Method for parallelly processing data in extensible manner for network I/O (input/output) virtualization

Abstract

The invention discloses a method for parallelly processing data in an extensible manner for network I / O (input / output) virtualization. The method has the advantages that data receiving and transmitting, protocol analyzing, data stream managing and upper-level application are regarded as network I / O links on the basis of hierarchical structures of networks, and parallel cooperative work is carried out in the various links in an asymmetric assembly line manner; sufficient computational resources are injected into computational resource bottleneck points in a symmetric parallel mode, and accordingly the data processing capacity is parallelized and is extensible; multi-core resources are flexibly managed according to system load, and accordingly performance bottleneck of I / O virtual resources due to the computational capacity of the traditional I / O driving methods can be eliminated.

Description

Technical field [0001] The invention relates to the field of network I / O virtualization data processing, and in particular to a parallel and scalable data processing method for network I / O virtualization. Background technique [0002] I / O virtualization is a software technology, which is an abstraction of network equipment between the underlying hardware and the workload. At present, there are three solutions for I / O virtualization: full virtualization, paravirtualization and hardware-assisted virtualization. Full virtualization is a technology that simulates hardware network I / O completely through software programs. This method is inefficient and cannot make full use of the physical line speed of 10GbE networks, and it is no longer capable of meeting the virtualization requirements of current high-performance network NIC devices. Currently, high-performance network interface devices (NICs) above 10GbE provide hardware-assisted single-root I / O virtualization (SR-IOV) support and...

AI Technical Summary

Problems solved by technology

When the single-threaded data processing method overhead exceeds the computing power of the physical CPU, the throughput performance of virtualized I / O resources is limited by the computing resource bottleneck of data processing

On the other hand, SR-IOV is assisted by hardware to complete Layer 2 data exchange, and the bottleneck of computing resources has been greatly improved. However, similar computing resource bottlenecks appear in the VF driver in the virtual machine.

The instantaneous network burst of a single virtual machine will cause a large amount of packet loss and performance degradation, and the overall I / O throughput performance will also drop significantly in the case of alternate throughput bursts of multiple virtual machines

In short, the data processing method of high-performance network I / O virtualization has high computing resource overhead, which makes it difficult to fully utilize the physical wire-speed performance of high-performance NICs to form virtualized I / O resources.

Images