Quantitative Availability Analysis Of Hierarchical Datacenter Under Power Oversubscription

From the perspective of economic and efficient benefits,modern data center oversubscribes power supplies to deploy as many servers as possible.The oversubscription is based on the varied loads among servers to modulate power demand.Nevertheless,power oversubscription has potential threats to system availability,the data center may collapse as a result of overloading.Current solutions to the oversubscription usually focus on managing the datacenter workload to avoid the peak power demand time in the data center.However,none of the current research considers the influence of the failure of the power or utility components,where the component failure may affect the effectiveness of these strategies.Meanwhile,none of these current research can answer the question that how many servers should be deployed in the data centers under an availability constraint.In this paper,we propose quantitative availability analysis of hierarchical datacenter under power oversubscription.To this end,we use Markov chain and Stochastic Reward Net(SRN)to model the failure and repair processes of data center components.Markov chain is used to model the components in the upper level(PDU,switches).The servers at the bottom level are distributed in two pools: main pool and backup pool,where running servers are in main pool and turned-off servers in backup pool.Migration from backup pool to main pool is conducted once any running server fails and the failed one is sent to repair.SRNs are implemented to model these two pools.The evaluation is based on the real-life Google and Wikipedia traces.The result shows the relationship between oversubscription and data center availability,which can guide the data center operators to choose the appropriate oversubscription ratio under the availability constraint.The contributions are as follows:1)We present a three tier hierarchical model for the availability analysis of data center,while considering the influence of components failure.We use Markov chains and SRNs to model the failure and repair processes of data center components.We use two state pools to represent the possible intermediate states of servers.2)We propose two SRN sub-models to simulate server level instead of the complicated Markov model.With different component configurations,network topology and state transition of the data center model,we deduce the relationship between data center availability and oversubscription.3)We use Google and Wiki traces to evaluate data center availability from the variation of oversubscription and find the appropriate oversubscription ratio for the data center operators to deploy servers under different availability constraints.4)We use the classic supercapacitor model to simulate UPS and study the influence of UPS on data center availability.By simulation,we deduce the quantitative relationship between data center availability and oversubscription when UPS exists.5)We extend the method of analyzing a single data center and apply it to analyzing multiple data centers.By analysis and simulation,we obtain the quantitative relationship between availability and oversubscription of multiple data centers.Furthermore,the experiment result shows that compared to assembling all servers into one data center,distributing servers to multiple data centers leads to higher availability.