High performance design method for network function virtualization

Abstract

A high performance design architecture for network function virtualization is an efficient network function virtualization framework which eliminates the processing redundancy in network functions byaggregating the operations of the network functions in a service chain on a data stream, so as to realize an efficient network data path. According to the design architecture, a local action matchingtable is established in each network function to record the operation behaviors on the data packets in the data stream and the states of the network functions, and a global action matching table is established to aggregate all the local action matching tables. Meanwhile, an event triggering table and a new data packet processing mode are used to guarantee the logical correctness of the network functions at run time, and the performance of the system is improved through the parallelization of network function functions. The processing delay and throughput of the network function service chain are significantly optimized under the condition of guaranteeing the logical correctness of the network functions.

Description

technical field [0001] The invention belongs to the field of network technology, relates to performance optimization of network function virtualization (NFV), in particular to a high-performance design method oriented to network function virtualization. Background technique [0002] Network Function Virtualization (Network Function Virtualization, NFV) refers to the realization of network functions (NF, such as intrusion detection, monitoring and load balancing, etc.) in software to achieve more flexible management and lower network system costs. Network functions are usually connected together to form a serial service chain, and NFV operators guide data packets to be delivered through the service chain. However, the software-based implementation of network functions also brings about major performance problems. Existing studies have shown that packet processing performance degrades severely when packets pass through long chains containing multiple network functions. [00...

AI Technical Summary

Problems solved by technology

Even if the above two types of optimization methods are used, performance degradation is inevitable

If we do not break the boundaries of NF processing, such as merging multiple modifications that act on packets, there may be some processing redundancy that cannot be eliminated by the existing two types of optimizations

Speeding up these operations does not completely eliminate redundancy; and parallelizing them is not necessarily feasible, since they may not be independent (for example, if they write to the same field of the packet)

This redundant operation may increase with the length of the chain, so when the service chain is long enough, the processing performance will drop significantly

Images