A data flow migration method in data center network update

Abstract

The invention discloses a method for migrating data streams in data center network update, which includes searching for all feasible shortest paths in the network as candidate paths for each new data stream, or searching out all feasible shortest paths in the network for each new data stream When a data flow is migrated, it is a candidate path that can accommodate the data flow. Since there is no need to consider how to find a feasible network state, there is no need to carefully design and execute an update sequence. By locally scheduling each new data flow, they are assigned the shortest possible path. And only at the expense of the migration of an extra data stream. At the same time, select a data flow that passes through all bottleneck links to migrate, improve update efficiency, ensure that the data flow of this additional data flow is as small as possible, and reduce the data flow of migration. There is no need to search the destination data flow matrix across the network, nor to search for a series of transition states between the initial and destination states of the network.

Description

technical field [0001] The invention relates to the network updating of the data center, in particular to a data flow migration method in the data center network updating. Background technique [0002] Infrastructure as a Service (IaaS) is a cloud computing model that allows tenants to multiplex computing, storage and networking resources in a data center. With the rapid growth of IaaS, service providers have to rent or build a large-scale data center. Inside each data center, a special data center network (DCN) connects thousands of switches and hundreds of servers. Network update events frequently occur in cloud computing data centers, such as network topology updates, data flow updates caused by network operators, network updates caused by application programs and even switch failures. For example, data center operators periodically upgrade switches in the network and add more switches to support more servers; thus, topology changes occur. For applications, normal VM m...

AI Technical Summary

Problems solved by technology

These data streams that need to be migrated will incur additional overhead and may also affect the running application

In other words, these data flows that need to be migrated will cause some additional data flows to be migrated, resulting in unnecessary rules that need to be installed on the switch

At the same time, link congestion and packet loss will also occur

Images