Dynamic EEE Coalescing: Techniques and Bounds

Abstract

Frame coalescing is one of the most efficient techniques to manage the low power idle (LPI) mode supported by Energy Efficient Ethernet (EEE) interfaces. This technique enables EEE interfaces to remain in the LPI mode for a certain amount of time upon the arrival of the first frame (time-based coalescing) or until a predefined amount of traffic accumulates in the transmission buffer (size-based coalescing). This paper provides new insights on the practical efficiency limits of both coalescing techniques. In particular, we derive the fundamental limits on the maximum energy savings considering a target average frame delay. Additionally, we present new open-loop adaptive variants of both time-based and size-based coalescing techniques. These proposals dynamically adjust the length of the sleeping periods in accordance with actual traffic conditions to reduce energy consumption while keeping the average delay near a predefined value simultaneously. Analytical and simulation results show that the energy consumption of both proposals is comparable to the fundamental limits. Consequently, we recommend the usage of the time-based algorithm in most scenarios because of its simplicity as well as its ability to bound the maximum frame delay at the same time.

Adaptive Frame Coalescing in Energy Efficient Ethernet with Model Predictive Control and Queuing Theory

“…Open loop control methods do not rely on the use of a gain parameter. The authors of [9] propose a dynamic tuning algorithm for both size and timer parameters to control the average queuing delay. The tuning is based on the average queuing delay expressions obtained in [20] with the assumption of Poisson arrivals.…”

“…Throughout this section, we present performance results both for MPC-mean(M ) and MPC-tail(M ) using simulations with static Poisson traffic and then Pareto interarrival times with shape parameter α = 2.5, the latter used to capture the self-similar nature of Ethernet traffic in some existing studies such as [9]. In these simulations, the frame size distribution is assumed to be of PH-type which is obtained by mixing the Erlang(10,100) and Erlang(10,1500) distributions with probabilities 0.3 and 0.7, respectively, where Erlang(k,v) stands for an Erlang distribution with order k and mean size v in units of bytes.…”

“…Throughout this subsection, we compared the effectiveness of MPC-mean with the open-loop dynamic timer-based coalescer which is proposed and shown to achieve a performance close to the lower bound in [9]. We only focused on the performance of MPC-mean(M ) with M = 2, 3 due to the fact that MPC-mean(1) is similar to the benchmark dynamic coalescer in the way that they both select the timer parameter at the instant of first arrival within the coalescing cycle and, energy consumption gains for M > 3 are relatively limited despite the increased computational burden.…”

“…This approach is in contrast with determining the timer parameter once and upon the arrival of the first frame when in the LPI state, known as timer-based coalescing studied in [8] when the arrival rates are assumed to be fixed. The dynamic version of timer-based coalescing is proposed in [9] where frame arrival rates and first two moments of service times are estimated for each coalescing cycle and the timer parameters are adjusted accordingly. In our proposed approach, for a given MPC instant at which a certain number of frames with a certain workload held in the coalescing buffer, we analytically model the residual cycle (remaining portion of the coalescing cycle) to derive expressions related to the coalescing cycle of interest including the expected waiting time, the delay violation probability, the expected duration of the coalescing cycle, etc., all obtained as a function of the timer parameter.…”

Leveraging energy saving capabilities of current EEE interfaces via pre-coalescing

“…This coalescing approach [4] clearly trades some delay for greater energy savings [5,6]. Many coalescing proposals employ the number of queued frames as the condition to exit LPI, but time-based coalescers, that use the time since the first frame arrival while in LPI, can also be configured to obtain identical results [7,8].…”