## Wavelength Division Multiple Access Channel Hypercube Processor Interconnection (1992)

Venue: | IEEE Transactions on Computers |

Citations: | 44 - 18 self |

### BibTeX

@ARTICLE{Dowd92wavelengthdivision,

author = {Patrick Dowd},

title = {Wavelength Division Multiple Access Channel Hypercube Processor Interconnection},

journal = {IEEE Transactions on Computers},

year = {1992},

volume = {41},

pages = {1223--1241}

}

### Years of Citing Articles

### OpenURL

### Abstract

A multiprocessor system with a large number of nodes can be built at low cost by combining the recent advances in high capacity channels available through optical fiber communication. A highly fault tolerant system is created with good performance characteristics at a reduction in system complexity. The system capitalizes of the self-routing characteristic of wavelength division multiple access to improve performance and reduce complexity. A hypercube based structure is introduced, where optical multiple access channels span the dimensional axes. This severely reduces the required degree, since only one I/O port is required per dimension. However, good performance is maintained through the high capacity characteristics of optical communication. The reduction in degree is shown to have significant system complexity implications. Four star-coupled configurations are studied as the basis for the optical multiple access channels, three of which exhibit the optical self-routing characterist...

### Citations

207 |
The cube connected cycles: a versatile network for parallel computation
- Preparata, Vuillemin
- 1981
(Show Context)
Citation Context ...h expansion through r, N + \DeltaN = r+1 Y i=1 m i (27) \DeltaN = N (m r+1 \Gamma 1) (28) With expansion through an increase in m i to m 0 i , N + \DeltaN = m 1 : : : m i\Gamma1 m 0 i m i+1 : : : m r =-=(29)-=- \DeltaN = N m 0 i m i \Gamma 1 (30) If the performance of a structure of size N is denoted as P(N), the Marginal Performance (MP) of the structure is defined as the incremental increase in performanc... |

154 |
Generalized Hypercube and Hyperbus Structures for a Computer Network
- Bhuyan, Agrawal
- 1984
(Show Context)
Citation Context ...imited system size has reduced the attractiveness of such configurations. The WMCH is a topological extension of the Spanning Multiaccess Channel Hypercube (SMCH) [8], the Generalized Hypercube (GHC) =-=[3]-=-, and the Spanning Bus Hypercube [35]. However, a significant improvement in performance is achieved through the physical interconnection. Topologically, the WMCH is equivalent to the SMCH and the GHC... |

115 |
A survey of interconnection networks
- Feng
- 1981
(Show Context)
Citation Context ... node communicating with nodes at a distance less than or equal to d@ , the average distance is given by: d = " OE P d@ j=1 K j # d@ X i=1 iK i + " 1 \Gamma OE N \Gamma P d@ j=1 K j # r X i==-=d@+1 iK i (12)-=- With a small locality, the graph of Figure 10 may provide a pessimistic assessment of average distance for the BC, NNMH and GHC. For example, many algorithms have been developed for the NNMH which on... |

76 |
Packet Switching in a Multi-Access Broadcast Channel with Application to Satellite Communication in a Computer Network
- Lam
- 1974
(Show Context)
Citation Context ...her continues with the current process or context switches to another process. The delay is comprised of three components: the queueing delay, the transmission time, and the user slot synchronization =-=[14,22]-=-. The average delay for the M/D/1 queueing model of TDMA is t = ae 2 = 2(1 \Gamma ae) + T p + m 2 T p (5) where the three terms represent the three components described above, ae = mT p , andsis the c... |

46 |
Communication Structures for Large Networks of Microcomputers
- Wittie
- 1981
(Show Context)
Citation Context ...ttractiveness of such configurations. The WMCH is a topological extension of the Spanning Multiaccess Channel Hypercube (SMCH) [8], the Generalized Hypercube (GHC) [3], and the Spanning Bus Hypercube =-=[35]-=-. However, a significant improvement in performance is achieved through the physical interconnection. Topologically, the WMCH is equivalent to the SMCH and the GHC: a processor has a unity distance co... |

28 |
Multiaccess Protocols in Packet Communication Systems
- Tobagi
- 1980
(Show Context)
Citation Context ...odes increases. The optical fanout of the star coupled system is about 256 [2,15] There has been much effort placed in examining optical bus-based systems, and protocols to provide access arbitration =-=[13,25,33]-=-. However, the low fanout has reduced their attractiveness with systems requiring interconnection of greater than a dozen nodes. The following section describes possible star-coupled configurations. T... |

21 |
Interconnection networks based on a generalization of cube-connected cycles
- Carlson, Cruthirds, et al.
- 1985
(Show Context)
Citation Context ...otal number of communication channels. With the WMCH, there are are S subchannel per physical link. Using earlier results, the average traffic density is ae = dN NrS=m (3) = m \Gamma 1 S N N \Gamma 1 =-=(4) sin-=-ce there are a total of Nr m channels in the system. When N AE 1, ae �� m \Gamma 1 S . B. WMCH Physical Interconnection The following section investigates the implementation of optical multiple ac... |

20 |
Optical bus and star-coupled parallel interconnection
- Dowd
- 1990
(Show Context)
Citation Context ...distance is useful since it provides a measure of the expected packet delay. Let K i represent the number of nodes at a distance i. The average distance d is defined as: d = 1 N \Gamma 1 r X i=1 iK i =-=(9)-=- As derived in [3,8], the number of nodes at a given distance for the BC, WMCH, and the GHC is: K i = ` r i ' (m \Gamma 1) i (10) since for i differing digits, each able to vary in (m \Gamma 1) ways, ... |

19 |
Performance of demand assignment multiple access schemes in broadcast bus networks
- Fine
- 1985
(Show Context)
Citation Context ... bounds for the resulting performance. A class of protocols known as Demand Assignment Multiple Access (DAMA) have been introduced in recent years to harness the high capacity nature of optical fiber =-=[13]-=-. The protocols exploited the unidirectional nature of optical fiber, but ignored the physical characteristics of the optical devices. This paper shows that the assumptions made by many such protocols... |

18 |
Multiwavelength Networks and New Approaches to Packet Switching
- Goodman
- 1989
(Show Context)
Citation Context ...o a channel is highly dependent on the characteristics of the devices. As the device characteristics improve, the number of nodes increases. The optical fanout of the star coupled system is about 256 =-=[2,15]-=- There has been much effort placed in examining optical bus-based systems, and protocols to provide access arbitration [13,25,33]. However, the low fanout has reduced their attractiveness with systems... |

18 |
Wavelength-tunable optical filters: Applications and technologies
- Kobrinski, Cheung
- 1989
(Show Context)
Citation Context ...e entire 1:3 \Gamma 1:56��m range, with a filter bandwidth of 1 nm. The setup delay for an acousto-optic device is on the order of ��S, whereas electro-optic device have on the order of nS set=-=up time [23]-=-. Tuning the acousto-optic devices occur through changes in the acoustic frequency (10-300 MHz), and precise tuning is possible. The electro-optic devices needs a drive voltage of about 10 v. Acousto-... |

11 |
A Non-Blocking Optical Interconnection Network Using Directional Couplers
- Hinton
- 1984
(Show Context)
Citation Context ...capacity. There is equal traffic on all channels, because of the uniform routing,si =sj for all i and j. The total arrival rate iss= L i , reducing the above equation to: t = d ��B \Gammasi +Q(d +=-= 1) (19)-=- The average channel arrival rate for each scheme is determined next. New arrivals,sh , generated by each local node processor are evenly distributed over its k I/O ports. Furthermore, assume that a p... |

11 |
Technology and system issues for a WDM-based fiber loop architecture
- Wagner, Lemberg
- 1989
(Show Context)
Citation Context ...he limit of the agility with currently constructed devices (estimated to be about 128 [20,23,24]) and power budget limitations restrict the number to about 128-256 processor nodes [1,2,10,16,17]. See =-=[34]-=- for a review of this technology. The system SC 3 , illustrated in Figure 3(c), avoids distributing the optical power among all output ports, and optically routes the data to the destination node. In ... |

10 |
Optical interconnects for high speed computing
- Haugen, Rychnovsky, et al.
- 1986
(Show Context)
Citation Context ...r coupled systems. The limit of the agility with currently constructed devices (estimated to be about 128 [20,23,24]) and power budget limitations restrict the number to about 128-256 processor nodes =-=[1,2,10,16,17]-=-. See [34] for a review of this technology. The system SC 3 , illustrated in Figure 3(c), avoids distributing the optical power among all output ports, and optically routes the data to the destination... |

10 |
Integrated Optics : Theory and
- Hunsperger
- 1991
(Show Context)
Citation Context ... of agility of the sources and receivers, restrict the maximum number of processors on the star coupled systems. The limit of the agility with currently constructed devices (estimated to be about 128 =-=[20,23,24]-=-) and power budget limitations restrict the number to about 128-256 processor nodes [1,2,10,16,17]. See [34] for a review of this technology. The system SC 3 , illustrated in Figure 3(c), avoids distr... |

10 |
Digital Optical Computing
- Sawchuk, Strand
- 1989
(Show Context)
Citation Context ...N), the Marginal Performance (MP) of the structure is defined as the incremental increase in performance when the network is expanded to size N + \DeltaN : MP = P (N + \DeltaN ) \Gamma P (N ) \DeltaN =-=(31)-=- For example, a measure of performance could be the product of the saturation traffic and the total number of processors. The Marginal Cost (MC) is similarly defined: MC = C(N + \DeltaN ) \Gamma C(N )... |

9 |
Optical interconnections for computer communications
- Dowd
- 1989
(Show Context)
Citation Context ...ed in greater detail in the next section, due to their superior fanout capability. Optical interconnects have been studied to achieve the cost and performance objectives discussed in the introduction =-=[9,10]-=-. There are many desirable characteristics of optical interconnects: increased fanout, a large bandwidth, high reliability, support longer interconnection lengths, exhibit low power requirements, and ... |

9 |
Twelve random access strategies for fiber optic networks
- Maxemchuk
- 1988
(Show Context)
Citation Context ...odes increases. The optical fanout of the star coupled system is about 256 [2,15] There has been much effort placed in examining optical bus-based systems, and protocols to provide access arbitration =-=[13,25,33]-=-. However, the low fanout has reduced their attractiveness with systems requiring interconnection of greater than a dozen nodes. The following section describes possible star-coupled configurations. T... |

8 |
Spanning multiaccess channel hypercube computer interconnections
- Dowd, Jabbour
- 1988
(Show Context)
Citation Context ...o power budget considerations. This limited system size has reduced the attractiveness of such configurations. The WMCH is a topological extension of the Spanning Multiaccess Channel Hypercube (SMCH) =-=[8]-=-, the Generalized Hypercube (GHC) [3], and the Spanning Bus Hypercube [35]. However, a significant improvement in performance is achieved through the physical interconnection. Topologically, the WMCH ... |

7 |
An approximate analysis of delay for fixed and variable length packets in an unslotted ALOHA channel
- Ferguson
- 1977
(Show Context)
Citation Context ...her continues with the current process or context switches to another process. The delay is comprised of three components: the queueing delay, the transmission time, and the user slot synchronization =-=[14,22]-=-. The average delay for the M/D/1 queueing model of TDMA is t = ae 2 = 2(1 \Gamma ae) + T p + m 2 T p (5) where the three terms represent the three components described above, ae = mT p , andsis the c... |

6 |
An N-stage planar optical permutation network
- Spanke, Benes
- 1987
(Show Context)
Citation Context ...ple, a measure of performance could be the product of the saturation traffic and the total number of processors. The Marginal Cost (MC) is similarly defined: MC = C(N + \DeltaN ) \Gamma C(N ) \DeltaN =-=(32)-=- The MC is obtained for the BC, NNMH, GHC and WMCH using the complexity function defined in Section 3.2.6. Figure 16 illustrates the difficulty of obtaining the excellent topological characteristics o... |

5 |
Optical clock distribution to silicon chips
- Clymer, Goodman
- 1986
(Show Context)
Citation Context ...ms represent the three components described above, ae = mT p , andsis the channel arrival rate generated per node. The average delay of SC 0 is t 0 = T p h m 2 + 1 i +so m 2 T 2 p 2[1 \Gammaso mT p ] =-=(6)-=- sinces=so whereso denotes the total arrival rate generated by each node. Assume there are S = m subchannels with SC 1 , each representing the path to a particular node. A packet transmission along wa... |

5 |
Digital high speed interconnects: a study of the optical alternative
- Hartman
- 1986
(Show Context)
Citation Context ...r coupled systems. The limit of the agility with currently constructed devices (estimated to be about 128 [20,23,24]) and power budget limitations restrict the number to about 128-256 processor nodes =-=[1,2,10,16,17]-=-. See [34] for a review of this technology. The system SC 3 , illustrated in Figure 3(c), avoids distributing the optical power among all output ports, and optically routes the data to the destination... |

5 |
The Topology, Algorithms and Analysis of a Synchronous Optical Hypergraph Architecture
- Ofek
- 1987
(Show Context)
Citation Context ...roblems addressed in [35] have ceased to be limiting factors. The goal of this work is to obtain the system complexity benefits of [35] and the topological characteristics of [3,8]. Refer to [26] and =-=[28]-=- for an analysis of the adaptability with changing traffic conditions, and the synchronization of multiple slotted channels for the the spanning bus hypercube, respectively. The format of the paper is... |

4 |
et al., "Multiwavelength Optical Crossconnect for Parallel Processing Computers
- Arthurs
- 1988
(Show Context)
Citation Context ...r coupled systems. The limit of the agility with currently constructed devices (estimated to be about 128 [20,23,24]) and power budget limitations restrict the number to about 128-256 processor nodes =-=[1,2,10,16,17]-=-. See [34] for a review of this technology. The system SC 3 , illustrated in Figure 3(c), avoids distributing the optical power among all output ports, and optically routes the data to the destination... |

4 |
HYPASS: An Optoelectronic Hybrid Package Switching System
- Arthurs, Goodman, et al.
- 1988
(Show Context)
Citation Context ...o a channel is highly dependent on the characteristics of the devices. As the device characteristics improve, the number of nodes increases. The optical fanout of the star coupled system is about 256 =-=[2,15]-=- There has been much effort placed in examining optical bus-based systems, and protocols to provide access arbitration [13,25,33]. However, the low fanout has reduced their attractiveness with systems... |

4 |
A semi-markov model for the performance of multiple-bus systems
- Mudge, Al-Sadoun
- 1985
(Show Context)
Citation Context ...ansion can be achieved through increases in r or m i for any 1sisr. Some topologies, such as the BC, can only be expanded through increases in r. With expansion through r, N + \DeltaN = r+1 Y i=1 m i =-=(27)-=- \DeltaN = N (m r+1 \Gamma 1) (28) With expansion through an increase in m i to m 0 i , N + \DeltaN = m 1 : : : m i\Gamma1 m 0 i m i+1 : : : m r (29) \DeltaN = N m 0 i m i \Gamma 1 (30) If the perform... |

3 |
A closed-form solution for the performance analysis of multiple-bus multiprocessor systems
- Irani, Onyuksel
- 1984
(Show Context)
Citation Context ...eached its destination, where p = 1=d. The total arrival rates of all ports except i, which have not reached their destination, are divided among k \Gamma 1 ports. Equation (20) reduces to:si = d ksh =-=(21)-=- This result is expected, since the channel arrival rate is greater with a larger average distance, and smaller for a larger degree. Equation (19) and (21) can be used to provide the saturation traffi... |

3 |
Wavelength-Tunable and Single Frequency Semiconductor Lasers for Photonic Communications Networks
- Lee, Zah
- 1989
(Show Context)
Citation Context ... of agility of the sources and receivers, restrict the maximum number of processors on the star coupled systems. The limit of the agility with currently constructed devices (estimated to be about 128 =-=[20,23,24]-=-) and power budget limitations restrict the number to about 128-256 processor nodes [1,2,10,16,17]. See [34] for a review of this technology. The system SC 3 , illustrated in Figure 3(c), avoids distr... |

3 | Cost-performance bounds for multimicrocomputer networks
- Reed, Schwetman
- 1983
(Show Context)
Citation Context ... r+1 Y i=1 m i (27) \DeltaN = N (m r+1 \Gamma 1) (28) With expansion through an increase in m i to m 0 i , N + \DeltaN = m 1 : : : m i\Gamma1 m 0 i m i+1 : : : m r (29) \DeltaN = N m 0 i m i \Gamma 1 =-=(30)-=- If the performance of a structure of size N is denoted as P(N), the Marginal Performance (MP) of the structure is defined as the incremental increase in performance when the network is expanded to si... |

2 |
Static Interconnection Network Extensibility based on a Marginal Performance/Cost Analysis
- Dowd, Dowd, et al.
- 1989
(Show Context)
Citation Context ...network expansion should be taken into consideration when examining performance and complexity measures. The marginal expansion of a network must be carried out while preserving the network structure =-=[11]-=-. Typically, topologies cannot be incrementally increased in size. Marginal Analysis denotes the incremental increase to performance or cost as the system is expanded from size N to N + \DeltaN . For ... |

2 |
Group Communication in Bus-Based Computer Networks
- McKinley
- 1989
(Show Context)
Citation Context ... of the problems addressed in [35] have ceased to be limiting factors. The goal of this work is to obtain the system complexity benefits of [35] and the topological characteristics of [3,8]. Refer to =-=[26]-=- and [28] for an analysis of the adaptability with changing traffic conditions, and the synchronization of multiple slotted channels for the the spanning bus hypercube, respectively. The format of the... |

1 |
Product-Form Solution Techniques for the Performance Analysis of Multiple-Bus Multiprocessor Systems with Non-uniform Memory Reference
- Chiola, Marsan, et al.
- 1988
(Show Context)
Citation Context ...e components: the queueing delay, the transmission time, and the user slot synchronization [14,22]. The average delay for the M/D/1 queueing model of TDMA is t = ae 2 = 2(1 \Gamma ae) + T p + m 2 T p =-=(5)-=- where the three terms represent the three components described above, ae = mT p , andsis the channel arrival rate generated per node. The average delay of SC 0 is t 0 = T p h m 2 + 1 i +so m 2 T 2 p ... |

1 |
Performance Evaluation of the Spanning Multiaccess Channel Hypercube Interconnection Network
- Dowd, Jabbour
- 1987
(Show Context)
Citation Context ...ensions. Each i dimensional channel has m i processors attached where the processors have identical addresses except in position i of the r-tuple, allowing simple, totally distributed routing schemes =-=[7,8]-=-. A WMCH with r = 3 and N = 4 \Theta 3 \Theta 2 is illustrated in Figure 1(a), and a 2-dimensional example with N = 4 2 is shown in Figure 1(b). When m i = m for all 1sisr, the total number of process... |

1 |
One-Sided Rearrangeable Optical Switching Network
- Hill
- 1986
(Show Context)
Citation Context ...ng assumes a system size of N nodes at a degree of k. The cost of the node processor is neutral since equal sized networks are used in the comparison. However, the term c np N is included in Equation =-=(18)-=- to represent the cost of N node processors. A communication processor is responsible for capturing and buffering packets during transmission and reception, providing media access control and logical ... |