Distributed Asynchronous Service Deployment in the Edge-Cloud Multi-tier Network

TL;DR

The paper tackles latency-constrained service deployment and migration in edge-cloud multi-tier networks with mobile users. It introduces DASDEC, a distributed asynchronous framework that operates without a global state, using Stage SFS, PU, and PD to place and migrate services while minimizing cost and migration overhead. The authors prove PMP is NP-hard, provide convergence guarantees with an F-mode mechanism, and show in trace-driven simulations that DASDEC achieves near-central performance with low control traffic and low overhead. This work enables scalable, cross-provider deployment and migration across edge and cloud datacenters, reducing reliance on a single orchestrator and improving resilience in dynamic environments.

Abstract

In an edge-cloud multi-tier network, datacenters provide services to mobile users, with each service having specific latency constraints and computational requirements. Deploying such a variety of services while matching their requirements with the available computing resources is challenging. In addition, time-critical services may have to be migrated as the users move, to keep fulfilling their latency constraints. Unlike previous work relying on an orchestrator with an always-updated global view of the available resources and the users' locations, this work envisions a distributed solution to the above problems. In particular, we propose a distributed asynchronous framework for service deployment in the edge-cloud that increases the system resilience by avoiding a single point of failure, as in the case of a central orchestrator. Our solution ensures cost-efficient feasible placement of services, while using negligible bandwidth. Our results, obtained through trace-driven, large-scale simulations, show that the proposed solution provides performance very close to those obtained by state-of-the-art centralized solutions, and at the cost of a small communication overhead.

Figures (5)

• Figure 1: Reference scenario: Each mobile user (illustrated by a car or a human) is associated with the nearest PoA, which is equipped with a co-located Mobile Edge Computing (MEC) server. The MEC servers are connected to higher-level servers with more computational capacity. Our model refers to such servers in the edge-cloud continuum as datacenters.
• Figure 2: A scenario with seven datacenters (the cloud is at the root of the topology). The PoAs of requests $r_0, r_1$, $r_2$, and $r_3$ are $s_3$, $s_5$, $s_6$, and $s_6$, respectively. A naive algorithm fails to place $r_3$ (Fig. \ref{['fig:need_bu_problem']}), while there exists a feasible placement (Fig. \ref{['fig:need_bu_sol']}). After $r_3$ finishes (Fig. \ref{['fig:need_bu_r_3_finish']}), an imprudent algorithm may push request $r_1$ up again to datacenter $s_2$ (Fig. \ref{['fig:need_bu_after_pu']}). This may result in falling back into a scenario that requires push-down (Fig. \ref{['fig:need_bu_fail_again']}).
• Figure 3: Run example of the PD procedure. The circle denotes the step $\in\{0,5\}$.
• Figure 4: Minimum required processing capacity when varying the ratio of RT service requests.
• Figure 5: Per-request signaling overhead due to DASDEC.

Theorems & Definitions (2)

• Theorem 1
• proof