Storage Migration Scheduling Algorithm Based On Workload-Aware

Many cloud adopters have had great successes in leveraging these capabilities to deliver services much faster than any of these users could have achieved if they had to build out their own infrastructure,with the development of the emerging open cloud computing.The huge amount of datum and access request requires the cloud services have the availability of rapid response capability,and high scalability.To maintain high performance and availability,migrations could be used to move virtual machines from one cloud to another cloud that has better resource availability.While this freedom leads too many potential benefits,the running services must be minimally disrupted by the migration.Unfortunately,current solutions for wide-area migration incur too much disruption as they will significantly slow down storage I/O operations during migration.The resulting increase in service latency could be very costly to a business.This thesis presents a novel storage migration scheduling algorithm that can greatly improve storage I/O performance during wide-area migration.In this thesis,the notion of storage migration scheduling is introduced to orchestrate the sequence in which storage is transferred.Scheduling allows us to take advantage of inherent access patterns such as temporal locality,spatial locality,and access popularity that are found in a wide range of I/O workloads to significantly optimize the data transfer and reduce performance degradation.The paper develops a novel workload-aware storage migration scheduling algorithm.The algorithm uses only simple records of a limited number of past I/O operations for workload characteristic inference.Using a trace-driven framework,the algorithm provides large performance benefits across a wide range of popular virtual machine workloads.In the pre-copy model,the algorithm can reduce extra traffic and postponed time;in the post-copy model,the algorithm can reduce the remote reads;and in the pre+post copy model,the algorithm take advantages of previous model,t he performance is exceeded to another three algorithms.Therefore,the algorithm is presented by the paper provides a large performance benefit in wide-area migration,using a trace-driven framework.