Saturday, November 21, 2026
Science
No Result
View All Result
  • Login
  • HOME
  • SCIENCE NEWS
  • CONTACT US
  • HOME
  • SCIENCE NEWS
  • CONTACT US
No Result
View All Result
Scienmag
No Result
View All Result
Home Science News Technology and Engineering

Serverless Framework Cuts Edge Telecom rApp Computing Overhead by Half

November 21, 2026
in Technology and Engineering
Denise Maddox
By Denise Maddox Scienmag Editorial Profile - Mechanical Engineering
Reading Time: 6 mins read
0
Serverless Framework Cuts Edge Telecom rApp Computing Overhead by Half

Serverless Framework Cuts Edge Telecom rApp Computing Overhead by Half

65
SHARES
587
VIEWS
Share on FacebookShare on Twitter
ADVERTISEMENT

Deploying event-driven management applications as serverless functions on multi-access edge computing platforms can reduce computational processor usage by thirty-five to fifty percent for intermittent telecom workloads, according to an experimental systems study conducted on an Open Radio Access Network testbed. Telecommunications operators adopting the O-RAN Alliance architecture increasingly host software programs known as rApps within the Non-Real-Time RAN Intelligent Controller to automate configuration, diagnose faults, and optimize radio resource efficiency. However, prevailing production deployments package every rApp as an always-on containerized microservice that continuously reserves processor and memory allocations on edge hardware, even when specific management tasks execute only periodically or during sporadic network anomalies. The experimental evaluation demonstrates that replacing static container deployments with an event-driven Function-as-a-Service execution model resolves severe idle resource waste on capacity-constrained edge nodes while remaining well within the non-real-time control loop deadlines established by telecom standards.

The study, published in the Journal of Network and Systems Management by researcher Roberto Riggio, investigates the execution mechanics of the Non-Real-Time RAN Intelligent Controller, abbreviated as Non-RT RIC, operating inside the broader Service Management and Orchestration framework. The Non-RT RIC oversees radio network policy guidance, artificial intelligence and machine learning analytics, and configuration updates over timescales exceeding one second, communicating with near-real-time controllers across standardized A1 and O1 programmatic interfaces. Third-party applications running within this layer utilize the R1 interface to access policy exposure, data management, and coordination services. While the reference software maintained by the O-RAN Software Community assumes persistent microservice execution, the research establishes that this long-running lifecycle is merely an implementation convention rather than a strict protocol requirement of the R1 service definitions, proving that functions can be instantiated on demand and released when idle without sacrificing interface compatibility.

To evaluate alternative execution paradigms on telecom edge hardware, the author constructed a physical Multi-access Edge Computing testbed running a single-node Kubernetes cluster version 1.28 on an edge-grade server configured with a sixteen-core Intel Xeon processor, sixty-four gigabytes of memory, and solid-state storage. The software stack integrated the Knative Kubernetes-native serverless platform with the O-RAN Software Community K-release Non-RT RIC services, including the A1 Policy Management Service, Information Coordination Service, and an Apache Kafka message broker. Synthetic telemetry was produced by an automated emulator generating performance measurement reports formatted in 3GPP XML schemas, while a custom generator emitted fault management alarms conforming to Virtual Event Streaming standards. An emulated near-real-time controller received and acknowledged policy directives, enabling precise end-to-end timing across the complete automated control loop from event ingestion to policy delivery.

The experimental framework introduced two specialized components to reconcile standard Function-as-a-Service semantics with telecom control requirements without altering the core O-RAN services. A dedicated Python-based Event Router bridged internal Kafka message topics to serverless HTTP endpoints, translating incoming performance counters, fault alarms, and scheduled synthetic triggers into function invocations according to declarative subscription mappings. Because Function-as-a-Service instances are stateless by design and lose in-memory variables when terminating during scale-to-zero events, the architecture incorporated an external Redis key-value store deployed within the same cluster namespace. This shared state store preserved operational context across invocations, permitting applications to track historical radio utilization counters and alarm correlation logs across cycles with sub-millisecond retrieval latencies that added negligible overhead to the overarching one-second operational budget.

The evaluation characterized three distinct rApp workload archetypes representing standard telecom operational patterns, deployed across always-on, fully serverless, and hybrid execution modes to form nine comparative profiles. A periodic energy-saving application evaluated cell sleep policies every fifteen to one hundred and twenty seconds by reading fifty performance counters, executing simple threshold algorithms, and dispatching A1 deactivation directives using an eighty-five megabyte container image. A reactive fault-remediation application monitored alarm notifications, correlating incident data with historical performance metrics from the state store to issue corrective policies from a ninety megabyte container image during network disturbances. A streaming traffic-steering application consumed continuous cell throughput statistics every second, executing policy updates from a ninety-five megabyte image to balance cell loads during real-time traffic variations.

Across two hundred measured initialization events per configuration, fully serverless scale-to-zero execution incurred median cold-start latencies of 1.8 seconds for the energy-saving application, 2.0 seconds for the fault-remediation application, and 2.1 seconds for the traffic-steering application, with ninety-fifth percentile figures reaching 2.2, 2.5, and 2.7 seconds respectively. Detailed latency decomposition revealed that pod scheduling and container setup governed by the Kubernetes control plane consumed approximately 0.6 seconds, node-local container image layer extraction required roughly 0.4 seconds, and Python interpreter startup combined with dependency imports and readiness probing absorbed the remaining 0.8 to 1.0 second. Crucially, all observed cold-start latencies remained within the operational boundaries of the Non-RT RIC, which is explicitly designed to handle control loops operating above one second where infrequent initialization delays do not disrupt network stability.

In steady-state closed-loop latency benchmarks conducted at an arrival rate of 0.1 events per second, the always-on mode delivered median response times of 0.16 seconds for energy management, 0.19 seconds for fault healing, and 0.12 seconds for traffic steering. Under fully serverless execution with a sixty-second scale-down delay, the periodic energy-saving application experienced cold starts on approximately thirty-eight percent of invocations, producing a bimodal response distribution with a median of 0.18 seconds and a ninety-fifth percentile of 2.2 seconds. The reactive fault-remediation application encountered cold starts across sixty-nine percent of invocations, raising its median latency to 2.0 seconds. Meanwhile, the sub-second cadence of the traffic-steering application prevented it from ever scaling down, maintaining warm instances that yielded a 0.12-second latency identical to always-on execution.

Under severe alarm bursts testing autoscaling responsiveness with up to five hundred concurrent notifications within a five-second interval, a single always-on replica processed events sequentially, ballooning total burst completion time to 62.1 seconds. Fully serverless execution engaged horizontal scaling up to twenty-eight concurrent replicas, shortening the completion time to 33.7 seconds, though initial scheduling contention caused 4.2 percent of events to breach the five-second operational deadline. The hybrid deployment mode—which preserves a single permanently warm replica while allowing elastic provisioning for surges—achieved the most effective performance. The hybrid configuration completed the five-hundred-alarm burst in 24.0 seconds using twenty-four peak replicas while reducing deadline drops to 1.1 percent, representing a seventy-four percent reliability improvement over the fully serverless model.

Long-term resource measurements spanning a twenty-four-hour diurnal traffic cycle demonstrated dramatic efficiency gains for sporadic workloads when abandoning static reservations. For the periodic energy-saving rApp, fully serverless deployment reduced total processor time from 86.4 to 42.1 CPU-kiloseconds, yielding a fifty-one percent compute reduction and a fifty-seven percent memory reduction from 2.8 to 1.2 gigabyte-hours. For the reactive fault-remediation application, which sat completely idle across ninety-five percent of the diurnal cycle, serverless execution slashed processor consumption by seventy-eight percent from 84.2 to 18.6 CPU-kiloseconds, lowering memory usage from 2.7 to 0.6 gigabyte-hours. Conversely, the continuous traffic-steering application registered an insignificant three percent processor reduction, demonstrating that streaming functions gain no tangible resource benefit from scale-to-zero infrastructure.

Investigating software optimization strategies, the author determined that switching from the baseline ninety-megabyte Python image to a forty-eight-megabyte Alpine base reduced median cold-start latency from 1.96 to 1.54 seconds, while a stripped thirty-one-megabyte distroless image with precompiled bytecode achieved 1.17 seconds. Crucially, runtime dependency pruning yielded greater latency reductions than physical layer compression, though container scheduling imposed an immutable platform floor of 0.57 seconds. Furthermore, pulling the baseline image from a remote registry rather than local cache expanded cold starts to 4.38 seconds, indicating that node-local registries are mandatory for viable edge serverless operations. Scaling the concurrent rApp population from three to forty applications showed that while the Redis store comfortably handled peak loads of one hundred and forty operations per second with 1.6-millisecond latency, Knative autoscaler reconciliation loops expanded from 0.4 to 2.8 seconds, demonstrating that growing application populations lengthen effective cold-start times.

Accounting for the underlying infrastructure footprint, the Knative and routing control components introduced an aggregate steady-state platform overhead of 688 millicores and 692 megabytes of memory, corresponding to 59.4 CPU-kiloseconds daily, with Knative accounting for seventy-eight percent of this cost. Break-even analysis revealed that for event-driven or balanced portfolios, the framework achieves net resource savings when hosting as few as two to three concurrent rApps, with a ten-rApp balanced deployment freeing 316 CPU-kiloseconds daily. However, an operator managing an exclusively streaming workload would require at least twenty-four concurrent applications to break even against the platform overhead, confirming that high-frequency streaming applications should remain assigned to conventional always-on containers.

The study highlights specific constraints and deployment trade-offs, emphasizing that tests were conducted on a single physical node where inter-node network synchronization, distributed storage latency, and multi-node container placement were not evaluated. Shared key-value stores also introduce security considerations, requiring strict Redis access-control lists and mutual transport-layer security to prevent unauthorized cross-application data leakage across network slices. Ultimately, the findings establish an operational taxonomy for Open RAN edge deployments: reactive rApps with event rates below 0.1 Hertz should leverage fully serverless scale-to-zero execution; periodic rApps operating between 0.01 and 0.1 Hertz achieve optimal reliability and efficiency through hybrid warm-instance provisioning; and high-frequency streaming functions should remain permanently active.

Subject of Research: Serverless execution models for event-driven O-RAN Non-RT RIC management applications on edge computing infrastructure

Article Title: Serverless rApps at the Network Edge: An Experimental Evaluation of Event-Driven Non-RT RIC Functions on Kubernetes-Based MEC

Article References: Riggio, R. (2027). Serverless rApps at the Network Edge: An Experimental Evaluation of Event-Driven Non-RT RIC Functions on Kubernetes-Based MEC. Journal of Network and Systems Management, 35(1), Article 3. https://doi.org/10.1007/s10922-026-10111-z

Image Credits: AI Generated

DOI: 10.1007/s10922-026-10111-z

Keywords: Open RAN, Non-RT RIC, rApps, Serverless Computing, Multi-access Edge Computing, Kubernetes, Knative, Function-as-a-Service, Edge Cloud, Resource Efficiency, Cold-Start Latency, Autoscaling, Telecom Infrastructure, Network Automation, State Management, Fault Remediation, Energy Saving

Cite Scienmag News

Denise Maddox. (November 21, 2026). Serverless Framework Cuts Edge Telecom rApp Computing Overhead by Half. Scienmag. https://scienmag.com/serverless-framework-cuts-edge-telecom-rapp-computing-overhead-by-half/

Denise Maddox. "Serverless Framework Cuts Edge Telecom rApp Computing Overhead by Half." Scienmag, 21 November 2026, https://scienmag.com/serverless-framework-cuts-edge-telecom-rapp-computing-overhead-by-half/. Accessed 21 November 2026.

Denise Maddox. "Serverless Framework Cuts Edge Telecom rApp Computing Overhead by Half." Scienmag. November 21, 2026. https://scienmag.com/serverless-framework-cuts-edge-telecom-rapp-computing-overhead-by-half/

Tags: autoscalingcold start latencyEdge Cloudenergy savingFault RemediationFunction-as-a-ServiceKnativeKubernetesmulti-access edge computingNetwork AutomationNon-RT RICOpen RANrAppsresource efficiencyserverless computingState Managementtelecom infrastructure
Share26Tweet16
Previous Post

Retrospective Study Examines Multimodal Imaging and Outcomes in Rare Gastrointestinal Melanoma

Next Post

Dual Plane Orchestration Framework Optimizes Distributed Artificial Intelligence Pipelines Across Networks

Related Posts

Optical Scanning Matches Traditional Sensors in Geotechnical Wall Deformation Testing
Technology and Engineering

Optical Scanning Matches Traditional Sensors in Geotechnical Wall Deformation Testing

November 21, 2026
Hybrid AI Pipeline Aligns Cross-Framework Cybersecurity Controls with 98.5 Percent Precision
Technology and Engineering

Hybrid AI Pipeline Aligns Cross-Framework Cybersecurity Controls with 98.5 Percent Precision

November 21, 2026
Synchrotron X-Ray Analysis Uncovers How Stretched Nanocomposites Reorganize Conductive Networks
Technology and Engineering

Synchrotron X-Ray Analysis Uncovers How Stretched Nanocomposites Reorganize Conductive Networks

November 21, 2026
Hybrid Deep Learning Models Accurately Classify Sugarcane Leaf Diseases Across Six Classes
Technology and Engineering

Hybrid Deep Learning Models Accurately Classify Sugarcane Leaf Diseases Across Six Classes

November 21, 2026
Fuzzy Screening Ranks Composite Record First for Potential External Prosthesis Covers
Technology and Engineering

Fuzzy Screening Ranks Composite Record First for Potential External Prosthesis Covers

November 21, 2026
Decorating Nitrogen-Doped Niobium Oxide Nanoplates with Gold Quadruples Solar Hydrogen Generation
Technology and Engineering

Decorating Nitrogen-Doped Niobium Oxide Nanoplates with Gold Quadruples Solar Hydrogen Generation

November 21, 2026
Next Post
Dual Plane Orchestration Framework Optimizes Distributed Artificial Intelligence Pipelines Across Networks

Dual Plane Orchestration Framework Optimizes Distributed Artificial Intelligence Pipelines Across Networks

  • Mothers who receive childcare support from maternal grandparents show more optimized

    Mothers who receive childcare support from maternal grandparents show more parental warmth, finds NTU Singapore study

    27656 shares
    Share 11059 Tweet 6912
  • University of Seville Breaks 120-Year-Old Mystery, Revises a Key Einstein Concept

    1061 shares
    Share 424 Tweet 265
  • Bee body mass, pathogens and local climate influence heat tolerance

    682 shares
    Share 273 Tweet 171
  • Researchers record first-ever images and data of a shark experiencing a boat strike

    546 shares
    Share 218 Tweet 137
  • Groundbreaking Clinical Trial Reveals Lubiprostone Enhances Kidney Function

    531 shares
    Share 212 Tweet 133
Science

Embark on a thrilling journey of discovery with Scienmag.com—your ultimate source for cutting-edge breakthroughs. Immerse yourself in a world where curiosity knows no limits and tomorrow’s possibilities become today’s reality!

RECENT NEWS

  • Evidence Synthesis Prioritizes Targeted Hepatitis A Vaccination Over Universal Routine Immunization in Iran
  • Optical Scanning Matches Traditional Sensors in Geotechnical Wall Deformation Testing
  • Study Shows Programmatic Aid Workers Hold Agency to Decolonise Humanitarian Practice
  • Amygdala Projections to Anterior Cingulate Cortex Promote Fear Memory Extinction

Categories

  • Agriculture
  • Anthropology
  • Archaeology
  • Athmospheric
  • Biology
  • Biotechnology
  • Blog
  • Bussines
  • Cancer
  • Chemistry
  • Climate
  • Earth Science
  • Editorial Policy
  • Marine
  • Mathematics
  • Medicine
  • Pediatry
  • Policy
  • Psychology & Psychiatry
  • Science Education
  • Science News
  • Social Science
  • Space
  • Technology and Engineering

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 5,151 other subscribers

© 2026 Scienmag - Science Magazine

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • HOME
  • SCIENCE NEWS
  • CONTACT US

© 2026 Scienmag - Science Magazine

Discover more from Science

Subscribe now to keep reading and get access to the full archive.

Continue reading