Open Source MANO and Sustainability
Open Source MANO (OSM) is an ETSI-hosted, production-grade NFV “Management and Orchestration” stack [1]. It automates lifecycle management of network services (NS) and virtualised functions. Traditionally, OSM optimises performance, availability, and cost. The carbon-aware extension described here integrates environmental metrics into orchestration: real-time carbon intensity or renewable availability (e.g. solar fraction) guide resource decisions [2].
Two representative mechanisms within OSM’s control loop are summarised as follows:
Service Throttling:
Define multiple Virtual Energy-Aware States (VEAS) for a workload, each with different CPU frequency, core count, and power states. OSM’s Eco-Control module dynamically switches between these states based on carbon signals. For example, during high grid carbon intensity (nighttime or low renewables), throttle the NS to a low-power state; when green energy is abundant, restore full performance. This uses ETSI NFV’s VEAS standard, altering only the MANO descriptor and orchestration without app changes [3].
Spatial Shifting:
Characterise different cloud sites (VIMs) by Power Usage Effectiveness (PUE) and Carbon Usage Effectiveness (CUE). OSM intercepts service instantiation requests and directs them to the site with the lowest CUE (i.e. cleanest energy). In practice, an Eco-Control micro-service polls each data center or edge node for current PUE/CUE. When launching or migrating a service, OSM chooses the greenest location. This means delayable workloads (like batch video analytics) can run in low-carbon sites, further cutting embodied and operational emissions.
Use Case of Maritime Port Video Analysis
This post shows how an advanced MANO platform can cut carbon footprint, in a specific use case of a maritime port. By embedding carbon-aware orchestration mechanisms (service throttling and spatial shifting) into OSM’s workflow, one can dynamically adjust compute workloads in response to energy/carbon metrics, without changing applications or hardware. In this port Automatic License Plate Recognition (ALPR) use case, this approach yielded dramatic savings of daily CO₂ emissions. To validate these claims, we outline a methodology using port surveillance video: extracting operational metrics (vehicle counts, idle times, equipment use) via computer vision, converting them to fuel use and CO₂ with standard emission factors. This rigorous evaluation highlights how tailoring MANO policies to workload characteristics enables major energy and emissions efficiency in smart ports.
In detail, the Service Throttling mechanism uses VEAS to control CPU usage. In experiments, a baseline high-performance state (e.g. 2.2 GHz, 8 cores) and a low-power state (e.g. 1.2 GHz, 8 cores) were defined. Power profiling on an Intel i7-8700 CPU showed (Table below) that lowering CPU frequency from 2.2 GHz to 1.2 GHz roughly halved power draw (8.9 W to 4.0 W for 4 cores). It also revealed that analysing one video frame at 2.2 GHz consumed ~13.57 J, and at 1.2 GHz ~13.06 J. This indicates that pure frequency throttling yields limited energy savings if the workload is CPU-bound. However, in practical workloads with idle gaps or deadline slack, throttling can lower instant power and match processing rate to green-energy availability. The orchestration logic uses a renewable-energy threshold: for instance, when solar contribution exceeds 50%, switch to the full state; when below, use the power-saving state. When implemented in the port ALPR use case, this adaptive throttling shifted most of the compute to daytime hours and idled GPU/CPU at night, flattening the carbon footprint without dropping throughput (since ALPR throughput was not needed uniformly).
The Spatial Shifting mechanism complements throttling by choosing cleaner infrastructure. Each compute site (on-premise DC or cloud region) publishes its current PUE (data-center efficiency) and CUE (kg CO₂ per kWh of power used). The OSM Eco-Control service periodically queries these values or ingests them via telemetry. When a new service (or migration) is requested, OSM evaluates all candidate sites’ CUE and directs deployment to the one with lowest carbon impact.
For instance, if the port has access to both a local server farm (brown power, PUE=1.7, high CUE) and a remote data center with mostly renewables (PUE=1.2, low CUE), spatial shifting will deploy the video analysis service remotely or schedule it for when the port’s power is greener. This approach requires extending the OSM northbound API with a “sustainable instantiation” endpoint. Internally, Eco-Control intercepts NS instantiation calls and applies a “find-min-CUE” policy. All service descriptors remain unchanged except that their placement constraints include the chosen VIM ID. The benefit is twofold: it uses existing algorithmic placement for basic load balancing but overrides it when carbon reduction is the priority. No application-level changes are needed. In effect, the MANO control loop now considers an extra dimension (carbon) besides CPU/memory cost.
Conclusion
The analysis confirms that embedding carbon feedback into MANO has a major environmental payoff. Service throttling alone (adjusting CPU states) already cut daily ALPR emissions by ~40%. When coupled with spatial shifting to greener infrastructure, total savings approached ~47% (nearly halving operational carbon). These gains come with minimal trade-offs: throughput and QoS were largely maintained because the ALPR workload was delay-tolerant. Crucially, all logic resided in the orchestration layer; no application or hardware changes were needed.
The video-based evidence strengthens the case: it shows that the smart orchestration policies translate into real-world operational metrics (reduced idling, more efficient equipment use) and thus verifiable CO₂ reductions. This kind of port-video-to-emissions pipeline can serve as ongoing monitoring, validating MANO’s impact or flagging drift if conditions change.
References
[1] A. Canete, M. Amor, and L. Fuentes, “HADES: An NFV solution for energy-efficient placement and resource allocation in heterogeneous infrastructures,” J. Netw. Comput. Appl., vol. 221, Art. no. 103764, 2024, doi: 10.1016/j.jnca.2023.103764.
[2] F. Antão, H. R. Chi, D. Corujo, V. A. Cunha, M. Sanchez and R. L. Aguiar, “Toward Decarbonized Network Function Virtualization-Based ICT Management: Standardization, Architecture, and Challenges,” in IEEE Communications Standards Magazine, vol. 10, no. 1, pp. 162-170, March 2026, doi: 10.1109/MCOMSTD.2025.3595284.
[3] ETSI, “Environmental Engineering (EE); Green Abstraction Layer (GAL); Power management capabilities of the future energy telecommu nication fixed network nodes; Enhanced Interface for power management in Network Functions Virtualisation (NFV) environments” ETSI ES 203 682 V1.2.1, Mar. 2024.
Read EXIGENCE D2.1 – Metrics for Energy Consumption and Efficiency Metering: HERE
Authors
Instituto de Telecomunicações
Haoran Chi has been with the Instituto de Telecomunicações, Portugal, since August 2019, where he is currently a Senior Researcher. During his career, he has obtained expertise knowledge on 5G (beyond) telecommunication, network automation, and machine learning. He has published >90 technical articles and successfully coordinated and managed multiple European projects. He is a Senior Editor of IEEE Transactions on Consumer Electronics, the Deputy Editor-in-Chief of IEEE Transactions on Consumer Electronics, and an Associate Editor of IEEE Transactions on Industrial Informatics and IEEE Communications Standards Magazine.
Instituto de Telecomunicações
Daniel Corujo received the Habilitation degree from the University of Aveiro, Aveiro, Portugal. He is currently an Associate Professor with the University of Aveiro and a Researcher with the Instituto de Telecomunicações, Aveiro. He has contributed to standardisation in the IETF/IRTF and IEEE. He is the author of over 130 SCOPUS-referenced articles. He teaches and researches in multiple network communications areas, such as 5G/6G mobile networks, their industrial applications, and mobile scenarios at different layers of the networking stack, particularly software-defined networks, network function virtualisation, and information-centric networking. He has been pursuing such interests in the scope of international research projects, within the Horizon Europe programme, and its previous installments. He is the Vice-Chair of the IEEE ComSoc PT Chapter.
Instituto de Telecomunicações
Filipe Antão received the bachelor’s degree in computer engineering from the University of Aveiro, Portugal, in 2023, where he is currently pursuing the master’s degree. He is currently a grant holder at IT, under the International Exigence project. His research interests include sustainable orchestration in NFV environments, 5G networks, and next-generation mobile systems.