Container Elasticity: Based on Response Time using Docker

Objectives: Cloud computing is an approach to provide the computing resources (machine) to end-users for running their application over the Internet. The computing resources consist of various things (e.g. RAM, Memory, CORE, etc.). These resources are allocated to an application without human intervention for managing the fluctuating workload. To manage the real-time fluctuating workload, cloud providers use VM based or Container-based virtualization to host the client services. Adding/removing resources dynamically as per the demand of application through cloud is known as elasticity. Cloud providers use the auto-scaling mechanism to implement elasticity. A machine that hosts an application can be either overloaded or under-loaded due to the real-time fluctuating workload. The cloud providers use an auto-scaling mechanism to automatically scale up or down the computing resources at the right moment for managing the real-time fluctuating workload. The failure of allocation/de-allocation of resources at the right time leads to SLA violation, service unavailability, customers lost, more power consumption, minimum throughput and maximum response time. Hence, the allocation/de-allocation of resources at right moment becomes critical for the successful completion of tasks in a dynamic environment efficiently. Methods: Resource provisioning for managing dynamic and fluctuating workload has been achieved through an algorithm (PID with dynamic HAProxy) which is based on decision-making approach that depends on the response time of container using mechanism of control theory. Results: The proposed work has improved performance of the system in terms of resource utilization and response time to manage the fluctuating workload. Conclusion: The addition/removal of containers dynamically to manage fluctuating workload can be achieved more efficiently.