User-access-frequency statistics based hotspot adjustment in all-optically interconnected metro-embedded datacenters

• We first introduce the metro-embedded DC architecture and the principle of “follow the user” low latency service provisioning. Then the corresponding control plane design for the floatable resource orchestration is discussed.
• An experimental prototype of the proposed metro-embedded DC architecture is presented.
• Finally, we measured user lower access latencies to validate the feasibility of the proposed resource scheduling schemes.

Traditional datacenters are typically deployed with highly clustered resources and built-in remote areas far away from users, requiring huge floor space and power consumption. The proposed metro-embedded datacenter architecture with distributed resources spreading across the metro area close to users could be a promising solution, which exhibits lower access latency and more flexibility for service orchestration. Based on metro-embedded datacenter, we propose the conception of “follow the user” with user access location aware dynamic hotspots adjustments. By enabling service hotspots adjustment according to user access frequency statistics, the quality of service can be significantly improved in terms of reduced latency. Finally, we experimentally demonstrate low-latency service provisioning based on the proposed principle.