Journal article
IEEE, Journal on Selected Areas in Communications (JSAC), vol. 32(1), 2014 Jan, pp. 16-27
APA
Click to copy
Bolla, R., Bruschi, R., Carrega, A., Davoli, F., & Lago, P. (2014). A Closed-Form Model for the IEEE 802.3az Network and Power Performance. Journal on Selected Areas in Communications (JSAC), 32(1), 16–27. https://doi.org/10.1109/JSAC.2014.140103
Chicago/Turabian
Click to copy
Bolla, Raffaele, Roberto Bruschi, Alessandro Carrega, Franco Davoli, and Paolo Lago. “A Closed-Form Model for the IEEE 802.3az Network and Power Performance.” Edited by IEEE. Journal on Selected Areas in Communications (JSAC) 32, no. 1 (January 2014): 16–27.
MLA
Click to copy
Bolla, Raffaele, et al. “A Closed-Form Model for the IEEE 802.3az Network and Power Performance.” Journal on Selected Areas in Communications (JSAC), edited by IEEE, vol. 32, no. 1, Jan. 2014, pp. 16–27, doi:10.1109/JSAC.2014.140103.
BibTeX Click to copy
@article{bolla2014a,
title = {A Closed-Form Model for the IEEE 802.3az Network and Power Performance},
year = {2014},
month = jan,
issue = {1},
journal = {Journal on Selected Areas in Communications (JSAC)},
pages = {16-27},
volume = {32},
doi = {10.1109/JSAC.2014.140103},
author = {Bolla, Raffaele and Bruschi, Roberto and Carrega, Alessandro and Davoli, Franco and Lago, Paolo},
editor = {IEEE},
month_numeric = {1}
}
🔬 IEEE 802.3az Energy Efficient Ethernet: Closed-Form Modeling of Power Consumption and Network Performance
Energy Efficient Ethernet (EEE), standardized through IEEE 802.3az, introduces Low Power Mode (LPM) mechanisms that reduce link energy consumption during periods of inactivity.
However, transitions between active and low-power states introduce non-negligible delays, creating a fundamental challenge for network designers: how can energy consumption and packet latency be accurately predicted under heterogeneous traffic conditions?
Our research addresses this problem by introducing a closed-form analytical model for jointly estimating power consumption and network performance in IEEE 802.3az Ethernet links.
The model supports three Ethernet technologies:
🔹 100BASE-TX: 100 Mbps Ethernet with independently controlled low-power transitions for the two communication directions.
🔹 1000BASE-T: Gigabit Ethernet, where entry into Low Power Mode requires both transmission and reception directions to be idle.
🔹 10GBASE-T: 10 Gigabit Ethernet with independent directional power-management behavior and technology-specific transition times.
📐 Analytical framework
The proposed methodology integrates several fundamental modeling components:
🔹 Batch-Based Traffic Modeling: Packet arrivals are characterized through burst arrival rates, batch-size distributions, and packet-size variability.
🔹 Queueing-Theoretic Representation: Each communication direction is modeled using an Mˣ/Dₘ/1/SET queueing system, incorporating batch arrivals, discrete deterministic service times, and stochastic setup delays.
🔹 Power-State Modeling: The analytical framework considers active transmission, sleep, quiet, refresh, and wake-up phases, accounting for their respective durations and power consumption.
🔹 Closed-Form Performance Estimation: Mathematical expressions are derived for average queue occupancy, mean packet waiting time, and link power consumption.
🔹 Bidirectional Link Modeling: The framework captures the coupling between transmission and reception directions required by the IEEE 802.3az operation of 1000BASE-T links.
The main advantage is the possibility of estimating energy and network performance metrics without relying on computationally expensive packet-level simulations.
🧪 Experimental validation
The model was validated using a physical testbed comprising IEEE 802.3az-compliant 1000BASE-T switches, an Ixia N2X traffic generator, and an Agilent U2353A data acquisition system for power measurements.
The evaluation employed real-world traffic traces from the MAWI archive, scaled to exercise the links' low-power operation, and considered both unidirectional and bidirectional communication scenarios.
📊 Key research findings
✅ Power estimation accuracy: Maximum relative error below 1% in the real-traffic validation.
✅ Latency estimation accuracy: Maximum relative error below 0.5%.
✅ Traffic awareness: The model accounts for heterogeneous packet sizes, arrival rates, and burst distributions.
✅ Standards coverage: The analytical formulation supports the three Ethernet speeds considered in the original IEEE 802.3az framework.
✅ Computational tractability: The use of closed-form expressions and stationary queueing analysis provides a low-complexity alternative to detailed simulation.
✅ Network planning applications: The analytical results can support the dimensioning of Ethernet infrastructures and evaluation of energy–latency trade-offs.
🎯 Key research contribution: A unified analytical framework capable of predicting both energy consumption and packet waiting times for IEEE 802.3az Ethernet links, including the distinct low-power transition mechanisms of different physical-layer technologies.
The study experimentally validates the accuracy of the proposed model. It does not establish a universal percentage of energy savings or experimentally validate a complete optimization controller. Such optimization is identified as a potential application of the analytical framework.

🌱⚡ Can Ethernet Networks Save Energy Without Sacrificing Performance? 🌐🔌
Imagine an Ethernet network that knows when to rest. Instead of consuming nearly the same amount of power regardless of traffic, its links can enter low-power states during idle periods and wake up when new packets arrive.
Sounds simple, right? The real challenge is understanding how these transitions affect energy consumption and network performance.
Our research investigates this problem through IEEE 802.3az Energy Efficient Ethernet (EEE), a standard designed to reduce unnecessary power consumption in wired networks.
The key idea is to combine intelligent power management with accurate mathematical modeling, allowing network designers to understand the relationship between energy efficiency and packet transmission delays.
⚡ Active Mode: Ethernet links transmit packets at their normal operating speed.
💤 Low Power Mode (LPM): Links reduce power consumption when no traffic needs to be transmitted.
🔄 Sleep and Wake-Up Transitions: Links move between operating states, introducing delays that must be considered.
🧠 Analytical Modeling: Mathematical expressions predict power consumption and packet waiting times under different traffic conditions.
📊 What did we discover?
✅ The proposed model estimates link power consumption with errors below 1% in the real-traffic validation.
✅ Packet latency predictions achieve errors below 0.5%.
✅ The framework supports three Ethernet technologies: 100BASE-TX, 1000BASE-T, and 10GBASE-T.
✅ The analytical approach accounts for traffic bursts, packet-size distributions, and sleep/wake-up behavior.
Our manga-inspired infographic illustrates how a seemingly simple idea—letting Ethernet links sleep when they are not needed—introduces interesting engineering challenges.
🎯 The message: Making networks greener requires more than reducing power consumption. It requires understanding exactly how energy-saving mechanisms influence service performance.
