Information processing apparatus, computer-readable recording medium having stored therein program, and method for processing information
In extendedly allocating a job to nodes subordinate to multiple leaf switch, the job is allocated to nodes of each of multiple leaf switches such that that the number of the nodes being allocated thereto the job and belonging to each of the multiple leas switches does not exceed the number of valid links between the leaf switch and the spine switches. This can avoid occurrence of a conflict accompanied by a link failure, so that degrading of the system can be avoided.
1. An information processing apparatus included in an information processing system comprising:
a plurality of nodes,
a plurality of first switches each of which is connected to a set of respectively unique nodes among the plurality of nodes, and
a plurality of second switches each of which is connected thereto the plurality of first switches via a plurality of links including valid links and failure links,
the information processing apparatus comprising a processor that allocates a job executed in a predetermined number of nodes to one or more nodes among the plurality of nodes, the processor is configured to execute a process comprising:
determining whether a single target switch of the plurality of first switches is connected to a plurality of unoccupied nodes equal to or more than the predetermined number of nodes, the plurality of unoccupied nodes are among the plurality of nodes;
upon determining that the single target switch is connected to the plurality of unoccupied nodes equal to or more than the predetermined number of nodes, allocating the job to the plurality of unoccupied nodes connected to the single target first switch; and
upon determining that the single target switch is not connected to the plurality of unoccupied nodes equal to or more than the predetermined number of nodes, allocating the job to unoccupied nodes connected to each of two or more target first switches among the plurality of the first switches,
a number of the unoccupied nodes connected to each of the two or more target first switches does not exceed a number of the valid links among the plurality of links connected to each of the two or more target first switches.
2. The information processing apparatus according to claim 1 , wherein the processor is further configured to:
sort the plurality of first switches in order of decreasing number of the failure links among the plurality of links connected to each of the plurality of first switches; and
wherein the determining comprises determining whether the plurality of unoccupied nodes are connected to each of the plurality of the first switches in order obtained by the sorting.
3. The information processing apparatus according to claim 2 , wherein the processor is further configured to:
sort the plurality of first switches in order of increasing number of the failure links among the plurality of links connected to each of the plurality of first switches; and
wherein the allocating the job to the unoccupied nodes connected to each of the two or more target first switches comprises allocating the job to the number of the unoccupied nodes connected to each of the two or more target first switches sequentially in order obtained through the sorting.
4. The information processing apparatus according to claim 1 , wherein the processor is further configured to:
sort the plurality of first switches in order of increasing number of the failure links among the plurality of links connected to each of the plurality of first switches; and
wherein the allocating the job to the unoccupied nodes connected to each of the two or more target first switches comprises allocating the job to the number of the unoccupied nodes connected to each of the two or more target first switches sequentially in order obtained through the sorting.
5. A non-transitory computer-readable recording medium having stored therein a program that causes a computer to execute a process comprising:
in an information processing system comprising a plurality of nodes, a plurality of first switches each of which is connected to a set of respectively unique nodes among the plurality of nodes, and a plurality of second switches each of which is connected thereto the plurality of first switches via a plurality of links, in allocating a job executed in a predetermined number of nodes to one or more nodes among the plurality of nodes,
determining whether a single target switch of the plurality of first switches is connected to a plurality of unoccupied nodes equal to or more than the predetermined number of nodes, wherein the plurality of unoccupied nodes are among the plurality of nodes;
upon determining that the single target switch is connected to the plurality of unoccupied nodes equal to or more than the predetermined number of nodes, allocating the job to the plurality of unoccupied nodes connected to the single target first switch; and
upon determining that the single target switch is not connected to the plurality of unoccupied nodes equal to or more than the predetermined number of nodes, allocating the job to unoccupied nodes connected to each of two or more target first switches among the plurality of the first switches,
wherein a number of the unoccupied nodes connected to each of the two or more target first switches does not exceed a number of the valid links among the plurality of links connected to each of the two or more target first switches.
6. The non-transitory computer-readable recording according to claim 5 , the process further comprising:
sorting the plurality of first switches in order of decreasing number of the failure links among the plurality of links connected to each of the plurality of first switches, wherein
wherein the determining comprises determining whether the plurality of unoccupied nodes are connected to each of the plurality of the first switches in order obtained by the sorting.
7. The non-transitory computer-readable recording according to claim 6 , the process further comprising:
sorting the plurality of first switches in order of increasing number of the failure links among the plurality of links connected to each of the plurality of first switches; and
wherein the allocating the job to the unoccupied nodes connected to each of the two or more target first switches comprises allocating the job to the number of the unoccupied nodes connected to each of the two or more target first switches sequentially in order obtained through the sorting.
8. The non-transitory computer-readable recording according to claim 5 , the process further comprising:
sorting the plurality of first switches in order of increasing number of the failure links among the plurality of links connected to each of the plurality of first switches; and
wherein allocating the job to the unoccupied nodes connected to each of the two or more target first switches comprises allocating the job to the number of the unoccupied nodes connected to each of the two or more target first switches sequentially in order obtained through the sorting.
9. A method of processing information, in an information processing system comprising a plurality of nodes, a plurality of first switches each of which is connected to a set of respectively unique nodes among the plurality of nodes, and a plurality of second switches each of which is connected thereto the plurality of first switches via a plurality of links, in allocating a job executed in a predetermined number of nodes to one or more nodes among the plurality of nodes, the method comprising:
determining whether a single target switch of the plurality of first switches is connected to a plurality of unoccupied nodes equal to or more than the predetermined number of nodes, wherein the plurality of unoccupied nodes are among the plurality of nodes;
upon determining that the single target switch is connected to the plurality of unoccupied nodes equal to or more than the predetermined number of nodes, allocating the job to the plurality of unoccupied nodes connected to the single target first switch; and
upon determining that the single target switch is not connected to the plurality of unoccupied nodes equal to or more than the predetermined number of nodes, allocating the job to unoccupied nodes connected to each of two or more target first switches among the plurality of the first switches,
wherein a number of the unoccupied nodes connected to each of the two or more target first switches does not exceed a number of the valid links among the plurality of links connected to each of the two or more target first switches.
10. The method according to claim 9 , further comprising:
sorting the plurality of first switches in order of decreasing number of the failure links among the plurality of links connected to each of the plurality of first switches, wherein
the determining comprises determining whether the plurality of unoccupied nodes are connected to each of the plurality of the first switches in order obtained by the sorting.
11. The method according to claim 10 , further comprising:
sorting the plurality of first switches in order of increasing number of the failure links among the plurality of links connected to each of the plurality of first switches; and
wherein the allocating the job to the unoccupied nodes connected to each of the two or more target first switches comprises allocating the job to the number of the unoccupied nodes connected to each of the two or more target first switches sequentially in order obtained through the sorting.
12. The method according to claim 9 , further comprising:
sorting the plurality of first switches in order of increasing number of the failure links among the plurality of links connected to each of the plurality of first switches; and
wherein allocating the job to the unoccupied nodes connected to each of the two or more target first switches comprises allocating the job to
the number of the unoccupied nodes connected to each of the two or more target first switches sequentially in order obtained through the sorting.