Locking scheme for four CPUs is often not scalable to 64 CPUS: For example, old Linux kernel was not scalable for 48 CPUs. The reasons are basically due to the lock related implementation, such as mount table read bottleneck in high contention using spin_lock, object reference counting spin check. More details in OSDI10's paper.
Locking scheme for four CPUs is often not scalable to 64 CPUS: For example, old Linux kernel was not scalable for 48 CPUs. The reasons are basically due to the lock related implementation, such as mount table read bottleneck in high contention using spin_lock, object reference counting spin check. More details in OSDI10's paper.
This comment was marked helpful 0 times.