Rapid Container Scheduling for Reactive Relocation of Individual HTTP Requests

As smartphones and PCs become more mobile and social networking services (SNSs) become more widespread, increasing opportunities are available to concentrate access to individual web sites depending on their contents. Many developers and researchers have assessed platforms that can withstand sudden access concentration without forcing service users and operators, even individuals, to miss such opportunities. A method of allocating resources to containers reactively and rapidly exists according to the number of accesses and loads without requiring specialized technical knowledge of users. Activating a plurality of containers across a plurality of accommodation servers is necessary to ensure availability to such an extent that no HTTP timeout occurs even if a host server that accommodates containers fails. Such assured capability still entails increased service usage costs and requires specialized knowledge of individual users. As described herein, we propose a rapid container scheduling method that relocates a single container automatically and reactively to another host server with each HTTP request according to the host server status. The proposed method creates an image of the web server process immediately using a general-purpose imaging method for web server software of various types after finishing the server process initialization to relocate a single container rapidly. When relocating a container, the web server process on the container starts rapidly from the image. Automatic scheduling requires no user to have specialized technical knowledge to increase availability. Moreover, obviation of the advance starting up of plural containers on plural accommodating servers reduces the resource costs for ensuring availability. Experimentation results confirmed that even a container that starts an application server which takes a long time to start can guarantee a level of availability that does not cause HTTP request timeouts when the server which accommodates containers stops.