부산대학교 도서관

This paper describes an experiment deployed as a use case on top of the IRIS testbed that implements two concurrent network slices on the same cloud infrastructure: one emulating a mobile virtual network operator public safety service with high throughput and low latency requirements and the other emulating an over-the-top service provider (delay tolerant - best effort slice). A scaling algorithm with the ability to dynamically allocate computational resources among the two slices is described, implemented, and evaluated in terms of performance gains in the priority slice when the shared computational resources are scarce. The results indicate that under depleted computational resources, this algorithm can significantly improve the TCP and UDP throughput performance of priority slices.