Optimization of computation and bandwidth resource allocation for deployed services

Determine optimal values for the computation resources f_m^{j,t} allocated by each edge server m to each deployed service j and the scheduling bandwidth R_{m,m'}^{j,t} allocated for tasks of service j from edge server m to edge server m' in each large-timescale frame t, rather than modeling these quantities as fixed functions of the service deployment decisions d_m^{j,t}.

Background

In the service deployment model, the paper introduces decision variables for computation resources f_m{j,t} allocated to each deployed service and inter-edge-server scheduling bandwidth R_{m,m'}{j,t}. These variables directly affect task processing delay but, to focus the paper on joint service deployment and task scheduling, they are not optimized; instead, they are modeled as functions of the deployment decisions.

The authors explicitly state that optimizing these resource allocation strategies is deferred, indicating a specific unresolved component needed to complete the overall optimization framework for minimizing expected processing delay in mobile edge computing networks.

References

The specific optimization of resource allocation strategies will be addressed in future research.

— Two-Timescale Dynamic Service Deployment and Task Scheduling with Spatiotemporal Collaboration in Mobile Edge Networks  (2508.16293 - Li et al., 22 Aug 2025) in Section II-B (Service Deployment Model)

An LLM-backed executor, i.e., agentic placement across heterogeneous resources, is a natural third locus for reasoning, but our single-node, homogeneous testbeds in this paper present no real placement choice, so we leave it to the multi-node setting for future.

— Avatar: Toward Autonomous End-to-End Orchestration of Scientific Workflows using LLMs  (2609.10509 - Raj et al., 9 Sep 2026) in Section 3, subsection “Policies, adapters, and modes” (Implementation)

Developing an alternative NSR-DP formulation that directly enforces physical capacity constraints and deriving corresponding theoretical guarantees remain future work.

— Contextual Bandit-Based Decomposition of Network Slice Requirements under Cumulative Resource Budget Constraints  (2609.09624 - Kobayashi et al., 9 Sep 2026) in Remark 2, Section IV-A, p. 11

Although this study focuses on predictive modeling rather than closed-loop control, ResLearn-XR outputs can support XR-aware management. Traffic predictions provide offered-load and burst-risk indicators for proactive bandwidth, queue, or rate-adaptation decisions, while calibrated QoE-risk probabilities can inform admission control, edge scaling, rendering adaptation, or migration. The bandwidth settings used here define controlled evaluation conditions and are not dynamically adjusted; closed-loop resource orchestration remains future work.

— ResLearn-XR: Residual Learning for Network Traffic and Quality-of-Experience-Aware Modeling in Extended Reality  (2609.04493 - Manjunath et al., 3 Sep 2026) in Section 5, subsection “Traffic Prediction” (discussion following Table VII)