P. Dowd, J. Perreault, J. Chu, D. C. Hoffmeister, R. Minnich, D. Burns, F. Hady, Y. J. Chen, and M. Dagenais, "Lightning network and systems architecture," J. Lightwave Technol. 14, 1371--1387 #1996#.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....queueing networks; IP packets 1. Introduction In recent years, optical broadcast and select local area network (LAN) architectures based on wavelength division multiplexing (WDM) have been a subject of research both theoretically [2,6,8 10,12,14 17,19,23,24,28,34] and experimentally [11,18,13,7]. The issues that have been addressed include the design of efficient media access control (MAC) protocols [6,8,31,34] the development of scheduling algorithms to mask the transceiver tuning latency [2,5,25,28,30] multicasting [4,25,29,35] and dynamic load balancing [3] Because of the ....

P. Dowd et al., Lightning network and systems architecture, J. Lightwave Technol. 14 (6) (1996) 1371--1387.

.... reduces the impact of the speed mismatch by partitioning the enormous bandwidth into multiple, multi access channels that operate at speeds compatible with the electronic interface [2] Research projects are under development at universities and labs to build systems based on these WDM techniques [3 7]. The objective of this paper is to examine the performance of WDM based network architectures for large scale communication networks. The main objective is low latency, low cost communication that is sufficiently scalable to support thousands of nodes. A primary design constraint, needed to ....

....is low latency, low cost communication that is sufficiently scalable to support thousands of nodes. A primary design constraint, needed to achieve the lowcost and scalability objectives, is the assumption that there will be only a few WDM channels (many less than the number of processors) [7]. This work was supported in part by the Research Grants Committee of the University of North Carolina at Greensboro, where part of the research was conducted. Part of the research utilized computing facilities at the North Carolina Supercomputing Center. The C parallel channels offered by WDM ....

[Article contains additional citation context not shown here]

P. Dowd, J. Perreault, J. Chu, D. Hoffmeister, R. Minnich, F. Hady, Y.-J. Chen, M. Dagenais, and D. Stone, "Lightning Network and Systems Architecture," Journal of Lightwave Technology, pp. 1371-- 1387, June 1996.

....the network architecture and the multiple access protocol based on reservation and scheduling. It also provides the formulation of the scheduling problem addressed in this paper. 2. 1 Network Architecture The network architecture is based on the multi level hierarchical architecture described in [17]. Each level in the network is organized as a star and uses a wavelength partitioner that forwards a set of wavelengths to the upper level and retains the remaining at the current level. In this paper we focus on a single level system where each node has one tunable transmitter and one tunable ....

.... if we try to minimize the effect of switching (tuning latency for WDM networks) in addition to minimizing transmission time [21] At higher transmission speeds, computational complexity of the scheduling algorithms assumes critical importance in achieving better performance as shown in [2, 17]. In this paper, we present two scheduling algorithms called SEQSAM (SEQuential Scheduling AlgorithM) and BALSAM (BALanced Scheduling AlgorithM) SEQSAM exploits the tunability of receivers by allowing receivers to tune to multiple channels within the data phase. BALSAM restricts tunability of ....

P. Dowd, J. Perreault, J. Chu, D. Hoffmeister, et al.: Lightning Network and Systems Architecture, Journal of Lightwave Technology, volume 14, no. 6, June 1996, pp. 1371--1387.

....in Section 5. Section 6 summarizes the paper and provides the conclusions. 2. Protocols incorporating Reservation This section describes the multiple access protocol based on reservation and scheduling. The protocols are designed for the multi level hierarchical architecture described in [8]. Each level in the network is organized as a star and uses a wavelength partitioner that forwards a set of wavelengths to the upper level and retains the remaining at the current level. HRP TSA [21] is defined here for a single level system where each node has one laser array transmitter and one ....

P. Dowd, J. Perreault, J. Chu, D. Hoffmeister, R. Minnich, F. Hady, Y.-J. Chen, M. Dagenais, and D. Stone, Lightning Network and Systems Architecture, IEEE/OSA Journal of Lightwave Technology (1996), no. 6, 1371--1387.

....clusters are organized using a WDM star topology connected through a central passive star coupler. The available wavelengths are partitioned into two categories: local access (labeled 1 . L ) and global access (labeled L 1 . C ) The star can be implemented using a wavelength partitioner [10] which filters the local access wavelengths at level 1 and allows only the global access wavelengths to be transmitted into the level 2 network. Each end node has a single tunable transmitter, and two tunable or fixed receivers (one for local and one for global access) Each node has a single ....

....node using the local wavelengths. This eliminates the single point of fault due to the router nodeand shares the processing among multiple nodes. Note: The optical network below the star can be a singlelevel structure as shown in the figure or a multi level hierarchical structure as studied in [10]. The figure shows a singlelevel structure for the sake of clarity. Level 2 ATM network: The level 2 wavelengths carrying global traffic are then connected to the input port of an ATM switch. Note that each star is connected to the ATM switch using a single fiber. A multi channel connector such ....

P. Dowd, J. Perreault, J. Chu, D. Hoffmeister, R. Minnich, F. Hady, Y.-J. Chen, M. Dagenais, and D. Stone, "Lightning Network and Systems Architecture," IEEE/OSA Journal of Lightwave Technology, June 1996.

....Fig. 2. Star Topology based on transmitter and receiver array. The architecture considered is a passive star topology, and each node has one tunable transmitter and one tunable receiver as shown in Fig. 2. This architecture can be extended to the multi level hierarchical architecture described in [10]. Here, each level in the network is organized as a star and uses a wavelength partitioner that forwards a set of wavelengths to the upper level and retains the remaining at the current level. The medium access control protocol studied for this architecture is LiteMAC [4] Transmission in LiteMAC ....

P. Dowd, J. Perreault, J. Chu, D. Hoffmeister, R. Minnich, F. Hady, Y.-J. Chen, M. Dagenais, and D. Stone, "Lightning Network and Systems Architecture," Journal of Lightwave Technology, pp. 1371--1387, June 1996.

....but it also means that we cannot continue sending according to the original bus schedule. Instead, the bus clock must be reset when initialization mode is entered. Moreover, if collisions are detected while in this mode, all nodes must reset their clocks [6] In the Lightning architecture [2], the lockstep method is used. During start up, a node is not allowed to send until its predecessor has transmitted, except for a dedicated first node. A time out is used to detect node failures and allow nodes to start sending even if their predecessor has failed. 4 System, Communication, and ....

P. Dowd, J. Perreault, J. Chu, D. Hoffmeister, R. Minnich, D. Burns, F. Hady, Y. Chen, M. Dagenais, and D. Stone. LIGHTNING network and systems architecture. IEEE/OSA Journal Of Lightwave Technology, 14(6):1371--1387, 1996.

....the implementation of pipelined optical buses [1] Alternatively, Wavelength Division Multiplexing (WDM) can be used to create multiple channels on which multiple messages can simultaneously be transmitted. The multiple channels can be either statically or dynamically allocated to processors [2]. Utilizing the ability to transmit multiple messages simultaneously on optical buses (in pipelined fashion or using multiple channels) researchers have described efficient algorithms for parallel computing based on the message passing paradigm (See [3] for an excellent survey) To the best of ....

D.C. Hoffmeister et al, "Lightning network and system architecture," in Parallel Computing using Optical Interconnections (K. Li, Y. Pan, S.Q. Zheng, ed.), Kluwer Academic Press, 1998.

....between them than in traditional, electronic interconnection networks, but in practice, the limitations imposed by transmitters, detectors, splitters, etc. permit only ten to twenty channels per fiber [3, 4] One approach to overcome this limitation is to organize processing nodes hierarchically [5, 6, 7, 8, 9, 10, 11, 12]. Nodes are interconnected in small clusters and these clusters are connected at a 1 new level either through interface nodes or through an entirely new network superimposed on the small clusters. This process is repeated as many times as necessary to achieve a network of the desired number of ....

....in parallel processing systems and this could result in some channels going unused in one portion of a IN while performance in another portion of the network is suffering due to overuse of the available channels. One approach to this problem that has been proposed is dynamic bandwidth reallocation [5, 10] wherein a system tracks some parameters in the network and dynamically reassigns channels to the areas with the greater needs. This reallocation approach, which we call dynamic channel allocation (DCA) should be analyzed with respect to need in HORN and a design algorithm should be specified. ....

[Article contains additional citation context not shown here]

P. Dowd, J. A. Perreult, J. Chu, D. C. H. R. Minnich, D. Burns, F. Hady, Y. Chen, M. Dagenais, and D. Stone, "LIGHTNING Network and System Architecture," IEEE Journal of Lightwave Technology, vol. 14, pp. 1371 -- 1387, June 1996.

....levels above the local rings it becomes practical to reduce the number of slots to a number significantly less than the number of nodes competing for channel access and therefore TDMA arbitration will only be used at levels above the lowest layer. 7. 3 FatMAC As mentioned previously, FatMAC 1 [38, 66] is also a reservation or arbitration protocol but it is more sophisticated than TDMA arbitration. Reservation based protocols fall into two broad categories: those that use a dedicated control channel for reservation and those that don t [66] Under TDMA arbitration, data channel arbitration (and ....

....7.3 FatMAC As mentioned previously, FatMAC 1 [38, 66] is also a reservation or arbitration protocol but it is more sophisticated than TDMA arbitration. Reservation based protocols fall into two broad categories: those that use a dedicated control channel for reservation and those that don t [66]. Under TDMA arbitration, data channel arbitration (and allocation) is performed electronically, effectively placing that protocol into the dedicated control channel category, even though a dedicated optical channel isn t used. FatMAC, however, performs the reservation optically, using the first ....

[Article contains additional citation context not shown here]

P. Dowd, J. A. Perreult, J. Chu, D. C. H. R. Minnich, D. Burns, F. Hady, Y. Chen, M. Dagenais, and D. Stone, "LIGHTNING Network and System Architecture," IEEE Journal of Lightwave Technology, vol. 14, pp. 1371 -- 1387, June 1996.

....unique delay reduces the risk of collisions, but it also means that we cannot continue sending according to the original bus schedule. Instead, the bus clock must be reset when initialization mode is entered and when collisions are detected while in this mode [6] In the Lightning architecture [3] the lock step method is used. During startup, a node is not allowed to send until its predecessor has transmitted, except for the first node. In this paper we propose two alternative methods, and compare them with the Lock Step method. To minimize disruption of the running applications, ordinary ....

....for timeslot assignments potentially causing infinite collisions are listed in Table 1. 4. Startup algorithms In this section three different startup algorithms will be described. Zero first (ZF) Increment Decrement (ID) and Lock step (LS) Algorithm LS has been proposed previously [3] and is included for comparison. For all algorithms, we assume that the nodes accept the time contained in the first message that is sent without collisions. Once this is received, there will not be any subsequent collisions. The goal is thus to send this first message without collisions. 4.1 ....

[Article contains additional citation context not shown here]

P. Dowd, J. Perreault, J. Chu, D. C. Hoffmeister, R. Minnich, D. Burns, F. Hady, Y. J. Chen, M. Dagenais, and D. Stone, "LIGHTNING network and systems architecture," IEEE/OSA Journal Of Lightwave Technology, vol. 14, no. 6, pp. 1371-1387, 1996.

....of potential protocols for this type of application but many are inappropriate or unacceptable for one reason or another. Preliminary analysis narrowed consideration and indepth analysis down to five promising protocols: time division multiple access (TDMA) TDMA with arbitration, FatMAC [2, 3], DMON [4] and token hierarchical optical ring network (THORN) These are analyzed in terms of average delay, system throughput, node complexity, bandwidth and scalability. They are also assessed more qualitatively with respect to their abilities to accomodate bursty communication traffic. The ....

.... severely limits their practicality in parallel processing systems [5, 6, 7] Many researchers have proposed circumventing this limit by arranging INs hierarchically in which PNs are interconnected to each other in clusters and these clusters are in turn connected one another in largers clusters [8, 9, 10, 11, 2, 3, 12]. Proposed topologies include star of stars, bus of buses, and ring of rings. Each cluster in a hierarchy follows the 20 channel limit but channel reuse [7, 2, 13, 14] across spatially separated structures creates a system in which many times that limit is possible, permitting optically connected ....

[Article contains additional citation context not shown here]

P. Dowd, J. A. Perreult, J. Chu, D. C. H. R. Minnich, D. Burns, F. Hady, Y. Chen, M. Dagenais, and D. Stone, "LIGHTNING network and system architecture," IEEE Journal of Lightwave Technology, vol. 14, pp. 1371 -- 1387, June 1996.

.... of the speed mismatch by partitioning the enormous bandwidth into multiple, multi access, more manageable channels that operate at speeds compatible with the electronic interface [13] This paper describes the media access protocol implemented in a WDM experimental testbed known as LIGHTNING [5]. This is a joint project involving the State University of New York at Buffalo, University of Maryland (College Park and Baltimore County, MD) Laboratory for Physical Sciences (College Park, MD) Supercomputing Research Center (Bowie, MD) and David Sarnoff Research Center (Princeton, NJ) The ....

.... is the assumption that there will be only a few WDM channels (many less than the number of processors) A scalable, hierarchical all optical architecture has been developed to achieve the system objectives of low latency and low cost through a combination of spatial reuse of WDM channels [5,6]. The intention is to provide a reconfigurable structure that supports DSM and capitalize on any reference locality above uniform. Cache coherence protocols (snooping or directory based) have to be provided to maintain cache consistency across all the processors [16] The performance of the DSM ....

[Article contains additional citation context not shown here]

DOWD, P., PERREAULT, J., CHU, J., HOFFMEISTER, D., MINNICH, R., HADY, F., CHEN, Y.-J., DAGENAIS, M., AND STONE, D. Lightning Network and Systems Architecture. IEEE/OSA Journal of Lightwave Technology 14, 6 (June 1996), 1371--1387.

No context found.

P. Dowd, J. Perreault, J. Chu, D. C. Hoffmeister, R. Minnich, D. Burns, F. Hady, Y. J. Chen, and M. Dagenais, "Lightning network and systems architecture," J. Lightwave Technol. 14, 1371--1387 #1996#.

No context found.

P. Dowd et al. Lightning network and systems architecture. Journal of Lightwave Technology, 14(6):1371--1387, June 1996.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC