Ilmu Komputer & AI editorial
METIS: A Declarative Slice Orchestrator for Application-Centric 5G/6G Networks
The core problem
Network slicing is positioned as the cornerstone of application-aware 5G and 6G networks, yet the dynamic lifecycle management of network slice instances with coordinated quality-of-service (QoS) enforcement across the radio access network (RAN) and core network (CN) remains unresolved. The authors identify three limitations in existing orchestrators: they rely on network-centric data models, imperative workflows, and static slice templates. In parallel, O-RAN addresses radio-side slice control independently of 3GPP core-side control, leaving slice-level QoS enforcement uncoordinated across domains.
METIS is introduced as a declarative slice orchestrator that manages the Day-0/1/2 lifecycle of network slice instances through cascaded reconciliation loops. It defines an application-centric data model for service profiles, enabling customers to describe the semantics and quality-of-experience (QoE) requirements of their applications. From these, METIS derives 3GPP-aligned slice profiles via hierarchical aggregation following the 5G QoS model, eliminating static templates, and jointly coordinates O-RAN and 3GPP slicing for slice instantiation and enforcement.
The paper's central fin
Innovation
METIS is evaluated on a 5G cloud-native testbed in a campus-event scenario. The reported lifecycle operation times are:
- Slice creation: 22.4 seconds
- Slice update: 5.1 seconds
- Slice upgrade: 52.2 seconds
- Slice deletion: 32.1 seconds
The orchestrator sustains full SLA satisfaction under concurrent multi-slice overload. It scales to 63 slice instances across nine zones while consuming under 0.03 processor cores total. It also recovers slices from injected failures across four levels in under 19 seconds.
These results indicate that declarative, application-centric orchestration with cascaded reconciliation can meet lifecycle and scaling requirements in a realistic multi-slice deployment. The scaling figure is particularly notable: 63 slice instances across nine zones with total consumption below 0.03 processor cores suggests that the control-plane overhead of the orchestrator is small relative to the slice data plane.
The failure-recovery result - under 19 seconds across four injected failure levels - supports the claim that cascaded reconciliation loops can restore slice state without manual intervention. The SLA satisfaction result under concurrent multi-slice overload ad
Why it matters
The central analytical contribution of the paper is the identification of a structural asymmetry in end-to-end slice control. Downlink traffic can be shaped at the core before it reaches the RAN, so core-side enforcement can influence downlink service levels. Uplink traffic, however, leaves the UE unregulated: the UE transmits before any core-side policy can be applied. Consequently, core-only slicing cannot reliably satisfy uplink SLAs, and radio-side enforcement is necessary rather than merely complementary.
This asymmetry reframes the relationship between O-RAN and 3GPP slicing. If radio-side and core-side control are addressed independently, as is common today, uplink enforcement falls outside the scope of core-only orchestration. METIS addresses this by jointly coordinating O-RAN and 3GPP slicing for both instantiation and enforcement, so that the derived 3GPP-aligned slice profiles are enforced at the appropriate domain.
The application-centric data model is the second key design choice. By letting customers describe application semantics and QoE requirements, and by deriving slice profiles through hierarchical aggregation following the 5G QoS model, METIS eliminates static slice templates. This shifts slice definition from a network-centric, template-driven process to a declarative, intent-driven one.
The cascaded reconciliation loops provide the mechanism for Day-0/1/2 lifecycle management. The measured lifecycle times - 22.4 s creation, 5.1 s update, 52.2 s upgrade, 32.1 s deletion - and the scaling result of 63 slice instances across nine zones under 0.03 processor cores indicate that this declarative approach is operationally feasible. The recovery result of under 19 seconds across four failure levels further supports the resilience of the reconciliation model.
Taken together, the findings suggest that coordinated, application-centric, declarative orchestration is a viable path toward application-aware 5G/6G slicing, provided that uplink enforcement is treated as a first-class requirement rather than an afterthought.
Who should read this
Opening member content…