Scheduling Inter-Datacenter Video Flows for Cost Efficiency

Abstract

As video streaming applications are deployed on the cloud, ISPs charge cloud providers for inter-datacenter transfers under the dominant percentile-based charging models. To minimize payment costs, existing works aim to keep traffic below the charging volume on each link. Scheduling Inter-Datacenter Video Flows for Cost Efficiency These methods, however, can not fully utilize the available bandwidth capacity of each link. We propose an economical and timely video flow scheduling system, called EcoFlow, as a solution.

Scheduling Inter-Datacenter Video Flows for Cost Efficiency Cloud providers (e.g. Amazon) offer various cloud-based pay-as-youuse services (e.g., Amazon Web Services) to cloud customers (e.g., Netflix). The Cloud has proven to be an effective infrastructure for hosting video streaming services with many benefits: First, cloud-based infrastructures provide easy-to-use scalability for hosting additional users and content by allowing cloud customers to purchase additional resources from the cloud provider. Second, distribution is similarly scalable and, in this context, a large number of globally distributed servers and streaming service providers (VSSPs) provide services in a vast array of different areas.

Conclusion

Cloud providers need to transfer video content between different datacenters to provide video streaming services to users across different regions. Under percentile based charging models, these inter-datacenter transfers are charged by ISPs. We take advantage of this particular characteristic of these models and propose EcoFlow to minimize the payment costs for inter-datacenter traffic by cloud providers. EcoFlow is an economical, timely video transfer strategy.