Virtualization scene flow migration method based on network function and system thereof

Abstract

The invention provides a virtualization scene flow migration method based on a network function. The method comprises the following steps of deploying a source flow migration processing unit and a target flow processing unit at a source network function place and a target network function place respectively, and arranging a two-byte label in each packet header of each data packet through the source flow migration processing unit and network exchange equipment so as to tab the data packet into four types, wherein the four types contain a normal type, an in-fly type, a last in-fly type and a redirect type; the target flow migration processing unit uses a classifier based on the label to classify the normal type of the data packets, the in-fly type of the data packets, the last in-fly type of the data packets and the redirect type of the data packets; deploying state machines in the flow migration processing unit and recording a current migration state, wherein a source state machine tabs and forwards the data packets according to the migration state; and a target state machine places the four types of the data packets in default buffer memory, in-fly annular buffer memory and redirect annular buffer memory in the target flow processing unit according to a migration state and a packet processing rule and finally delivers to the target network function place through the default buffer memory.

Description

technical field [0001] The present invention relates to the field of network function virtualization and the field of software-defined network technology, and in particular to a flow migration method and system based on network function virtualization scenarios. Background technique [0002] With the evolution of the network architecture and the continuous development of services, more and more dedicated devices are deployed in the network, such as security devices such as firewalls, intrusion detection, and access authentication, and performance-enhancing devices such as load balancing and TCP acceleration. Same level as routing and switching equipment. The massive deployment of these dedicated devices not only increases the cost of purchase and operation, but also increases the difficulty of management, operation and maintenance. Moreover, the development cycle is long, which cannot meet the requirements of rapid innovation and dynamic deployment of networks and applicatio...

AI Technical Summary

Problems solved by technology

[0006] Split / Merge(Rajagopalan S,Williams D,Jamjoom H,et al.Split / merge:System support for elastic execution in virtual middleboxes[C] / / Presented as part of the 10th USENIX Symposium onNetworked Systems Design and Implementation(NSDI 13). 2013:227-240.): When the migration starts, the controller instructs the Switch to immediately forward the data packets of the flow to be migrated to the controller for centralized caching, and then the controller copies the relevant state of the flow to be migrated in the source network function to the target In the network function, during this process, all the data packets of the migration flow will be cached in the controller. When the state copy is completed, the controller releases all the cached data packets to the target network function, and instructs the switch to directly forward the data packets to the target Network function, the whole process realizes the migration of flow and related states, but there are two problems. First, the packets released from the controller and the packets directly forwarded by the switch will be processed by the target network function out of order, which will cause network The error handling of the function, secondly, when starting to migrate the related flow state in the source network function, there are still some data packets that have been forwarded by the Switch but not yet processed by the source network function, these packets may update the flow state, but these states Updates to the target network function will never be reflected in the target network function, resulting in inconsistent flow state between the source and target network function

[0007] OpenNF: When the migration starts, the flow state is directly copied from the source network function. At this time, the source network function will forward the received migration flow data packet to the controller for caching. When the state copy is completed, the controller releases all cached packets. , the Switch also directly forwards the packet to the target network function, but at this time the target network function will first cache all the data packets from the Switch, and wait for all the data packets released from the controller to be processed before processing. Through this delicate The design can solve the problem of split / merge state inconsistency and packet disorder. However, this is a centralized system. When the migration scale is large, the controller will become the bottleneck of the entire system (single point processing and caching all Migration packets), in addition, the way migration packets are delivered to the target network function via the controller introduces additional forwarding

[0008] Improved OpenNF is different from OpenNF. The source network function only forwards the packets that upgrade the flow state to the controller for caching, which greatly reduces the cache pressure of the controller while ensuring state consistency. However, this method is still a centralized In addition, this method cannot guarantee that data packets will reach the target network function without disorder

[0009] like figure 1 As shown, the existing technology has shortcomings in terms of state consistency, migration efficiency, centralized caching requirements, and additional overhead of network functions.

Images