Principles

Based on the native SPDK+Ceph software, the following optimizations are implemented to improve cloud disk performance.

No.	Optimization Solution	Description
1	Thread-core binding	CPUs are specified when the SPDK is started to implement device I/O request polling. Thread-core binding is performed on the Ceph client so that Ceph threads can efficiently use the CPU resources, improving performance.
2	TCMalloc	The TCMalloc function is enabled to optimize memory allocation and release.
3	Gazelle	The Gazelle function is enabled to optimize the client kernel network.

Parent topic: Feature Description