Project topics and materials
Get Complete Project Material File(s) Now! »
Layered Structure of Telecommunications
According to the Open Systems InterconnectionModel (OSI model) [12], present day telecommunications rely on layered structure of communication protocols, where each protocol is responsible for providing a level-specific link abstraction. Layered structure representation is shown on Fig. 1.2. Application Layer represents the highest layer of protocols, which are specific for an application that is being used, be it a browser showing aweb-page, Base Band Processing Unit (BBU) in C-RAN connected to a Radio Base Station located and processing In- Phase/Quadrature (I/Q) samples of radio signal, or just an File Transfer Protocol (FTP) server. This layer ensures the functioning of an application according to its needs and consists of transporting and managing data as a whole, e.g. a photo of a cat encoded in a specific format. The data is then partitioned and transmitted by protocol specific-data units (PDU), with each layer operating with their specific types of PDU. Transport Layer provides support for application layer protocols and relies on the network layer. Transport Layer protocols control the transmission of application data partitioned on a transport layer PDUs. Examples of the transport layer protocols are next: Transport Control Protocol (TCP), with TCP Segment as UDP, or User Datagram Protocol (UDP), with UDP datagram as UDP. Each TCP Segment or UDP Datagramm is then encapsulated onto IP packet, a UDP of Network layer.
Network Layer is represented by the IP protocol, which manages individual IP packets whereas Transport Layer manages a set of them. IP packets contain source and destination addresses of nodes, where the application resides. These addresses help to route an IP packet in the network: it may traverse several routers and nodes in the network before will be delivered to the destination. If transport layer provide the end-toend communications, IP has and hop-by-hop service, with intermediate nodes reading their IP addresses in order to route a packet further.
Optical Network Layers Stack Overview
Currently, the communications consist of interdependent layers of protocols depicted in Fig. 1.3 and the majority of them are managed in the Electronic domain whereas only Physical Layer can be managed in the optical domain. Without the loss of generality one can establish next scheme of the data transmission: TCP regulates the transmission of IP packets, which are put onto Ethernet, SONET or SDH frames, which are then encapsulated onto Optical Payload Unit (OPU) which are then sent in an Optical Transport Network (OTN) [16] (cf. Fig. 1.4). OTN is a set of protocols that relies on the optical media for transmitting the signal but still is managed in the electronic domain. However, the optical media is a specific one, which is represented by DenseWavelength DivisionMultiplexing (DWDM) channels. DWDM [17] primarily consists in transmitting information on up to 80 wavelengths in C-band (1530–1565 nm), where each wavelength behaves as an independent channel. DWDM technologies rely heavily on the use of an optical device, called Reconfigurable Add and Drop Multiplexer (ROADM). ROADMs can separate and multiplex wavelength channels and route them through specific ports thanks to a Wavelength Selective Switches [18]. An example of the ROADM node is represented in Fig. 1.5. There we can see an example with optical channel on 1 passing through a node without any Optical-Electronic-Optical (OEO) conversion; and add or drop channels on 2 or 3 from/on Network Interface Cards (NIC) that convert data from electronic domain onto optical one through Optical Electronic (OE) and Electronic Optical (EO) conversions. This way one can create wavelength specific topologies in the optical network.
Possible Evolution of Optical Network towards Hybrid Optical Packet Switching
In the heart of all electronic domain layers lies the technology called Electronic Packet Switching (EPS). EPS defines how the data is routed in the network to achieve its destination. EPS relies on routing the packets of data, where each packet has information about its destination, e.g. IP address in IP packet. When the packet of data enters the router or network switch, the address is read and data is sent on the required output.
Depending on the layer, such a packet can have different names and types, for example for the Network layer it is IP-packets, in the case of Ethernet/SONET/SDH/OTN layers it is frames, but the principle is the same and happens always on electronic domain: a switch/router receives a chunk of data, reads the address, decides to which output to send, and sends it.
There is a persistent interest in the scientific community in moving higher layers of communication on the optical level [20]. By going away from electronic domain dominance, one could achieve several major improvements in several characteristics, such as better latency, transparency to advance modulation formats (i.e. possibility to increase traffic volume per fiber), and energy-savings due to disappearance of numerous OEO conversions. But it’s easier said than done, which explains the research stage of such kinds of proposals. All of them have something in common: they try to replace the dominant switching mechanism of networks – EPS.Most of the proposals are centered around the introduction of a new type of switching, e.g. Optical Circuit Switching (OCS), Optical Burst Switching (OBS), Optical Packet Switching (OPS) [21] or any other hybrid solution with a mix of this (e.g. EPS) and that (e.g. OCS).
Table of contents :
Abstract
Résumé
Acknowledgments
Contents
List of Figures
List of Tables
Glossary
1 Introduction
1.1 Introduction into Optical Networks
1.2 Optical Networks for Data Centers
1.3 Thesis Structure
1.4 Thesis Contributions and List of Publications
2 State of the Art on Hybrid Optical Packet Switching
2.1 Network Strategies for Contention Resolution
2.2 Hybrid Optical Packet Switching
2.3 Conclusion
3 Research Framework
3.1 Optical Packet SwitchingModel
3.2 Hybrid Optical Packet SwitchingModel
3.3 Electronic Packet SwitchingModel
3.4 TCP SAW– Reference CCA for OPS DCN
3.5 Simulation of Communications in a Data Center Network
3.6 Conclusion
4 Congestion Control Algorithms and their Performance in DC Network
4.1 TCP Stop-And-Wait-Longer
4.2 TCP SACK and TCP mSACK
4.3 Buffer Size of a Hybrid Switch and Its Influence on Latency
4.4 Conclusion
5 Energy Consumption in DC Network on Switching
5.1 On theMetric for Energy Consumption For Data Transport
5.2 Energy Consumption on Switching in HOPS Data Center
5.3 Conclusion
6 Latency in DC Network on Different SwitchingMechanisms
6.1 DCTCP basics
6.2 Study of DCTCP performance in HOPS and OPS DCN
6.3 Latencies of different CCAs achievable in DC
6.4 Conclusions
7 Hybrid and Optical Packet Switching Supporting Different Service Classes in DC Network
7.1 Class Specific Switching Rules in OPS and HOPS
7.2 Study Conditions
7.3 Advantages of using Class Specific Switching Rules
7.4 Conclusions
8 Conclusions and FutureWork
8.1 Summary and Conclusions
8.2 Future Research Directions
A DCTCP parameters influence on Network Performance
Bibliography
