Method and apparatus for performing change-over control to processor groups by using rate of failed processors in a parallel computer

• US 5,652,833 A
• Filed: 03/09/1995
• Issued: 07/29/1997
• Est. Priority Date: 03/15/1994
• Status: Expired due to Fees

• Current Assignee: Hitachi System Engineering Limited (Hitachi, Ltd.)
• Original Assignee: Hitachi Information Networks Limited, Hitachi, Ltd., Hitachi System Engineering Limited (Hitachi, Ltd.)
• Inventors: Tanaka, Natsuro, Takizawa, Mitsuyoshi, Meguro, Yasushi, Minamisawa, Akinori
• Primary Examiner(s): Beausoliel, Jr., Robert W.
• Assistant Examiner(s): Elmore, Stephen C.
• Application Number:
  US08/401,261
• Time in Patent Office: 873 Days
• Field of Search: 395/182.08, 395/375, 395/650, 395/700, 371/72, 364/474.16, 364/474.19, 364/551.01, 364/228.3, 364/230, 364/230.3, 364/264, 364/264.4, 364/268, 364/268.3, 364/268.9, 364/269.2, 364/281.9 , 364/285.3, 364/944.61, 364/949.4
• US Class Current: 714/10
• CPC Class Codes: G06F 11/2028 eliminating a faulty processor or activating a spare
• CPC Class Codes: G06F 11/2041 with more than one idle spare processing component

Abstract

In a change-over control method for a parallel processor system including a current processor group having a plurality of processors and a network connecting the processors to each other and a standby processor group configured in the same way as for the current processor group, a processor control section is disposed in the parallel processor system, and a monitor processor is arranged for each of the current and standby processor groups. A faulty processor ratio determined according to the amount of job processing is set to the processor control section. On receiving a report notifying occurrence of a failure in a processor from the monitor processor disposed in the current processor group, the processor control section determines a ratio of failed processors in the current processor group. When the ratio is equal to or more than the faulty processor ratio, the processor control section effects a change-over operation of transferring job processing from the current processor group to the standby processor group.

Claims

1. A parallel processor system, including a current processor group having a plurality of processors and a network connecting the processors to each other and a standby processor group including a plurality of processors and a network connecting the processors to each other, said parallel processor system comprising:

a monitor processor disposed in said current processor group and a monitor processor disposed in said standby processor group, said monitor processors controlling processing information of all processors in said current and standby processor groups, respectively;

a processor control section for executing a change-over operation from said current processor group to said standby processor group;

an address control table for storing therein physical and logical addresses of each processor of each of said current and standby processor groups with correspondence established therebetween;

a scheduling table for setting therein a faulty processor ratio determined according to an amount of job processing in said parallel processor system;

a storage section for storing therein processing information of each processor reported at a predetermined point of time from a monitor table disposed in said current processor group;

a change-over control section for calculating a ratio of failed processors in said current processor group at a pertinent point of time and comparing the calculated ratio with the faulty processor ratio set in the scheduling table when a report notifying occurrence of a failure in a processor is received from said monitor processor disposed in said current processor group, and executing a change-over operation of transferring the job processing from said current processor group to said standby processor group when the calculated ratio is equal to or more than the faulty processor ratio; and

an operator'"'"'s console connected to said processor control section for arbitrarily setting therefrom the faulty processor ratio to the scheduling table.

2. A parallel processor system according to claim 1, wherein said processor control section includes:

a timer for measuring time;

a scheduling table for having set therein, at a predetermined interval time, a faulty processor ratio determined according to an amount of job processing in said parallel processor system; and

a change-over control section for calculating a ratio of failed processors in said current processor group at a pertinent point of time and comparing the calculated ratio with the faulty processor ratio set in the scheduling table corresponding to a point of time indicated by said timer when a report notifying occurrence of a failure in a processor is notified from said monitor processor disposed in said current processor group, and conducting a change-over operation of transferring the job processing from said current processor group to said standby processor group when the ratio is equal to or more than the faulty processor ratio.

3. A parallel processor system according to claim 1, wherein said processor control section further includes:

a counter for counting the number of failed processors when a report notifying occurrence of a failure in a processor is received from said monitor processor;

a scheduling table for setting therein the number of faulty processors determined according to an amount of job processing in said parallel processor system; and

a change-over control section for comparing the number of failed processors counted by the counter with the number of faulty processors set in the scheduling table when a report notifying occurrence of a failure in a processor is received from said monitor processor disposed in said current processor group and conducting a change-over operation of transferring the job processing from said current processor group to said standby processor group when the counted number of failed processors is equal to or more than the number of faulty processors.

4. A parallel processor system according to claim 3, wherein said processor control section further includes:

a timer for measuring time;

a scheduling table for setting therein, at a predetermined interval of time, the number of faulty processors determined according to an amount of job processing in said parallel processor system; and

a change-over control section for comparing the number of faulty processors set in the scheduling table corresponding to a point of time indicated by said timer with the number of failed processors counted by the counter when a report notifying occurrence of a failure in a processor is received from said monitor processor disposed in said current processor group and conducting a change-over operation of transferring the job processing from said current processor group to said standby processor group when the counted number of faulty processors is equal to or more than the number of faulty processors.

5. A parallel processor system according to claim 1, wherein the current processor group and the standby processor group each have the same number of processors.

6. A change-over control method for use with a parallel processor system including a current processor group having a plurality of processors and a network connecting the processors to each other, and a standby processor group having a plurality of processors and a network connecting the processors to each other, and said current processor group and said standby processor group both have a monitor processor disposed therein, and said parallel processor system further includes a processor control section, said method comprising the steps of:

determining a faulty processor ratio according to an amount of job processing in said parallel processor system;

setting the faulty processor ratio in said processor control section;

calculating, in said processor control section, a ratio of failed processors in said current processor group at a pertinent point of time when a report notifying an occurrence of a failure in a processor is received from the monitor processor disposed in said current processor group;

comparing the ratio of failed processors calculated by said processor control section with said faulty processor ratio; and

executing a change-over operation of transferring job processing from said current processor group to said standby processor group when the calculated ratio of failed processors is equal to or more than said faulty processor ratio.

7. A change-over control method according to claim 6, including the further steps of:

reporting to said processor control section, at a preset point of time by the monitor processor disposed in said current processor group, processing information regarding each of the processors belonging to said current processor group;

storing the reported processing information in said processor control section;

notifying, by said processor control section, termination of communication to each of the processors belonging to said current processor group when the job processing is transferred from said current processor group to said standby processor group;

reporting logical addresses corresponding to respective physical addresses of each of the processors belonging to said standby processor group and initiating communication to each of the processors belonging to said standby processor group;

reading, by the monitor processor disposed in said standby processor group, the processing information from said processor control section and transmitting the processing information to each of the processors belonging to said standby processor group; and

receiving, at each of the processors belonging to said standby processor group, the processing information from each of the processors belonging to said current processor group.

8. A change-over control method according to claim 6, further including the steps of:

determining a number of failed processors, according to an amount of job processing;

setting the determined number of failed processors in said processor control section;

calculating, by said processor control section, the number of failed processors each time a report notifying occurrence of a failure in a processor is received from the monitor processor disposed in the current processor group;

comparing the calculated number of failed processors, in place of said faulty processor ratio, with the determined number of faulty processors; and

executing, when the calculated number of failed processors matches the number of faulty processors, a change-over operation of transferring job processing from said current processor group to said standby processor group.

9. A change-over method according to claim 6, wherein said current and standby processor groups both have the same number of processors therein.

14. A change-over method according to claim 7, further including the steps of:

determining, by said processor control section, whether or not the calculated number of failed processors before the pertinent point of time is different from a posterior number of failed processors to be utilized thereafter whenever a time indicated by said timer is changed;

comparing the calculated number of failed processors with the posterior number of failed processors;

comparing, when the posterior number of failed processors is less than the calculated number of failed processors, the number of failed processors at a present point in time in said current processor group with the posterior number of failed processors; and

executing a change-over operation of transferring job processing from said current processor group to said standby processor group when the number of failed processors up to the present point in time in said current processor group is equal to or more than the posterior number of failed processors.

10. A change-over control method for use with a parallel processor system including a current processor group having a plurality of processors and a network connecting the processors to each other, and a standby processor group having a plurality of processors and a network connecting the processors to each other, said current processor group and said standby processor group both include a monitor processor therein, and said parallel processor system includes a processor control section having a timer therein, said method comprising the steps of:

determining a faulty processor ratio at a predetermined interval of time according to an amount of job processing in said parallel processor system;

setting the determined faulty processor ratio in said processor control section;

calculating, by said processor control section, a ratio of failed processors in said current processor group at a pertinent point of time when a report notifying an occurrence of a failure in a processor is received from said monitor processor disposed in said current processor group;

comparing said calculated ratio of failed processors with said determined faulty processor ratio corresponding to a point of time indicated by said timer; and

executing a change-over operation of transferring job processing from said current processor group to said standby processor group when said calculated ratio of failed processors is equal to or more than said determined faulty processor ratio.

11. A change-over method according to claim 10, including the further steps of:

determining, in said processor control section, whether or not said determined faulty processor ratio before the pertinent point of time is different from a posterior faulty processor ratio to be employed thereafter whenever a time indicated by said timer is changed;

comparing said determined faulty processor ratio with the posterior faulty processor ratio;

calculating a second ratio of failed processors up to a particular point of time in said current processor group when the posterior faulty processor ratio is less than said determined faulty processor ratio;

comparing the second calculated ratio of failed processors with the posterior faulty processor ratio; and

executing a change-over operation of transferring job processing from said current processor group to said standby processor group when the second calculated ratio of failed processors is equal to or more than the posterior faulty processor ratio.

12. A change-over control method according to claim 10, further including the steps of:

reporting to said processor control section, at a preset point of time by the monitor processor disposed in the current processor group, processing information regarding each of the processors belonging to said current processor group;

storing the reported processing information in said processor control section;

notifying, by said processor control section, termination of communication to each of the processors belonging to said current processor group when the job processing is transferred from said current processor group to said standby processor group;

reporting logical addresses corresponding to respective physical addresses of each of the processors belonging to said standby processor group and initiating communication to each of the processors belonging to said standby processor group;

reading, by the monitor processor disposed in said standby processor group, the processing information from said processor control section and transmitting the processing information to each of the processors belonging to said standby processor group; and

receiving and executing, at each of the processors belonging to said standby processor group, the job processing from each of the processors belonging to said current processor group.

13. A change-over control method according to claim 10, further including the steps of:

determining a number of failed processors, in place of said faulty processor ratio, according to an amount of job processing;

setting the determined number of failed processors in said processor control section;

calculating, by said processor control section, the number of failed processors each time a report notifying occurrence of a failure in a processor is received from the monitor processor disposed in the current processor group;

comparing the calculated number of failed processors with the determined number of faulty processors corresponding to the time indicated by said timer; and

executing, when the calculated number of failed processors matches the number of faulty processors, a change-over operation of transferring job processing from said current processor group to said standby processor group.

15. a change-over method according to claim 10, wherein said current and standby processor groups both have the same number of processors therein.