POWER EFfiCIENCY AND DELAY TRADEOFF OF 10GBASE-T ENERGY EFfiCIENT ETHERNET PROTOCOL
ABSTRACT
In this paper, we study the power efficiency anddelay performance of the burst mode transmission (BTR) strategyfor the IEEE 802.3az energy efficient Ethernet (EEE) protocol.In the BTR strategy, the Ethernet interface goes to sleeponce its transmission buffer becomes empty and wakes upas soon as the first arrival has waited for time t or theN-th frame arrives at the interface. Based on the number ofarrivals during the vacation time, a new approach is proposedto analyze the M/G/1 queue with vacation times that aregoverned by the arrival process and the t and N parametersof BTR strategy. Our key idea is to establish the connectionbetween the vacation time and the arrival process to account fortheir dependency. We first derive the distribution of the numberof arrivals during a vacation time based on an event tree of theBTR strategy, from which, we obtain the mean vacation time andthe power efficiency. Next, from the condition on the number ofarrivals at the end of a vacation period, we derive a generalizedP-K formula of the mean delay for EEE systems, and prove thatthe classical P-K formula of the vacation model is only a specialcase when the vacation time is independent of the arrival process.Our analysis demonstrates that the t policy and N policy of theBTR strategy are compensating each other. The t policy ensuresthe frame delay is bounded when the traffic load is light, whilethe N policy ensures the queue length at the end of vacationtimes is bounded when the traffic load is heavy. These results,in turn, provide the rules to select appropriate t and N.Ouranalytical results are confirmed by simulations.
EXISTING SYSTEM:
Previous WorkSeveral analytical models have been developed to study theEEE protocol in recent years. The simplest case is consideredin , in which t =0and N =1, that is, the Wakeup istriggered by the first arrival frame. This kind of BTR strategyis also called frame transmission (FTR) scheme . Thetime is slotted according to the frame transmission time and adiscrete-time Markov chain is developed in, but it is verydifficult to solve for a closed-form solution. Thus, this paperonly obtains the power efficiency and does not derive the meandelay. derives the power efficiency through a simplemodel, which cannot be extended to study the mean delay of aframe. To obtain both the power consumption and mean delay,a thorough analytical framework for the EEE systems with thelink rate ranging from 100Mb/s to 10Gb/s is proposed i.However, previous results reported in show thatthe FTR strategy does not work well and around 86% of thepower is consumed by the Sleep and the Wakeup operationsin the worst case.Thus, uses a deterministic model to estimate theperformance of the BTR strategy with t →8and N>1,that is, only the counter is used. Despite this model obtainsboth the power efficiency and the mean waiting time, it isless accurate when N is small. On the other hand, considers the BTR strategy where only the timer is used. Thewaiting-time distribution and the power efficiency are derivedin based on a recursive relation between the waiting timesof two consecutive frames.
PROPOSED SYSTEM:
In this paper, we analyze the BTR strategy of 10Gbase-TEthernet, in which the two transmission directions are independentof each other. Our goal is to develop a unified model topredict the performance of the EEE protocol such that we canprovide rules to select the proper values of parameter t and N.The BTR strategy is modeled as an M/G/1 queue withvacation times, which is governed by the frame arrival processvia the timer threshold t and the counter threshold N.First, we show that the classical M/G/1 queue with vacationtimes cannot be directly applied to delineate theEEE protocol. We then develop a new approach to analyzethe performance of the EEE protocol. Our key ideais to establish a connection between the vacation time andthe arrival process to account for their dependency. Differentfrom the previous works, our derivation startswith the distribution of the number of arrivals during thevacation time. In particular, we formulate an event tree todelineate the complete sample space of all possible eventsgenerated from the t&N policy. This sample space of theBTR strategy enables us to derive the complete distribution ofthe number of arrivals during a vacation time. Based on thisdistribution, we obtain a generalized P-K formula of the meandelay, in which the classic P-K formula becomes a specialcase. Furthermore, we obtain the unified power efficiency andmean delay over the entire traffic rate region, from whichwe are able to forge a set of rules for selecting appropriateparameter t and N.For a fixed traffic load, we find that the power efficiencyconverges to a constant while the mean delay is unboundedwhen t and N increase, which means that t and N shouldnot be too large in the practical application of BTR strategy.Our results clearly show that the timer and the counterare compensating for each other and play different roles indifferent traffic rate regions. The timer t bounds the delayincurred by the vacation periods when the traffic rate issmaller than (N -1)/t, while the counter N limits the queuelength during the vacation time when the load is larger than(N -1)/t. In this paper, we show that, when the input trafficfluctuates around a mean traffic rate λ, a proper choice of thet and N parameters should satisfy the condition (N – 1)/t = λ. Our specific contributions are summarized asfollows:1. Based on the complete distribution of the number of arrivalsduring vacation time, we develop a new approach to analyzethe M/G/1 queue with the vacation times that are dependenton the arrival process and governed by the t and Nparameters of the BTR strategy.2. We derive a generalized P-K formula of mean delay overthe entire traffic rate region for the M/G/1 queue withvacation times governed by the arrival process. We showthat the classical P-K formula of mean delay is only aspecial case when the vacation time is independent of thearrival process.3. We show that the t policy and N policy of the BTR strategyare compensating each other. Our analysis demonstrates theimpacts of parameters t and N on power efficiency andmean delay. These results provide the rules to select theappropriate BTR parameters t and N, and they are effectiveeven under bursty input traffic.
CONCLUSION
This paper presents an analytical model for the BTR strategyof 10GBase-T Energy Efficient Ethernet and derives the powerefficiency and mean delay of the system. The proposed modelis the first approach that can dispose of the dependencybetween the vacation time and the arrival process as well as theparameters of the BTR strategy. In our model, we start withthe distribution of the number of arrivals during a vacationtime based on an event tree, and then derive the powerefficiency and a generalized P-K formula of the mean delayfor EEE systems.Our analysis shows that the counter N and the timer t playdifferent roles in performance guarantees and compensate eachother to adapt to traffic fluctuations. Therefore, the t&N pol-icy can better adapt to the traffic on Ethernet networks, whichtypically are bursty. Based on these properties, we provide therules to select appropriate values for parameters t and N.Themethodology developed in this paper can be generalized andapplied to the analysis of queueing models with vacation timesthat are governed by the arrival process.
REFERENCES
[1] M. Gupta and S. Singh, “Greening of the Internet,” in Proc. Conf. Appl.Technol. Archit. Protocols Comput. Commun. (SIGCOMM),NewYork,NY, USA, Aug. 2003, pp. 19–26.
[2] R. Bolla, R. Bruschi, F. Davoli, and F. Cucchietti, “Energy efficiencyin the future Internet: A survey of existing approaches and trends inenergy-aware fixed network infrastructures,” IEEE Commun. SurveysTuts. , vol. 13, no. 2, pp. 223–244, 2nd Quart., 2011.
[3] P. Winzer, “Beyond 100G Ethernet,” IEEE Commun. Mag., vol. 48, no. 7,pp. 26–30, Jul. 2010.
[4] P. Reviriego, K. Christensen, J. Rabanillo, and J. A. Maestro, “An initialevaluation of energy efficient Ethernet,” IEEE Commun. Lett., vol. 15,no. 5, pp. 578–580, May 2011.
[5] B. Kohl, 10 GBASE-T Power Budget Summary, IEEEStandard 802.3az Task Force Presentation Material,Piscataway, NJ, USA, 2007. [Online]. Available:http://www.ieee802.org/3/eee_study/public/mar07/kohl_01_0307.pdf
[6] M. Gupta, S. Grover, and S. Singh, “A feasibility study for powermanagement in LAN switches,” in Proc. 12th IEEE Int. Conf. Netw.Protocols (ICNP), Oct. 2004, pp. 361–371.
[7] B. Nordman. (May 2007). EEE Savings Estimates. [Online]. Available:http://www.ieee802.org/3/eee_study/public/may07/nordman_2_0507.pdf
[8] IEEE Standard for Information Technology—Local and MetropolitanArea Networks—Specific Requirements—Part 3: CSMA/CD AccessMethod and Physical Layer Specifications Amendment 5: Media AccessControl Parameters, Physical Layers, and Management Parameters forEnergy—Efficient Ethernet, IEEE Standard 802.3az-2010, Oct. 2010,pp. 1–302,
[9] P. Reviriego, J.-A. Hernadez, D. Larrabeiti, and J. A. Maestro, “Bursttransmission for energy-efficient Ethernet,” IEEE Internet Comput.,vol. 14, no. 4, pp. 50–57, Jul./Aug. 2010.
[10] K. J. Kim, S. Jin, N. Tian, and B. D. Choi, “Mathematical analysis ofburst transmission scheme for IEEE 802.3az energy efficient Ethernet,”Perform. Eval., vol. 70, no. 5, pp. 350–363, May 2013.
[11] D. P. Bertsekas, R. G. Gallager, and P. Humblet, Data Networks,vol.2.Englewood Cliffs, NJ, USA: Prentice-Hall, 1992.