Research on elastic control problem of multi-tenant data storage System in cloud

Research on elastic control problem of multi-tenant data storage System in cloud

Shandong University Huliansu

This article focuses on metrics related to data access time. The above two conditions should be ensured in the data elastic control, which increases the difficulty of the elastic control. 2 in order to achieve the flexible control of a multi-tenant storage system, we need to be able to propose control strategies to determine which data to move and where to move. Large-scale application load fluctuation, adaptive load balancing requirements can be scale-out at peak load, load low peak scale-down. This is achieved through data movement (e.g., partitioned or coalesced), but frequent movement can lead to new performance problems for systems that have already been overloaded. 3 to have a load forecasting mechanism to make the data extension strategy can be developed in advance. The supply of resources in the cloud has to go through a start-up delay, which requires that the control strategy should be taken as early as possible. Performance metrics in monitoring and Control service level agreements (SLAs) are susceptible to ambient noise and are not easily measured, and the performance requirements of different tenants in different SLAs are more complex than the same performance requirements. To solve the above problems, we propose a Adaptscala system based on MPC control, which can monitor the load of different tenant data access and load forecasting based on an exponential smoothing method. Combined with the performance requirements of data access response time in different tenant SLAs, The performance model of constructing multi-tenant data storage is used to determine whether each server can meet the performance requirements of different tenants for data access; Finally, a data elastic control Strategy generation algorithm is presented, which is used to calculate the final data storage adjustment strategy, which has less impact on the overall performance of the system and minimizes the overall resource cost. This paper uses Berkeley DB as the Key/value type of data storage engine in cloud, uses open source software to build the Adaptscala system, simulates multi-tenant data access and collects performance data to build a multi-tenant data storage model, Several tests have been carried out to verify that the system can meet the proposed elastic control target.

Research on elastic control problem of multi-tenant data storage System in cloud