Zero-copy

Zero-copy technology refers to when a computer performs an operation,The CPU does not need to first copy the data from one specific area of memory to another. This technique is often used to save CPU cycles and memory bandwidth when transferring files over a network. Zero-copy is a technique used in computer operations to avoid the CPU from being involved in the task of copying data between memory areas. It also eliminates unnecessary copies of data, thereby saving CPU cycles and memory bandwidth. This approach is particularly useful for tasks that involve transferring large amounts of data, such as high-speed file transfers over a network.

Hardware Implementation

The earliest implementation was for IBM OS/360, where a program could instruct the channel subsystem to copy blocks of data from one file or device to another without first moving the data.

Techniques for creating zero-copy software include using direct memory access (DMA) and memory mapping through a memory management unit (MMU). These features require specific hardware support and often involve specific memory alignment requirements.

A newer approach is to use Heterogeneous System Architecture (HSA) to facilitate passing pointers between the CPU and GPU and other processors, which requires the CPU and GPU to use a unified address space.