A Method for Optimizing Data Center Switching Network Costs

Abstract

The invention discloses a method for optimizing the switching network cost of a data center, which includes the following steps: an input stage switch accumulates received data packets to form a flow matrix B; and the ADTPT algorithm is used to establish a data packet delay and a core switch Compromise relationship between quantities; calculate the minimum number of core switches for a given delay; calculate the minimum cost of the data center switching network. The present invention has the following advantages: a compromise design between the delay and the number of core switches, the number of core switches can be reduced by appropriately increasing the delay, the design is flexible, and the scalability of the data center switching network design is enhanced; the core switch The reconfiguration overhead of the switching network is included in the total cost of the switching network, which improves the cost design of the switching network; the traffic matrix decomposition algorithm is extended from the scheduling of a single switch to the scheduling of multiple switches, and realizes the scheduling under the delay constraint. The batch scheduling in the data center switching network improves the batch scheduling process of multi-level switching.

Description

technical field [0001] The invention relates to a method for optimizing the cost of a switching network in a data center. Background technique [0002] With the rapid development of cloud services, the data center has received extensive attention as a business support platform. In order to solve the problems of high energy consumption, complex internal connections, high cost, and difficult to guarantee switching performance faced by the design of the internal switching network of the data center, the new architecture of the data center using optical switching and optical connections has become the main direction of technological development. These solutions are characterized by using OpenFlow switches as ToR (or Aggregation) switches, and using crossbar optical switches to implement core switches, and using high-speed optical fiber connections in the data center switching network, thus simplifying the structure of the switching network and enhancing system scalability. [...

AI Technical Summary

Problems solved by technology

[0005] However, the ADAPT algorithm is only applied to the scheduling of a single switch at present; in addition, an important problem faced by the above-mentioned crossbar optical switch as the core switch is the reconfiguration overhead of the system (mainly including the cost of switching the cross-connect configuration of the core switch) required time overhead, and the time it takes to synchronize between switches), with the continuous expansion of the scale of the data center and the increase in the complexity of the switching network, the problem of reconfiguration overhead will become more and more prominent. However, currently for data The design research on the cost optimization of the switching network inside the center has always ignored the reconfiguration overhead of the system, and the design is not reasonable

Images