Banner image placeholder
Banner image
Site avatar
Alex Carrega
Ph.D. IT & Network Engineer

Curriculum vitae



+39 348-74.85.497


TNT Lab @ UniGe

CNIT

via Opera Pia 13, 16154 Genoa, Italy



Trading off Power Consumption and Delay in the Execution of Network Functions by Dynamic Activation of Processing Units


Conference paper


Raffaele Bolla, Roberto Bruschi, Alessandro Carrega, Franco Davoli, Chiara Lombardo
2022 IEEE 8th International Conference on Network Softwarization (NetSoft), 2022, pp. 1-6


Cite

Cite

APA   Click to copy
Bolla, R., Bruschi, R., Carrega, A., Davoli, F., & Lombardo, C. (2022). Trading off Power Consumption and Delay in the Execution of Network Functions by Dynamic Activation of Processing Units. In 2022 IEEE 8th International Conference on Network Softwarization (NetSoft) (pp. 1–6). https://doi.org/10.1109/NetSoft54395.2022.9844039


Chicago/Turabian   Click to copy
Bolla, Raffaele, Roberto Bruschi, Alessandro Carrega, Franco Davoli, and Chiara Lombardo. “Trading off Power Consumption and Delay in the Execution of Network Functions by Dynamic Activation of Processing Units.” In 2022 IEEE 8th International Conference on Network Softwarization (NetSoft), 1–6, 2022.


MLA   Click to copy
Bolla, Raffaele, et al. “Trading off Power Consumption and Delay in the Execution of Network Functions by Dynamic Activation of Processing Units.” 2022 IEEE 8th International Conference on Network Softwarization (NetSoft), 2022, pp. 1–6, doi:10.1109/NetSoft54395.2022.9844039.


BibTeX   Click to copy

@inproceedings{bolla2022a,
  title = {Trading off Power Consumption and Delay in the Execution of Network Functions by Dynamic Activation of Processing Units},
  year = {2022},
  pages = {1-6},
  doi = {10.1109/NetSoft54395.2022.9844039},
  author = {Bolla, Raffaele and Bruschi, Roberto and Carrega, Alessandro and Davoli, Franco and Lombardo, Chiara},
  booktitle = {2022 IEEE 8th International Conference on Network Softwarization (NetSoft)}
}

🔬 Trading Off Power Consumption and Delay in Network Function Execution Through Dynamic Processing-Unit Activation

Network softwarization and virtualization enable flexible deployment of network functions on general-purpose computing infrastructures.

However, increased programmability does not automatically translate into improved energy efficiency. Processing resources may continue consuming power even when network traffic is insufficient to justify keeping all available computing units active.

Our research addresses this challenge through a queueing-based smart-sleeping control approach, in which physical or virtual processing units are dynamically activated or deactivated according to traffic conditions.

The fundamental objective is to investigate the relationship between power consumption, processing capacity, and packet-processing delay.

⚙️ Analytical Framework

The proposed approach relies on several interconnected concepts.

🔹 1. Multi-Server Queueing Model

Network-function execution is represented through a multi-server queueing system.

Incoming packets require processing by a set of computing resources, while the number of available servers corresponds to the number of active processing units.

This abstraction enables the analysis of how changes in processing capacity affect queueing behavior.

🔹 2. Dynamic Processing-Unit Activation

Physical or virtual processing units can transition between active and inactive states.

Reducing active capacity may lower power consumption, while increasing the number of available processing units can help accommodate higher traffic demand.

🔹 3. Smart-Sleeping Control

The framework considers a parametric control strategy for adjusting active processing capacity.

Control decisions operate on a time scale compatible with longer-term traffic-flow dynamics, rather than requiring continuous reconfiguration for every arriving packet.

🔹 4. Energy–Performance Trade-Off

Deactivating processing resources can create energy-saving opportunities, but lower processing capacity may increase the time packets spend waiting for service.

Consequently, resource-management decisions must consider both energy consumption and performance requirements.

🔄 Conceptual Control Workflow

The technical infographic illustrates the underlying control principle through a sequence of operations:

  1. Observe Traffic: Characterize the demand imposed on network-function processing resources.

  2. Evaluate Capacity: Consider how many processing units are required under the current traffic conditions.

  3. Adjust Active Resources: Activate additional processing units or allow unnecessary ones to enter low-power states.

  4. Assess Performance: Consider packet-processing delay alongside power consumption.

  5. Adapt Over Time: Reconsider resource allocation as longer-term traffic conditions change.

This is an explanatory representation of the smart-sleeping concept, not a verified reproduction of the complete implementation algorithm.

📊 Research Significance

✅ Energy Awareness: Introduces explicit processing-resource management rather than relying exclusively on virtualization and consolidation.

✅ Queueing-Based Analysis: Connects processing capacity to packet waiting and service behavior.

✅ Adaptive Resource Management: Allows the number of active processing units to vary with traffic conditions.

✅ Power–Delay Trade-Off: Recognizes that lower energy consumption and shorter processing delays may require different resource configurations.

✅ Softwarized Networking: Addresses energy management for network functions executed using general-purpose processing hardware.

✅ Future Network Relevance: Provides analytical foundations relevant to energy-aware virtualization and the sustainability objectives of evolving network architectures.

🎯 Key Research Contribution

A modeling and control approach for energy-aware execution of network functions through dynamic activation of processing units, investigating how adaptive computing capacity can balance power consumption and packet-processing delay.

This research highlights a fundamental challenge in sustainable network engineering: resource consolidation is beneficial only when supported by control mechanisms that account for actual workload requirements and their performance implications.

This research highlights a fundamental challenge in sustainable network engineering: resource consolidation is beneficial only when supported by control mechanisms that account for actual workload requirements and their performance implications.

🌱⚡ What If Network Processors Could Sleep When They Are Not Needed? 🌐🚀

Imagine a network where processing resources automatically adapt to traffic demand. During quiet periods, unnecessary CPU cores enter low-power states. When traffic increases, additional processing capacity becomes available.

Sounds simple, but there is a challenge: How can we reduce power consumption without introducing excessive packet-processing delays?

Our research explores this question in “Trading off Power Consumption and Delay in the Execution of Network Functions by Dynamic Activation of Processing Units.”

Network softwarization and virtualization increase flexibility and programmability. However, deploying network functions on general-purpose hardware does not automatically guarantee energy efficiency.

💤 Smart Sleeping: Dynamically adjust the number of active processing units according to traffic demand.

⚙️ Multi-Core Processing: Model network-function execution through multiple servers representing physical or virtual processing units.

📦 Packet Queues: Account for packets waiting to be processed when computing capacity changes.

📊 Traffic-Aware Adaptation: Adjust processing resources according to longer-term traffic-flow dynamics.

⚖️ Energy–Delay Trade-Off: Balance potential reductions in power consumption against the need for responsive packet processing.

The central idea is to avoid keeping unnecessary resources active while preserving sufficient capacity for network-function execution.

Our manga-inspired infographic illustrates how intelligent resource management can contribute to more energy-aware softwarized networks.

🎯 The message: Virtualization alone does not make networking greener. Energy efficiency also requires intelligent control over the processing resources that execute virtualized network functions.

The infographic presents the analytical concept rather than experimentally established savings.

Virtualization alone does not make networking greener. Energy efficiency also requires intelligent control over the processing resources that execute virtualized network functions.


Share

Text Only Owlstown
QR Code
Translate to