Petabit-per-second routers: optical vs. electronic implementations

Tucker, R.S.

doi:10.1109/ofc.2006.215992

Cited by 8 publications

(4 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Delay-line buffers of this size would require 1 Tm of optical fiber, that enough reach approximately from the Sun to the planet Saturn [17]. If we use eDRAM with 100 Tb/m 2 storage density a read/write cycle time of 0.2 ns [18], 250-Tb buffer would only need 2.5 m 2 areas. Thus, these technologies can provide feasibility for our approach.…”

Section: Algorithm 1 Three-dimensional Label Scheduling Algorithmmentioning

confidence: 99%

A novel distributed scheduling approach powered by central databases in optical burst networks

Xie

Huang

et al. 2008

Photon Netw Commun

View full text Add to dashboard Cite

In terms of the growth rate, network capacity is slower than network traffic. To solve the mismatch between them, utilization of available capacity can be maximized to provide "best-fit" service for emerging applications (e.g., eBanking, Peer-to-Peer (P2P) file exchanges, etc.). To satisfy such a requirement, we propose a three-dimensional label scheduling algorithm (TD-LSA), which is a distributed scheduling approach powered by central databases. In the proposal, all connections are provisioned in three dimensions (i.e., space, time and wavelength) in a distributed way and uniquely identified by three-dimensional labels. And in this way a three-dimensional connection is more robust than a single-dimensional connection because the three-dimensional approach can avoid potential scheduling conflicts as much as possible by using central databases in a network. Thus, the central databases can reduce bandwidth consumption and offer "best-fit" service for three-dimensional connections. Furthermore, in order to further reduce bandwidth consumption and to match with a priority of traffic, K -least hop first path (K -LHPF) and a rescheduling mechanism are applied to our proposal. Simulation experiments demonstrate that our proposal achieves the expected performance gain against existing alternates.

show abstract

Section: Algorithm 1 Three-dimensional Label Scheduling Algorithmmentioning

confidence: 99%

A novel distributed scheduling approach powered by central databases in optical burst networks

Xie

Huang

et al. 2008

Photon Netw Commun

View full text Add to dashboard Cite

show abstract

“…Operations shown in parentheses are not yet commercially available. The packet and LSP layers at the top of the GMPLS stack require large buffers, primarily to deal with contention [6]. These layers also process the data stream to implement packet and label switching.…”

Section: The Photonic Bottleneck In Optical Communicationsmentioning

confidence: 99%

“…The "all-optical" packet switches in these test-beds electronically process the packet header after splitting it away from the packet payload which is optically buffered but not processed.) a1208_1.pdf OThI1.pdf It has been shown that all-optical buffering is competitive with electronic (CMOS) buffering only when the buffers are very small [6]. The processing and buffering requirements of the LSP and Packet layers make it clear that the O/E interface cannot be moved into these layers cost-effectively.…”

Section: The Photonic Bottleneck In Optical Communicationsmentioning

confidence: 99%

The Photonic Bottleneck

Hinton¹,

Farrell

Tucker³

2007

OFC/NFOEC 2007 - 2007 Conference on Optical Fiber Communication and the National Fiber Optic Engineers Conference

View full text Add to dashboard Cite

show abstract

“…It is notable, however, that the long lengths of fiber are also heavy cost. For example, an 1-Tb/s router, with link rate of 40Gb/s and 1µm single-wavelength fiber delay line buffers at each port, would have a total of 8,000 km of fiber delay line [13]. This trend may become worse for future higher capacity router applications.…”

Section: Introductionmentioning

confidence: 99%

Constructing optical LIFO buffers of size B with 3 log<inf>2</inf> B fiber delay lines

Wang

Jiang

Pattavina

2010

2010 14th Conference on Optical Network Design and Modeling (ONDM)

View full text Add to dashboard Cite

Abstract-The lack of optical buffer is still one of the main problems that hinder the development of all optical network. The current works on this topic mainly focus on emulating optical buffers with fiber delay line (FDL). Recent advances have shown the feasibility of emulating many kinds of optical buffers, such as the First In First Out (FIFO) buffer, priority buffer, etc. The Last In First Out (LIFO) buffer is another important network components that can be used for providing QoS guarantee and congestion control. Huang et al. introduced a recursive construction of LIFO buffer with buffer size B by using no less than 9 log 2 B FDLs [1]. In this paper, we try to reduce the number of required delay lines there. We will show that by combining every 3 fiber delay lines into one group and applying exponential growth of FDLs among groups, one needs only 3 log 2 B delay lines to emulate a LIFO buffers of size B.

show abstract

Petabit-per-second routers: optical vs. electronic implementations

Abstract: We compare optical and electronic buffers and switches for use in future 1-Pb/s routers. Based on the scaling properties of optical and electronic devices, we conclude that electronics will be the technology of choice for future Pb/s routers.

Cited by 8 publications

References 9 publications

A novel distributed scheduling approach powered by central databases in optical burst networks

A novel distributed scheduling approach powered by central databases in optical burst networks

The Photonic Bottleneck

Constructing optical LIFO buffers of size B with 3 log<inf>2</inf> B fiber delay lines

Contact Info

Product

Resources

About