Conference Paper

Power consumption issues in future high-performance switches and routers

Inst. of Broadband Commun., Vienna Univ. of Technol., Vienna
DOI: 10.1109/ICTON.2008.4598688 In proceeding of: Transparent Optical Networks, 2008. ICTON 2008. 10th Anniversary International Conference on, Volume: 3
Source: IEEE Xplore

ABSTRACT The underlying demand for network capacity can only be satisfied by contemporaneously increasing the transmission bit rate and the switching capacity. An increase of data transmission and processing speed unavoidable leads to high requirements on power supply. Especially in the case of high-capacity electronic routers, the question of power consumption will be the major issue and probably the most important limiting factor as routers become larger in the future. Using optical switching fabrics could relax the limitations to some extent, but optical buffers occupy larger area and dissipate more power than electronic ones. In this paper, we address power consumption issues in future high-capacity switching elements. Different technologies and switching paradigms such as electronic packet/circuit switching or optical circuit/burst/packet switching are considered. In particular, we consider four different node architectures as candidates for future high-performance network elements. These architectures are described and examined with regard to power consumption.

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In the next 10 to 15 years the Internet will undergo a substantial increase especially with respect to the bandwidth required by end-users. Since the current Internet already consumes a not-negligible percentage of the total world electricity, reducing the energy consumption of telecom networks is expected to become an increasingly-important challenge, being unacceptable that the Internet energy consumption grows proportionally to the served bandwidth. In this paper we focus on backbone transport networks, that serve large aggregated amount of traffic. We compare three different network architectures which implement the Wavelength Division Multiplexing (WDM)-based transport of IP packets over optical fiber links (IP-over-WDM networks). The differences between these three architectures, which we identify as IP with no Bypass (IP-NB), IP with Bypass (IP-B) and IP with Bypass and Grooming (IP-BG), concern the capabilities of performing aggregation of traffic (grooming) and optical switching. IP-NB architecture performs grooming in every network node, where the traffic is electronically processed and forwarded by the IP routers. IP-B enables switching of wavelength channels directly in the optical domain, thus bypassing the processing of IP packets in the intermediate IP routers. This architecture does not provide grooming capabilities, but it just allows aggregation of different traffic demands to be established between the same source/destination pairs. IP-BG architecture represents an intermediate solution between the previous, since it provides both grooming capabilities (as for IP-NB) in order to efficiently exploit network capacity, and optical switching (as for IP-B) to reduce expensive electronic processing operations. We perform a comparative study between these three architectures showing the trade-off between the reduction of the power consumption or the cost of the networks, and we analyze how minimizing one of these two factors can influence the other.
    Communications (LATINCOM), 2010 IEEE Latin-American Conference on; 10/2010
  • [Show abstract] [Hide abstract]
    ABSTRACT: This paper outlines the market for transparent optical networks and how these have developed, from their simple origins of saving the costs of OEO regeneration in point-to-point WDM links, to today's ring and mesh networks with flexible/reconfigurable nodal traffic access, using multi-way ROADM's. The paper then reviews and compares the technology options for multi-way ROADM's and the impact that full optical transparency has on the overall network design. The paper also considers alternative approaches, using OEO conversion and electrical switching, and suggests where these or transparent optical switching may be most applicable in an increasingly cost-sensitive environment.
    Transparent Optical Networks, 2009. ICTON '09. 11th International Conference on; 08/2009
  • [Show abstract] [Hide abstract]
    ABSTRACT: Power consumption is expected to become the main limiting factor for scaling the current network architectures to capabilities of hundreds of terabit or even petabits. The use of optical switching fabrics (SFs) could relax the limitations to some extent but large optical buffers occupy larger area and dissipate more power than electronic ones. In this paper, we evaluate the power consumption of bufferless optical packet switches (OPSs), using the wavelength conversion to solve the output packet contentions. Sophisticated analytical models are introduced to evaluate the power consumption of synchronous and asynchronous OPSs (SOPSs and AOPSs) versus the offered traffic, the main switch parameters, and the used device characteristics. The power consumption in SOPSs and AOPSs is compared when commercial semiconductor optical amplifiers are used to implement SFs and wavelength converters (WCs). The obtained results show that the high power consumption in synchronization stage makes SOPS less effective than AOPS in terms of power consumption. For instance, when the OPSs are dimensioned with a sufficient number of WCs and offered traffic is 0.8, SOPS consumes 140% of power more than does the AOPS. Finally, though power consumption due to cooling system is not considered in the proposed model, we observe that both SOPSs and AOPSs consume much less power per gigabit per second carried than commercial routers.
    Journal of Lightwave Technology 01/2010; 28(5):847-857. · 2.56 Impact Factor