Elastic all-optical multi-hop interconnection in data centers with adaptive spectrum allocation
2017 (English)In: Optics Communications, ISSN 0030-4018, E-ISSN 1873-0310, Vol. 383, 478-484 p.Article in journal (Refereed) Published
In this paper, a novel flex-grid all-optical interconnect scheme that supports transparent multi-hop connections in data centers is proposed. An inter-rack all-optical multi-hop connection is realized with an optical loop employed at flex-grid wavelength selective switches (WSSs) in an intermediate rack rather than by relaying through optical-electric-optical (O-E-O) conversions. Compared with the conventional O-E-O based approach, the proposed all-optical scheme is able to off-load the traffic at intermediate racks, leading to a reduction of the power consumption and cost. The transmission performance of the proposed flex-grid multi-hop all-optical interconnect scheme with various modulation formats, including both coherently detected and directly detected approaches, are investigated by Monte-Carlo simulations. To enhance the spectrum efficiency (SE), number-of-hop adaptive bandwidth allocation is introduced. Numerical results show that the SE can be improved by up to 33.3% at 40 Gbps, and by up to 25% at 100 Gbps. The impact of parameters, such as targeted bit error rate (BER) level and insertion loss of components, on the transmission performance of the proposed approach are also explored. The results show that the maximum SE improvement of the adaptive approach over the non-adaptive one is enhanced with the decrease of the targeted BER levels and the component insertion loss.
Place, publisher, year, edition, pages
Elsevier, 2017. Vol. 383, 478-484 p.
Data center network, Flex-grid, Multi-hop, Optical circuit switching, Optical interconnect, Bit error rate, Electric switches, Insertion losses, Intelligent systems, Monte Carlo methods, Optical interconnects, Optical switches, Plasma diagnostics, Data center networks, Multihop, Adaptive optics
IdentifiersURN: urn:nbn:se:kth:diva-195118DOI: 10.1016/j.optcom.2016.09.027ISI: 000386870700073ScopusID: 2-s2.0-84988970672OAI: oai:DiVA.org:kth-195118DiVA: diva2:1048317
Funding Details: VR, Swedish Research Council. QC 201611212016-11-212016-11-022016-11-29Bibliographically approved