Enhanced garbage collection in a multi-node environment

摘要

Embodiments of the invention enhance a garbage collection process running on a parallel system or distributed computing environment. Using a garbage collector in such an environment allows a more in-depth analysis to be performed than is possible on other systems. This is because the number of compute nodes present in many parallel systems, and the connections between them, allows the overhead of doing advanced analysis to be spread across the nodes and the results of that analysis to be shared among the nodes.

主权项

1. A method of performing a garbage collection cycle on a parallel computing system having a plurality of compute nodes each running an instance of a computing job and a garbage collector, comprising:
identifying, by the instance of the garbage collector running on each of the plurality of compute nodes, an object space in a memory of a respective compute node, wherein the object space stores one or more objects allocated by the instance of the computing job running on the respective compute node; evaluating each object in the object space to determine whether the memory allocated to each respective object is eligible to be collected; upon determining that the memory allocated to a given object is eligible to be collected, removing the given object from the object space and returning memory allocated to the given object to a pool; determining a set of garbage collection statistics associated with at least one evaluated object in the object space; and transmitting the set of garbage collection statistics to a master garbage collector running on a compute node of the parallel computing system, wherein the master garbage collector distributes the transmitted set of garbage collection statistics received from the garbage collector on each respective compute node to the instance of the garbage collector running on the other compute nodes, of the plurality and wherein the master garbage collector running on the compute node is configured to assign one or more object types to the garbage collectors running on the respective compute nodes, wherein the garbage collector running on the respective compute nodes determine the set of garbage collection statistics for the assigned types.