process proceeds to step 108, where "
If it is not equal to or less than 0, the process proceeds to step 110. Here, since the pointed evaluation value of the processor 10-11 is larger than "0", the process proceeds to step 110 for explanation. "The following cases will be described later.

In step 110, scheduler 1
5 judges whether or not the CPU utilization rate of the pointed processor is equal to or more than a threshold value. The CPU utilization is equal to the load determined by the processor load collection daemon 13. If the CPU utilization is equal to or higher than the threshold, the process proceeds to step 120; otherwise, the process proceeds to step 112. If the CPU utilization is equal to or greater than the threshold value, no more jobs can be allocated to the corresponding processor, so the process proceeds to step 120, and the request is returned to the queue.
The request assignment process ends.

The processing from step 102 is repeated to point to another assignable processor.

Here, it is assumed that no job has been assigned to the processor 10-11,
Since the PU utilization (load) is “0”, the threshold T
Since it is lower than “90%” set as
It is assumed that the process proceeds to step 12.

Next, in step 112, the scheduler 15 increments the counter by "1". This counter has been cleared to "0" at the beginning of the process. The counter counts up the number of pointed processors, and counts the “number of required processors” in the job table 17 ′ shown in FIG.

At step 114, scheduler 1
In step 5, it is determined whether or not the value of the counter counted in step 112 is smaller than the required number of processors indicated by "number of required processors" in the job table 17 'shown in FIG. If the value of the counter is smaller than the required number of processors, the process proceeds to step 116, and if it is equal to or greater than the required number of processors, the process proceeds to step 122.

When the value of the counter becomes equal to the required number of processors, it means that the processor required for job assignment has been pointed, and the process proceeds to step 122. In step 122, the scheduler 15 assigns a job to the selected processor. At this time, “the execution module path name” of the job table 17 ′ shown in FIG. 6 is given to the assigned processor.

Here, the number of requested processors is assumed to be "4", and the description of the processing flow will be continued.
In step 114, the value of the counter is “1”,
Since the number of requested processors is set to “4”, step 11
Proceed to 6.

At step 116, scheduler 1
5 determines whether the processor is still left.
If there is any remaining, the process proceeds to step 118, and if not, the process proceeds to step 120.

If no request remains, the process proceeds to step 120, where the scheduler 15 returns the request to the queue. In step 124, the process of allocating the request is temporarily terminated.

Here, it is assumed that the processor other than the processor 10-11 is still left, so that the process proceeds to step 118. In step 118, the scheduler 15 selects the pointed processor, and then points to the processor having the higher evaluation expression value (evaluation value). Here, the processors 10-12, 10-13, and 10-1
4 have the same evaluation value, so any one of them is pointed out.

Thereafter, from step 110 to step 11
8 is repeated, and the processors 10-11, 10-12,
Four processors 10-13 and 10-14 are pointed out. When the points of the four processors are completed, in step 114, the value of the counter becomes equal to the required number of processors, so the process proceeds to step 122.

In step 122, scheduler 1
5 assigns a job to the selected processor. That is, the four processors 10-11, 10-12, and 10-
The execution request jobs are assigned to 13, 10-14.

Next, a description will be given assuming that another job execution request J2 has been made and the node set to be used is node set 1. Here, it is assumed that the state of the load of each processor is in the second stage ST2 in FIG. That is, the processors 10-11 and 10- of the processor group G1
At 12, 10-13, and 10-14, a job has already been allocated and executed, and the load LG1 at that time is:
30%. Here, for convenience of explanation, processors 10-11, 10-12, and 10-1 of processor group G1 are used.
The load L of 3, 10-14 is the same, and 3
The description will be made assuming that it is 0%.

For the new job execution request J2, in step 102, the evaluation formula 18 is applied to the node set 1 used by the execution request job registered in the job queue 17 with reference to the node set definition table 16. Evaluation is performed to obtain an evaluation value E.

As shown in FIG. 8, since the load LG1 of the processor group G1 is set to 30%, the evaluation value EG1 for each processor of the processor group G1 is "30 × 100". The evaluation value EG2 for each processor in the processor group G2 is “60 × 50”. That is, the evaluation value EG1 for each processor in the processor group G1 is equal to the evaluation value EG2 for each processor in the processor group G2.

At step 104, scheduler 1
5 indicates that the processors are counted in descending order of the value (evaluation value) of the evaluation expression, and the processor having the highest evaluation value is pointed. That is, since the evaluation value of the processor group G1 is equal to the evaluation value of the processor group G2 as evaluated in step 102, the processors 10-11, 10-12, and 10-1 constituting the processor group G1 or the processor group G2.
3,10-14,10-21,10-22,10-2
Points to any processor among 3, 10-24. If the evaluation value of each processor is the same, any processor may be pointed.

As described above, when a job is not assigned to either the processor group G1 or the processor group G2, the job is preferentially assigned to the processor group G1 having the larger weight coefficient W. Processors 10-21 and 10-2 configuring processor group G2
2, 10-23 and 10-24 are a processor group shared with the node set 2 which is another node set, and therefore, the weight count W is made smaller with respect to the processor group G1, so that the initial job allocation is performed. Is not done. However, the evaluation value of the processor group G2 is
When the value becomes equal to or more than the evaluation value of the processor group G1, the job is also assigned to the processor group G2. That is, the throughput for the processor groups G1 and G2 is improved.

For the job assignment to both the processor group G1 and the processor group G2, each evaluation value E is "0".
Is executed until Next, a case where the evaluation values EG1 and EG2 of the processor group G1 and the processor group G2 become "0" will be described with reference to FIG.

In the third stage ST3 of FIG. 8, for example,
When the load LG1 of the processor group G1 becomes "60%", the evaluation value EG1 becomes "0" at that time. When the load LG2 of the processor group G2 becomes "60%", the evaluation value EG2 becomes "0" at that time. In this state, when a new job execution request is made, the scheduler 15 obtains the evaluation values E of the processor groups G1 and G2 in step 102, and determines the processor having the highest evaluation value in step 104. Point. Here, the subsequent processing differs depending on whether the pointed processor is the processor of the processor group G1 or the processor of the processor group G2. When the processor of the processor group G1 is pointed, the scheduler 15 determines in step 106 that the evaluation value is “
It is determined that it is 0 "or less, and the process proceeds to step 108.

In step 108, scheduler 1
Step 5 determines whether the pointed processor is shared with another node set. That is, the processor group G
As will be understood from the node set definition table 16 shown in FIG. 5, the first processor proceeds to step 110 because it is not shared with another node set. Job allocation is performed until the usage rate exceeds the threshold.
That is, the processor group G1 that is not shared with another node set
As shown in the fourth stage ST4 of FIG. 8, a new assignment becomes possible until the load LG1 of the processor group G1 becomes "90%". Here, when the load LG1 of the processor group G1 is, for example, "89%", a new assignment is possible, so that the load LG1 of the processor group G1 can be up to 100%.

However, when the processor of the processor group G2 is pointed, the scheduler 15
In step 106, it is determined that the evaluation value is “0” or less, and the process proceeds to step 108.

In step 108, scheduler 1
5 determines whether or not the processor pointed to is shared with another node set, and the processor of the processor group G2 executes the node set definition table 16 shown in FIG.
As will be understood from, the request is returned to the queue because it is shared with the other node set 2 and the process goes to step 120. That is, the new assignment of the job to the processor group G2 shared with the other node set is until the evaluation value EG2 is larger than "0". That is, as shown in the third stage ST3 of FIG. 8, the load LG2 of the processor group G2 reaches "60%". Here, the processor group G
When the load LG2 of the processor group G2 is, for example, "59%", a new assignment is possible, and the load LG2 of the processor group G2 is
Can be up to about 70%. That is, the throughput for the business purpose of the node set 1 of the processor group G2 shared with another node set is guaranteed up to 60%. This numerical value is equivalent to the above-described occupation guarantee value (T-α).

Similarly, when looking at the node set 2, in the evaluation formula ((T−α) −L) × W, the threshold value T is set to “90%” and the vacant value α is set to “60%”. Then, for the node set 2, the processor group G2
Means that occupancy is guaranteed up to (T−α) = 30%.

As described above, in the same node set, by setting the weight coefficient W, when the load is small, the job is assigned preferentially to the processor with the larger weight coefficient, and the evaluation values become equal. After that, by executing job assignment to each processor, the throughput of the entire system can be improved.

For a processor shared with another node set, a guaranteed exclusive value (T-α) for the own node set is set. The allocation can be made possible. On the other hand, when viewed from the other node set side, the vacancy guarantee value α can secure the exclusive use, so that the throughput of a specific business process is guaranteed.

According to this embodiment, the throughput of the entire system is improved, and the throughput of a specific business process can be guaranteed.

[0070]

According to the present invention, in the processor load control method, the throughput of the entire system can be improved and the throughput of a specific business process can be guaranteed.

[Brief description of the drawings]

FIG. 1 is a block diagram of a parallel computer system using a processor load control method according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a software configuration of each processor constituting a parallel computer system using a processor load control method according to an embodiment of the present invention.

FIG. 3 is an explanatory diagram of a software configuration of a processor to which a job execution request is input among the processors constituting the parallel computer system using the processor load control method according to the embodiment of the present invention;

FIG. 4 is an explanatory diagram of a node set definition table provided in a processor to which a job execution request is input among processors constituting a parallel computer system using a processor load control method according to an embodiment of the present invention;

FIG. 5 is an explanatory diagram of an example of a node set definition table using a processor load control method according to an embodiment of the present invention.

FIG. 6 is an explanatory diagram of job information registered in a job queue using a processor load control method according to an embodiment of the present invention.

FIG. 7 is a flowchart illustrating a flow of a job allocation process based on a processor load control method according to an embodiment of the present invention.

FIG. 8 is an explanatory diagram of loads and evaluation values based on a processor load control method according to an embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 ... Parallel computer system 10-11, ... 10-MN ... Processor 12, 20 ... Network 30-1, ..., 30-n ... External client terminal 13 ... Processor load collection daemon 15 ... Scheduler 16 ... Node set definition table 17 ... Job queue 18 ... Evaluation formula

Claims (2)

[Claims] An aggregation of processors among a plurality of processors constituting a parallel computer system is defined as a node set, a weight coefficient is set for each processor in the node set, and a weight coefficient is set based on the weight coefficient. In the processor load control method of allocating a job to each processor in each node set, an exclusive use guarantee value is set for each node set in a plurality of node sets, and is shared with other node sets. For the processor,
A processor load control method, wherein a new job assignment exceeding the exclusive use guarantee value is not performed. 2. The processor load control method according to claim 1, wherein an evaluation value E = ((threshold value T−empty guarantee value α) −load L) × weight coefficient W is given for each node set. An evaluation formula is set, and at the time of job assignment, an evaluation value E is obtained based on the load L measured for each processor. The jobs are assigned in the order of the processor having the higher evaluation value E, and other jobs are assigned. For processors shared with the node set,
If the evaluation value E is equal to or less than zero, no new job assignment is performed.