Information processing apparatus, information processing system, information processing method, and storage medium

A system, comprising:

• a plurality of compute nodes;
• and
• a management apparatus coupled to the compute nodes by a network and configured to manage the plurality of information processing apparatuses,
• the management apparatus is configured to
• select a third compute node which is used when degradation occurs on a communication path, the communication path being used when a first compute node transmits data to a second compute node, and
• the first compute node includes a first processor configured to:
• obtain a maximum amount of memory for a job to be started, as a desired amount of buffer,
• determine, for each of the plurality of compute nodes, whether an amount of unused memory in a compute node is greater than the desired amount of buffer,
• determine the compute node as a relay node when it is determined that the amount of unused memory is greater than the desired amount of buffer,
• start the job,
• determine whether a size of data is equal to a predetermined threshold or less when degradation occurs on a communication path to a destination compute node of the data,
• transmit the data to the determined relay node coupled to the destination compute node when it is determined that the size of the data is more than the predetermined threshold, and
• transmit the data to the destination compute node through the communication path when it is determined that the size of the data is the predetermined threshold or less, and
• the second compute node includes a second processor configured to:
• acquire the data from the third compute node when the degradation occurs on the communication path and when the size of the data is more than the predetermined threshold, and
• receive the data from the first compute node when the degradation occurs on the communication path and when the size of the data is not more than the predetermined threshold.

An information processing method executed by a processor of an information processing apparatus, the information processing method comprising:

• obtaining a maximum amount of memory for a job to be started, as a desired amount of buffer;
• determining, for each of a plurality of compute nodes, whether an amount of unused memory in a compute node is greater than the desired amount of buffer;
• determining the compute node as a relay node when it is determined that the amount of unused memory is greater than the desired amount of buffer;
• starting the job;
• determining whether a size of data is equal to a predetermined threshold or less when degradation occurs on a communication path to a destination compute node of the data;
• transmitting the data to the determined relay node coupled to the destination compute node when it is determined that the size of the data is more than the predetermined threshold;
• and
• transmitting the data to the destination compute node through the communication path when it is determined that the size of the data is the predetermined threshold or less.

A non-transitory computer-readable storage medium storing a program that causes a processor included in an information processing apparatus to execute a process, the process comprising:

• obtaining a maximum amount of memory for a job to be started, as a desired amount of buffer;
• determining, for each of a plurality of compute nodes, whether an amount of unused memory in a compute node is greater than the desired amount of buffer;
• determining the compute node as a relay node when it is determined that the amount of unused memory is greater than the desired amount of buffer;
• starting the job;
• determining whether a size of data is equal to a predetermined threshold or less when degradation occurs on a communication path to a destination compute node of the data;
• transmitting the data to the determined relay node coupled to the destination compute node when it is determined that the size of the data is more than the predetermined threshold;
• and
• transmitting the data to the destination compute node through the communication path when it is determined that the size of the data is the predetermined threshold or less.