Journal article
Raffaele Bolla, Alessandro Carrega, Matteo Repetto, Giorgio Robino
MDPI, Computers, vol. 7(2), 2018, p. 36
MDPI, Computers, vol. 7(2), 2018, p. 36
Cite
Cite
APA
Click to copy
Bolla, R., Carrega, A., Repetto, M., & Robino, G. (2018). Improving Efficiency of Edge Computing Infrastructures through Orchestration Models. Computers, 7(2), 36. https://doi.org/10.3390/computers7020036
Chicago/Turabian
Click to copy
Bolla, Raffaele, Alessandro Carrega, Matteo Repetto, and Giorgio Robino. “Improving Efficiency of Edge Computing Infrastructures through Orchestration Models.” Edited by MDPI. Computers 7, no. 2 (2018): 36.
MLA
Click to copy
Bolla, Raffaele, et al. “Improving Efficiency of Edge Computing Infrastructures through Orchestration Models.” Computers, edited by MDPI, vol. 7, no. 2, 2018, p. 36, doi:10.3390/computers7020036.
BibTeX Click to copy
@article{bolla2018a,
title = {Improving Efficiency of Edge Computing Infrastructures through Orchestration Models},
year = {2018},
issue = {2},
journal = {Computers},
pages = {36},
volume = {7},
doi = {10.3390/computers7020036},
author = {Bolla, Raffaele and Carrega, Alessandro and Repetto, Matteo and Robino, Giorgio},
editor = {MDPI}
}
🔬 Context-Aware Orchestration for Energy-Efficient Edge Computing Infrastructures
Edge computing is a fundamental enabler of distributed, latency-sensitive applications. However, mobility, heterogeneous workloads, and proactive resource provisioning introduce significant challenges for infrastructure efficiency and Quality of Service (QoS).
Our research introduces and experimentally evaluates an energy-efficient edge computing architecture that extends conventional Infrastructure-as-a-Service (IaaS) management with application context, programmable networking, and power-aware resource control.
The proposed framework integrates:
🔹 Application-Aware Resource Classification: Virtual machines are classified according to their service requirements and operational roles: red (critical), yellow (flexible), and green (idle).
🔹 Extended OpenStack Management: Enhanced metadata and management interfaces expose application context and support energy-aware consolidation decisions.
🔹 Software-Defined Networking: OpenDaylight and OpenFlow enable programmable forwarding paths and more efficient bandwidth utilization.
🔹 Power-State Management: ACPI S3 sleep states and Wake-on-LAN mechanisms allow unused servers and network devices to reduce power consumption while remaining recoverable.
🔹 Monitoring and Orchestration: Power measurements, CPU utilization, and VM state information support context-aware decisions about workload placement and resource activation.
📊 Experimental evaluation and findings:
✅ Energy savings: Context-aware consolidation achieved approximately 55–60% savings during low-workload periods in the evaluated scenario, compared with roughly 25–30% for conventional consolidation without application context.
✅ Performance–energy trade-off: At full workload, spreading VMs across additional servers improved processing performance by approximately 28%, but increased infrastructure power consumption by 35–44% compared with aggressive consolidation.
✅ Network efficiency: Explicit SDN path configuration reduced network congestion and packet losses relative to flooding-based forwarding at higher traffic rates.
✅ Service continuity: Live-migration experiments demonstrated that careful configuration of migration parameters can substantially reduce migration overhead while limiting application processing delays.
✅ Energy proportionality: Incorporating application context produced a closer relationship between actual computing utilization and infrastructure power consumption.
The architecture was validated through a real OpenStack/OpenDaylight experimental testbed and an elastic video-transcoding application comprising one dispatcher and nine worker virtual machines.
🎯 Key contribution: Demonstrating that awareness of application requirements and resource usage can improve the effectiveness of edge workload consolidation while supporting differentiated QoS requirements.
🌱⚡ Smarter Edge Computing: Less Energy, More Intelligence! ☁️🚀
What if edge computing infrastructures could deliver fast, reliable services while dramatically reducing unnecessary energy consumption?
As applications move closer to users, edge infrastructures must handle mobility, fluctuating workloads, and increasingly demanding Quality of Service (QoS) requirements. Traditional resource provisioning keeps virtual machines and servers running even when they are not needed, leading to significant energy waste.
Our research explores a smarter approach: context-aware orchestration and workload consolidation.
The key idea is simple but powerful: not every virtual machine needs to be active all the time!
🔴 Critical VMs: Strict performance and availability guarantees.
🟡 Flexible VMs: Controlled resource sharing and optimization.
🟢 Idle VMs: Paused until needed, allowing their hosting infrastructure to enter low-power states.
By combining OpenStack, Software-Defined Networking (SDN), and intelligent power management, the proposed framework can dynamically reorganize workloads, reduce unnecessary resource consumption, and support responsive edge services.
📊 Experiments on a real testbed demonstrate substantial energy savings, particularly during low-workload periods, while highlighting the importance of balancing consolidation aggressiveness with application performance.
🎯 The message: A greener edge is possible when infrastructure management understands not just how many resources are allocated, but how they are actually used!
