Anda di halaman 1dari 6

Routing Solutions for Future Dynamic Networks

Auristela Silva, Tarciana Silva, Reinaldo Gomes, Luciana Oliveira, Igor Canana, Djamel Sadok , and 2Martin Johnsson
1

Federal University of Pernambuco, Brazil, E-mail: {auristela, tarciana, reinaldo, lpo, icc, jamel}@gprt.ufpe.br 2 Ericsson Research, Sweden, E-mail: martin.johnsson@ericsson.com

AbstractDynamic Networks with users from heterogeneous networks, different operator, provider or technology domains are needed for scenarios such as emergency and disaster recovery, field work among others. One of the challenges is to design routing protocols that satisfy the consequent requirements. In this paper we present extensions for Bluetooth, AODV, TORA, and OSPF protocols, called Blue-AN, AODV-AN, TORAN and OSPF-AN that could be used in future network design. These new protocols add services information in their databases and also operate in their respective environments. We used OMNet++ simulator to conduct our experiments. Keywords- Routing Protocols, Ambient Networks, Services

These new protocols were designed to disseminate service information, associating service to routes in order to flood this new information through the network. Under our approach, we classify services into two classes: user services (e.g. VoIP, VoD) and support services (e.g. QoS, security). More details of each one of these are presented in next sections. We implemented and simulated TORAN and OSPF-AN protocols using the OMNet++ simulator [2], thus validating them and verifying their performance and interoperability. This article is organized as follows: section 2 overviews routing protocols used in dynamic networks, section 3 explains the ones we have considered in our work and show their adaptation to work in Ambient Networks and the specific AN requirements they attend, section 4 presents simulation results of the evaluated protocols, and finally section 5 presents some conclusions and future works. WHY CURRENT PROTOCOLS ARE UNSUITABLE TO DYNAMIC NETWORKS? Current routing protocols are typically built to work in a certain select networking environment and were not designed to cope with the challenges of highly dynamic scenarios. Although some technologies already addressed desirable dynamic routing features, they do not consider issues such as interoperability and transparence towards other technologies as well as service level information dissemination. On the one hand, when using ad-hoc routing protocols, organization and reconfiguration are transparent to the users once the protocols are designed to allow the discovery of dynamically constructed networks and to adapt to their changes. On the other hand, they do not support native mechanisms to disseminate information about services offered by the nodes, and hence extensions are necessary to allow their operation over new dynamic networks. Routing protocols in these new scenarios need to have a high level of adaptability to topological changes detecting them as fast as possible without the need for any user interaction. This functionality is extremely important to allow the adaptation to some usual processes in ANs such as networks and users composition and decomposition [3][4]. Networks are about the services they offer, hence another important feature to dynamic networks is their support for the dissemination of service level information, instead of working only with end-points addresses, as is currently their case. This is important as AN users may demand service access (not nodes) at any time and a suitable routing becomes paramount to the success of such demand. Consequently, AN routes must be associated to services offered by each network node. 2.

1. INTRODUCTION The concept of dynamic networks appears in many contexts and is also currently seen associated with future Internet design. A major discussion point for these future networks is how different technologies including wired structures, wireless communication, optical networks, embedded equipments such as MEMS (Micro-ElectroMechanical Systems as sensors) or SoC (System on Chip) and many other technologies will cooperate and exchange information to attend potential users of these new network environments. In order to enable future dynamic networks, all these networks have to cooperate to satisfy a set of features, architectures and solutions, created to offer interoperability among heterogeneous networks belonging to different operators, providers or technology domains. Solutions necessary to allow this cooperation must be defined and validated while taking into consideration important desirable features for future Internet tendencies such as automatic policies definition, devices mobility, security, selforganization, self-configuration, new routing paradigms and several others depending of the set of applied technologies. In this context, the Ambient Networks [1] project has worked on a framework for networks beyond 3G and defined the dynamic composition between networks and users. Ambient Networks (AN) aim to seamlessly interconnect different technologies, obtaining a resultant network that is homogeneous for the potential users of its services. Based on the composition concept, we proposed a case study considering four different routing environments, with support for devices requirements and several network technologies: short-ranged networks, ad-hoc networks, and structured networks. We chose Bluetooth, AODV/ TORA, and OSPF, respectively, as protocols acting in each of the aforementioned environments. These new protocols received the acronym AN to differ them from their original non ANcapable versions. They are: Blue-AN, AODV-AN, TORAN, and OSPF-AN.

ISBN 978-89-5519-136-3

-212-

Feb. 17-20, 2008 ICACT 2008

In addition, protocol messages must be simple and extensible in order to not overload the network with control information traffic and to be able to accommodate future routing mechanisms. Besides, simple messages help to preserve the backward compatibility, since the new protocols probably will be applied in legacy networks with current routing protocols. Because of this, our protocols use, whenever possible, almost the same messages as traditional existent protocols, except for the addition of optional information that is considered by the new protocols and ignored by the old ones. Considering all of these features, the design of routing protocols to work with Ambient and Dynamic Networks (ANs) represents a very challenging task. Routing information must be disseminated among different technologies, allowing their nodes to compute routes that could take into consideration path lengths, QoS constraints, security, policies, inter-domain routing and so forth. All this while also taking into account the services offered and requested by each user. Since ANs can comprise heterogeneous devices with different capabilities, leading to many different technologies, the routing requirements for ANs are more complex to guarantee compared to the ones for current networks. In the next section, we are going to describe the protocols defined in details as well as show how they work. 3. ROUTING PROTOCOLS Routing protocols are typically classified as either Interior Gateway Protocols (IGPs) used to exchange information inside the same Autonomous Systems (ASs) (e.g., OSPF [5], IS-IS [6], RIP [7]) or as Exterior Gateway Protocol (EGP) used to exchange information between networks (e.g., BGP [8]). Both IGPs and EGPs share the common goal of exchanging routing information between routers, but differ in the scope of their actions and the features necessary to control information and the needs of each specific environment. Based on the research results, published in the form of deliverables, from the Ambient Network project [1], we concluded that the approach of a single routing protocol to be used in all environments might not be feasible, since each environment has specific requirements. Each particular environment needs different strategies and a single protocol would have to adapt to all possible environments, bringing a significant overhead to the routing process. However, interoperability between different routing protocols has a greater appeal, since each protocol could operate in a specific environment and could communicate with other protocols using a common subset of protocols, which would be independent of the native environment. Considering different technologies, we classified existing dynamic networks into 4 environments: structured networks, ad-hoc networks, service based networks and Inter-AN communication. Based on these environments we separated the protocols as intra-AN (Intra Ambient Network protocol) and inter-AN (Inter Ambient Network protocol), which are used in similar roles to IGP and EGP, respectively.

Structured networks are typically wired (but can be wireless) and remain mostly stable (fixed) during their existence. Typical examples are Ethernet LANs, network backbones and managed wireless hotspots. Although wireless structured networks are somewhat different from wired networks, they all have a high degree of stability, small user mobility, reliability, scalable growth, fixed nameto-address mapping and (mostly) fixed physical locations. Wireless networks may also have power constraints. Ad-hoc networks are wireless networks with no central management point (hotspot). Every node on the network must rely on its neighbors to be able to access services from far away nodes. Their main characteristics include a high degree of instability, moderate-to-high user mobility, power constraints, fixed host-to-node mapping, ungoverned growth and low reliability Service based networks are typically short-ranged ad-hoc wireless networks where nodes connect with each other to use a service. It is possible to create a network with many nodes and route information between them but this is not generally the case. Typical technologies that enable this kind of network are Bluetooth1 and UPnP2. Because they are also ad-hoc networks, all previous characteristics also apply to them. The main difference regarding typical Ad-Hoc networks is the more restraining power control, since most devices that participate in this type of network have a lower processing performance and a higher power cost associated to it. In order to verify how to adapt routing protocols with the objective of supporting new requirements of future dynamic networks, we developed Blue-AN, AODV-AN, TORANIntra and OSPF-AN as Intra-AN protocols, whereas TORAN-Inter was developed as Inter-AN. We chose these Intra-AN protocols to represent as much heterogeneous environments as possible and because they support extensions in its messages without interfering in current implementations. Each of them is suitable for a specific environment, thus providing the necessary information to the upper Inter-AN protocol, which is responsible for doing the coordination and interoperation among possibly heterogeneous environments. In Figure 1 we show two different environments, using AODV-AN and OSPF-AN, and TORA-AN as an inter domain protocol.

Fig.1.Routing Protocols designed for AN.

1 2

Bluetooth site: http://www.bluetooth.com UpNP site: http://www.upnp.org

ISBN 978-89-5519-136-3

-213-

Feb. 17-20, 2008 ICACT 2008

All intra and inter-AN protocols were adapted, designed and developed to cover some AN features presented in section 2. One common aspect is that all of them act in a service-oriented networking, since the routing is based on services. Each protocol is going to be described further in more details. Another requirement considered important was compatibility. In all defined environments we selected existent protocols to keep the compatibility with legacy devices. This would allow for a more natural replacement process, keeping current investment and, at the same time, adding devices with new functionalities until these new protocols replace all legacy elements. The unique exception was in the case of Inter-Domain communication. In the current Internet, BGP is used as InterDomain routing protocol but this was designed to work in structured networks that lack dynamism and are composed by high capacity routers. On the other hand future networks are supposed to be very dynamic and made by several different types of devices, creating a heterogeneous environment where not necessarily only high capacity routers are allowed to perform InterDomain communication. These new operating characteristics are not very suitable to a BGP network, since its routes advertisement and maintenance mechanisms do not support a high level of dynamism in the network, creating updates storms in BGP that may compromise its performance. Because of such shortcomings we do not consider BGP as a possible solution for the Inter-AN communication and designed instead a completely different protocol. However, we are currently studying a mechanism to turn compatible the communication between TORAN and BGP considering two main points: routing policies and networks data dissemination. These two requirements are necessary to design a framework to be used by TORAN to communicate and cooperate with BGP. a. BAN Bluetooth for Ambient Networks Bluetooth is a wireless communication technology designed to connect low power devices in small networks with at most 8 nodes, called a piconet. Bluetooth provides some useful features for AN routing such as device discovery (Bluetooth inquiry mode allows devices to be discovered) and consequently their relatively easy adaptation for handling topological changes, and robustness. However, Bluetooth does not support native mechanisms to disseminate information about services offered by the networks, as well as the handling of a larger number of nodes than what is supported by a single piconet. These features will be supported by Blue-AN, a routing protocol based on services and scatternet (union of piconets) topologies. Blue-AN acts as an intra-AN protocol suitable for shortrange wireless environments. It is designed to attend the following AN requirements: path and topological changes adaptation, specific type of composition and decomposition of short-range ANs (the composition of piconets forming

scatternets), multipath and routing based on services. This protocol provides the necessary information to an upper general protocol, such as the TORAN-Inter domain protocol. Blue-AN is composed by a Scatternet Formation Protocol and a Routing Protocol. In the former one, we eliminate the topologies requiring the nodes in the same range and the ones representing bottlenecks such as a single path to reach any given node. In the same manner as BlueAN, the Blueline algorithm [9] represents a graph topology, and is the closest solution to our routing mechanism. Both build and maintain scatternets in a distributed way, without the need for all nodes to be in the same transmission range. Blueline also defines a routing mechanism, to enable nodes to exchange data using the Bluetooth physical address as a target node. On the other hand, BlueAN performs routing based on services, and hence, services are associated to the routes and are the targets of a request, instead of using a physical address. b. AODV-AN AODV (Ad-Hoc On-Demand Distance Vector) [10] is a routing protocol for Ad-Hoc Networks. As implied by its name, it operates using an on-demand algorithm to find the available routes to existing nodes. In its current implementation, only single-path routing is supported, although enough information is gathered that can enable multi-path routing, and no context/service information is retrieved. AODV-AN (Ambient Network On-Demand Distance Vector) is based on AODV with modifications and extensions in routing tables, protocol messages and searching mechanism for routes. The routing tables were modified to allow the association of multiple paths to the same service. When there is a request for a service, the protocol checks to see if there is a route for that service and then, the packets are forwarded. Otherwise, a route request message is sent with the service identification as its target id. AODV defines an extension mechanism [11], which was used in AODV-AN to enable the announcement of support services for the AN-nodes and to enable the request of routes with specific QoS requirements. The extended messages are only readable by nodes implementing AODV-AN protocol and simply ignored and forwarded by legacy nodes that implement AODV. c. OSPF-AN OSPF-AN [12] extends OSPF protocol in order to be used like an Internal Gateway Protocol in Ambient Networks, introducing service information in its routing. For enabling OSPF-AN to work with services, we modified the Link State Advertisements (LSAs) for flooding the new information (service information) throughout the routing domain. Each router in the Autonomous System (ASs) originates one or more LSAs and the set of all LSAs forms the database of links in OSPF networks. From this new database, considering links states and services, each router builds a routing table, where service is

ISBN 978-89-5519-136-3

-214-

Feb. 17-20, 2008 ICACT 2008

the key, by calculating a shortest-path tree to reach that specific service, with the root of the tree being the calculating router itself. d. TORAN TORA (Temporally Ordered Routing Algorithm) [13] is a routing protocol suited for networks that have unstable connectivity. It operates mostly on-demand and is very robust to topology changes, because it locates them as early as possible, i.e. a link/node failure will typically trigger update messages only to the 1-hop neighbors. In addition to TORA, BGP was also considered as a possible base protocol candidate for the Inter-AN protocol. BGP is the routing protocol that connects current Internet ASs and therefore is very robust and reliable. In spite of this, BGP was not considered a good candidate because it has issues with dynamic networks [14]: if the network topology changes very often, the control message overhead in BGP rises considerably and takes too much time to stabilize. Therefore, TORA performs better because it is an ondemand protocol, whereas BGP is a proactive one. Additionally BGP operates with policies, whereas TORA does not. TORA does not need to implement a mechanism to exchange, create and maintain policies; it must only enforce the policies by consulting them and see if they allow routing through a known route. TORA operates by creating a DAG (directed acyclic graph) rooted at the destination node based on a concept of height. Therefore DAG represents the route which the packets may flow from node A to node B if node A is higher than node B. This restriction guarantees that the discovered routes are necessarily loop free. TORA-AN (Temporarily-Ordered Routing algorithm for Ambient Networks) was created in order to make TORA accomplish AN requirements. The first step involved modifying the path finding algorithm to look for services. Upon finding a service, the available support services are also retrieved and disseminated to the nodes in the network. When a request for that service is made, the algorithm will check if that service is known and if the supported QoS level satisfies the request. The route maintenance mechanism was enhanced to take into consideration that a route to a service may exist, but the necessary QoS requirement is not available. In this case, the algorithm now reacts to service failures, i.e. a user stopped offering the service, a new support service is achievable, etc. Although this increases the number of times maintenance needs to be made and therefore the number of control messages being sent, the changes can be very concentrated and together with the established routing policies, can be minimized. 4. SIMULATIONS AND RESULTS For the simulation of the OSPF-AN we used the discrete event simulation environment, OMNeT++, which is free of charge for teaching and research use. We choose two protocols in our simulations, one intra-domain (OSPF-AN) and another inter-domain (TORAN).

In the simulations we used two different network scenarios: (1) Gant Dark Fibre network with sixteen routers and (2) Gant network with thirty-four routes. In our experiments we adopted a number of different services per router. This number varied from one up to eight. a. OSPF-AN In the first simulation, we modified the amount of services in each router, and measured convergence time (Fig.2) and the number of exchanged LSAs (Fig.3) in the Gant Dark and Gant networks using OSPF-AN. In the scenarios with one service each router had one service. We used static services and verified that in both scenarios there was a slight increase of the convergence time from Gant Dark to Gant networks. On the other hand, the quantity of exchanged LSAs remained the same in both scenarios with the same number of service.

Fig.2. Convergence time using OSPF-AN

Fig.3. Total of LSAs using OSPF-AN In the second simulation (Fig.4), we compared OSPF and OSPF-AN, using the same scenarios prior, measuring the number of exchanged LSAs, total of bytes sent and convergence time. As can be seen in Figure 4, the number of exchanged LSAs is the same, when comparing OSPF and OSPF-AN in each scenario, independently of the number of routers and network complexity. In Figure 5 we show the amount of exchanged bytes in each scenario, and verify that the OSPF-AN protocol transmitted more data than traditional OSPF. The results were expected, since OSPF-AN transmits additional service information in all transmitted LSAs. The increase of sent bytes in the networks was: 21% in

ISBN 978-89-5519-136-3

-215-

Feb. 17-20, 2008 ICACT 2008

Gant Dark and 26% in Gant network.

and OSPF-AN basically uses point-to-point communication. However, this difference decreases as the network grows. This happens because TORAN has smaller messages than the ones of OSPF-AN and then, even transmitting more messages, TORAN traffic became closer to OSPF-AN traffic. TORAN presented very good results considering the convergence time, as shown in Figure 8. In smaller topologies the average time was less than 2 milliseconds and for bigger topologies, 7 milliseconds.
Geant_DarkFibre
1600

Geant

Transmitted Data (Kbytes)

Fig.4. Total of LSAs comparing OSPF and OSPF-AN

1400 1200 1000 800 600 400 200 0

Networks

Fig.7. Transmitted data using TORAN intra-domain TORAN obtained these results since there is no traffic in the startup of the network and, consequently, all bandwidth stays available for use by control messages. Besides, TORAN does not necessarily obtains the best route to a destination, trying to obtain a route that allows the communication independently of being the best or not.
Geant_DarkFibre
Convergente Time (ms)

Fig.5. Transmitted data comparing OSPF and OSPF-AN b. TORAN Intra-domain In the simulation with TORAN intra the amount of control messages (Query and Update) transmitted by all routers increases as long as the number of routers increases (Fig.6).
Geant_DarkFibre 15
Total of Control Messages x(103)

Geant

8 6 4 2 0 Networks

Geant

10 5 0 Networks

Fig. 8 Convergence time using TORAN intra-domain Another question that helped TORAN is due to its reactive route discovery process. In reactive routing protocols, it is only necessary to obtain information about the desired destination and not for all destinations in the network, which enables smaller convergence time. This represents a great advantage in Inter-AN routes establishment. c. TORAN Inter-domain After the studies considering only one Ambient Network, we started to study the interaction between different ANs to verify how the Inter-AN routing protocol can work in different situations, such as: with two ANs using the same Intra-AN routing protocol, two ANs using different Intra-

Fig.6. Total of Control Messages using TORAN intradomain This is an expected result, since OSPF-AN presented the same behavior in its results, but in the case of TORAN we can note higher values if compared to OSPF-AN. This happens because TORAN uses a flooding process to discover information more than OSPF-AN, and hence transmits and receives more messages. TORAN starts transmitting more information (Fig.7) than OSPF-AN because, as said previously, it floods its messages

ISBN 978-89-5519-136-3

-216-

Feb. 17-20, 2008 ICACT 2008

AN protocols, control access policies, and so forth. We collected four main metrics to evaluate all the applied mechanisms: convergence time, bandwidth overhead (Control messages), size of routing tables (to evaluate service aggregation), number of dropped messages (to evaluate routing policies). We defined sub-scenarios to evaluate the interaction between different Ambient Networks: using the same IntraAN routing protocol, using OSPF-AN or TORAN in both networks or using one different Intra-AN protocol in each AN. In all the cases TORAN is the unique algorithm used to work as Inter-AN protocol. In first sub-scenario we used TORAN to work as IntraAN protocol. Regarding the convergence time, the obtained results presented values in the same time scale of TORANIntra evaluation. For TORAN-Intra the convergence time was 6.3 ms and in the case of TORAN-Inter, 10.8 ms. The time for TORAN-Inter presented higher values because it has to wait for the TORAN-Intra discovery process to then generate the route to the external service. Another point that contributes to this difference is the fact that TORA increments exponentially its Query Timer (from 0.1s to 6.4s) in order to reduce networks overhead. With regard to the generated traffic, we have two pieces of information that should be considered: number of control messages and used bandwidth. TORA transmitted a total of 2500 control messages (2250 from TORAN-Intra and 250 from TORAN-Inter). In the same way of the control messages, bandwidth in Intra-AN is much more used by routing traffic than in Inter-AN. In Intra-AN environments TORAN used 261.6 Mb/s and to allow Inter-AN communication was necessary only 18.2 Mb/s. The routing table in the case of TORAN presented information only from the requested destination by the node, since TORAN works in a reactive way. 5. CONCLUSIONS AND FUTURE WORKS This paper presented four routing protocols for working in Ambient Networks: Blue-AN, AODV-AN, TORANIntra, TORAN-Inter and OSPF-AN. They represent different routing environments and were extended from the traditional Bluetooth, AODV, TORA and OSPF protocols respectively. A very important characteristic were implemented to put new functionalities in the original routing algorithms and to enable them to work better in the typical scenarios of Ambient Networks at the services information level. This and route information were disseminated not only based on addresses of the desired end-points, but also on service(s) required by the user(s). Another contribution was the simulation of two versions of the TORAN protocol, one to work in intra domain environments and another one to work as an inter domain routing protocol, together with OSPF-AN (an intra-AN protocol) to allow services dissemination.

In future works we intend to design a service aggregation mechanism to allow to tribute/store/locate routing information, trying to reduce possible problems brought by the increase of the amount of information caused by storage of services data in routing. Other challenge is the addition of access control policies for being used by this new routing. REFERENCES
[1] [2] [3] WWI-AN Ambient Networks Project WWW Server. http://www.ambient-networks.org OMNeT. OMNeT++ Discrete Event Simulation System. www. omnetpp.org, 2003. C. Kappler, et al, A Framework for self-organized Network Composition, Proc. 1st IFIP TC6 WG6.6 International Workshop on Autonomic Communication (WAC 2004), Springer LLNC Series, Berlin, Okt. 2004. Norbert Niebert, Andreas Schieder, Henrik Abramowicz, Gran Malmgren, Joachim Sachs, Uwe Horn, Christian Prehofer and Holger Karl. Ambient Networks An Architecture for Communication Networks Beyond 3G. IEEE Wireless Communications, Vol. 11, No. 2, April 2004, pp. 14-21. RFC 2328, J. Moy. OSPF Version 2. Request For Comments 2328, Internet Engineering Task Force, April 1998. R. Callon. RFC 1195 - Use of OSI IS-IS for routing in TCP/IP and dual environments, 1990. C. Hedrick. RFC 1058 - Routing Information Protocol, 1988. Y. Rekhter and T. Li. RFC 1771 - A Border Gateway Protocol 4 (BGP-4), 1995. R.-S. Chang and M.-T. Chou. "Blueline: A Distributed Bluetooth Scatternet Formation and Routing Algorithm", in Journal Of Information Science and Engineering 21, 479-494 (2005) Charles Perkins. Ad Hoc On Demand Distance Vector (AODV) routing.Internet-Draft, draft-ietf-manet-aodv-00.txt, November 1997. Work in progress. C. E. Perkins, E. M. Belding-Royer, and S. R. Das, Ad hoc ondemand distance vector (aodv) routing, Published Online, Internet Engineering Task Force, RFC Experimental 3561, July 2003. [Online]. Available: http://rfc.net/rfc3561.txt A. Silva, T. Dias, L. Oliveira, R. Gomes, I. Canana, D. Sadok and M. Johnsson, OSPF-AN: An Intra Domain Routing Protocol for Ambient Networks, IPOM 2007, LNCS 4786, pp. 188191, 2007. Vincent D. Park and M. Scott Corson. Temporally-Ordered RoutingAlgorithm (TORA) version 1: Functional specification. Internet-Draft, draft-ietf-manet-tora-spec-00.txt, November 1997. Work in progress. J. Rexford, J. Wang, Z. Xiao, and Y. Zhang, BGP routing stability of popular destinations, in Proc. Internet Measurement Workshop, November 2002.

[4]

[5] [6] [7] [8] [9]

[10]

[11]

[12]

[13]

[14]

ISBN 978-89-5519-136-3

-217-

Feb. 17-20, 2008 ICACT 2008

Anda mungkin juga menyukai