Research Of Emergency Backup In Inter-Datacenter Networks

Nowadays,data centers(DCs)are becoming the key infrastructure to support cloud computing,big data and data-intensive applications.To provide users with high quality,low latency and nondisruptive network services,enterprises such as Google,Microsoft and Amazon have placed their own data centers globally and built the inter-data center networks(Inter-DCNs)by interconnecting these DCs.However,an Inter-DCN is vul-nerable to disasters such as earthquakes,hurricane and tsunami,which will affect the services of thousands of users,thus causing huge losses to network operators.Hence most enterprises have their data backup strategies,which include regular and emergency backup.Regular backup regularly receives data redundancy to cope with possible disas-ters in the future,while emergency backup is triggered urgently when a disaster is going to occur.The process of emergency backup is quite different from regular backup,and its research is relatively lacking.This topic studies the emergency backup problem in Inter-DCNs.Specifically,we firstly utilize the time expanded network(TEBN)method and modify it to the vari-ant TEN(VTEN),and transform the dynamic network optimization in the emergency backup scenario into a static one.Secondly,based on the VTEN network,we formu-late an optimization to maximize the emergency backup profit with considerations of data value and backup costs.Then we use alternating direction method of multipliers(ADMM)algorithm to solve the large-scale optimization model in parallel.However,a sub-problem of ADMM is costly to solve exactly,we therefore turn to obtain an ap-proximate solution of it,thus developing a new inexact ADMM algorithm.Finally,we theoretically prove the convergence of the proposed algorithm and evaluate its per-formance through simulation experiments.The results show that our inexact ADMM algorithm has good convergence and robustness.In the meanwhile,it outperforms sev-eral existing algorithms in terms of backup profit and running time.