The application discloses a private cloud-based
server resource scheduling method and
system, and relates to the technical field of
server resource scheduling. The method comprises the following steps: performing multi-dimensional resource modeling on physical
server nodes in a private cloud cluster,
cutting a physical GPU into a plurality of virtual GPU instances through
virtualization technology, and constructing a unified
resource pool; receiving a computing task submitted by a user, analyzing the QoS level identifier and multi-dimensional resource requirement of the task; based on the real-time state of the unified
resource pool and the multi-dimensional resource requirement, adopting
parallel optimization to solve a globally optimal task placement scheme for each candidate node, and generating a scheduling decision; according to the scheduling decision, allocating and binding physical CPU cores and virtual GPU instances for the computing task on a target node, and configuring corresponding
resource isolation parameters according to the QoS level of the task. The technical problem of low overall utilization of server resources in the prior art is solved, and the technical effect of improving the overall
resource utilization of the server is achieved.