S ince IP networks become larger and larger it would be impossible for them to operate without routing protocols. Routing protocols provide the appropriate paths for site to site communication. There are many Interior Gateway Protocols developed based on known algorithms for IP networks, preventing rooting loops. Some of the most known interior gateway protocols are OSPF, RIP and EIGRP. Each of these protocols has its drawbacks and benefits and the election between them is based on many parameters. In this thesis we present a survey and a performance test of the OSPF routing protocol using IT Guru OPNET simulation tool. We will examine its network convergence duration, packet delay variation, CPU utilization, voice jitter and the need of Multi Areas.
C omputer-based communication systems exchange data via links within a set of connected devices using routing protocols. Routing protocols are used for the transmission of packets over IP networks and are based on algorithms which find the best path for transmitting data over the network relied on various metrics. The most common metrics are packet delay, bandwidth, cost, maximum transmission unit and hop count. These metrics results are saved in routing tables by the protocols. Even if the routing protocol is between ASs or within an AS, there are two types of routing protocols, external gateway protocol (EGP) and interior gateway protocol (IGP). The most used Interior gateway protocols are OSPF, RIP and EIGRP which every protocol has different performance from each other.
This study examines Open shortest path first protocol challenges on convergence, routing table and the need for Multiple Areas. Open shortest path first (OSPF) is an adaptive routing protocol that distributes routes and information within Single autonomous systems (AS). The network is divided by OSPF into Areas which every Area is consisted of one or more segments. Each segment introduces the set of connected routers over a common communication channel. When a failure occurs the paths are recalculated and the topologies are regenerated by every router within that Area. This procedure needs time until the routers discover the Area’s new topology and it is known as convergence time. Two topology scenarios will be implemented with the OSPF routing protocol in order to observe the behavior. Specifically, OSPF’s Single Area and Multi Area performance will be compared both in basic and scaled topology based on four quantitative metrics which are the network duration convergence, packet delay variation, CPU utilization and voice jitter.
T he main aim of this thesis is to study the OSPF challenges on convergence, routing table and the need for Multiple Areas. In order to collect more accurate results the performance of the Open Shortest Path Protocol (OSPF) will be examined extensively based on specific quantitative metrics which are the network convergence duration, the packet delay variation, the CPU utilization and the voice jitter. In a controlled, virtualized environment which is the Riverbed Modeler 17.5 Academic Edition, also called OPNET, two topologies will be designed, created as well as evaluated based on the default characteristics of the OSPF routing protocol due to the fact that the academic edition has various restrictions. Particularly, these quantitative metrics will be measured in a basic as well as a scaled network topology.
The specific thesis will examine whether the Single Area OSPF performs better on simple topologies compared to Multi Area as well as if the Multi Area OSPF has better behavior in scaled topologies in comparison to Single Area. This hypothesis generated based on most researchers especially RFC 1247 (1991) assertions that within an Area each router preserves a topological database for the Area to which it belongs as well as it has complete knowledge of the entire network.
Thus, it is implied that the Single Area OSPF performs much better than Multi Area OSPF in a small, simple, basic topology due to the fact that the last mentioned has to conduct multiple processes simultaneously in order to flood the LSAs in other Areas through the Area Boarder Routers. However, as the topology grows with multiple entries, link state changes occur and the Single Area OSPF has to recalculate the link state database in a multiple degree compared to Multi Area OSPF which distributes router into Multiple Areas. This automatically reduces the size of the LSDB as well as the size of the routing table due to the fact that routers have no detailed information about external Areas.
Network convergence duration
The convergence of OSPF Single Area in scaled topology will be slower compared to OSPF Single Area in simple topology as well as to OSPF Multi Area in scaled topology.
The convergence of OSPF Multi Area in simple topology will be slower compared to OSPF Multi Area in scaled topology as well as to OSPF Single Area in simple topology.
Packet delay variation
Packet delay variation of OSPF Multi Area in simple topology will be slightly higher in comparison to OSPF Single Area in the same topology.
IP Voice jitter
Voice jitter of OSPF Single Area in scaled topology will be higher than OSPF Single Area in simple topology.
OSPF Single Area in simple topology will need more CPU utilization compared to OSPF Multi Area in the same topology, however, it will not require as much processing power as OSPF Single Area in scaled topology
I n order to fulfill the scope of this thesis which refers to the OSPF challenges on convergence, routing table and the need for Multiple Areas, extensive critical evaluation and literature review implemented almost completely from RFC 2328 and RFC 1247 as well as IEEE. Specifically, several books were studied, conference proceedings, journals and technical publications were examined based on the performance of OSPF in Single Area and Multi Areas. Additional research was conducted using key words relevant to network convergence duration, packet delay variation, CPU utilization and voice jitter Single Area, Multi Areas, interior routing protocols, OPNET and Riverbed Modeler Academic Edition.
Despite the fact that various literature based on the OSPF analysis exists, it must be noted the restricted literature of OSPF comparison performance between Single Area and Multi-Areas implemented on the Riverbed Modeler simulation tool is presented. Proceeding, after the examination of the related work and OSPF features, simulation scenarios were designed in order to verify from the results the hypothesis of this thesis. The simulation infrastructure was created and the results were congregated and examined in order to depict the outcome as well as to accomplish the research aim of this thesis.
I n the first part of our research OSPF which is an industry-standard open source protocol will be examined. Developed by Dijkstra, OSPF uses the Shortest-Path-First algorithm and it is supported by almost all network equipment companies because of it open source nature. Although, SPF algorithm consumes more memory and CPU sources from the router, it has great performance. Later in this thesis, we will design a simple OSPF topology to examine its behavior and after that we will expand this topology from Single Area to Multi Area to see if its behavior is improved or not (Mirzahossein et al, 2013).
T here are not many works analyzing routing protocols performance. In this literature section some of these studies are grouped and, finally, some of these routing protocols performance studies will be shown.
OSPF, as it is presented by Shaikh et al, provided by its authors a study about OSPF behavior in large network infrastructure that is based on hierarchical structure which is form by fifteen Areas and five hundred routers. The main goal of this infrastructure is to provide highly reliable and available connectivity from facilities to databases and applications residing in a data center. It is introduced a methodology for analyzing the link state advertisements (LSA) traffic that is generating when a network topology occurs. Its authors, also provide a method to predict the refreshing rate of LSAs from the configuration information of routers. Moreover, it is observed that some types of topologies can provoke duplicate LSA asymmetric traffic.
Another OSPF study by A. Basu et al, examines the stability of OSPF under steady state and with link interferences. It will be shown in this study the given effects by traffic engineering extensions focused on stability of an OSPF network infrastructure. As the OSPF has long extensibility because of its design principles, Goyal et al, focused on OSPF robustness against failures and OSPF scalability, supports in its survey that these objectives are achieved by the division of the routing domain into Multiple Areas which will limit the CPU processing overhead of OSPF protocol. These great feature would allow large OSPF infrastructures to exist even with not so powerful routing devices and avoid meltdowns even if the infrastructure faces frequent and multiple topology changes. The same statement about overhead reduction is also referred by the survey of Jun et al, which provides results for scalable MANETs using Multi Areas presenting their effects. In this survey, it is also supported by theoretical and simulation results that there is an optimum number of Areas that plays critical number in OSPF scalability. Findings of this study can be applied to other table driven protocols including OSPF extensions (MPR, MDR) and OLSR as the reduction of the overhead from using these Areas number is independent of Single Area approaches. On survey of Pereira et al, the authors have studied network’s performance with OSPF routing based on bandwidth estimation that was analyzed in terms of delay and network throughput under the OPNET simulator.
They have shown that adaptive OSPF routing can significantly increase the network throughput by 25% in simulation scenarios that traffic load is highly concentrated, proving the advantages of an adaptive OSPF routing. Although, they proved that, they also suggest further analysis by taking account simultaneously more factors such as information loss rate and delay. Another research of Zhao et al, formally analyze the convergence dynamics of OSPF in presence of multiple failures. According to authors’ analysis the convergence time is delayed mainly and largely by the protocol’s timers. From their point of view, the cause of these dynamics commonly lies about the detection of this failures are asynchronous. They state that the reduction of detection times would relief the impact, since smaller timers will probably overreact to small network changes and increase the network’s instability.
This research, also demonstrate in Multiple failures, a great convergence delay is present which suggests that network administrator has to take into account the possible failure scenarios and dependency when he configures the OSPF network to achieve faster convergence. Finally, in the journey of our survey, has been found the research of Bojewar et al, which proposes an OSPF extension to reduce packet loss in Single link failures. They proposing a protocol that differs than from OSPF as OSPF attempts to converge after a link failure. The proposed idea is, instead of relying to handle link failures on convergence, the routing devices that are connected to the failed links, compute the new routes without including the failure and add a complete new path in the packets received by them. This way, any routing device that will receive these packets can route with safety the packets to the inserted routes.
L arge scale IP networks are impossible to provide the appropriate site to site paths without routing protocols. Interior Routing Protocols (IGPs) are used within a domain for routing i.e. a network that is within the control of an organization. An autonomous system (AS) is comprised of multiple individual networks which belong to organizations, companies or schools. The routing between ASs or within the individual networks themselves is conducted with IGPs. The interior routing protocols are classified in two categories, Link state routing protocols and distance vector routing protocols (Sendra et al, 2011). Some IGPs are EIGRP, RIP and OSPF, each of them has its drawbacks and its benefits and the best selection between them to be implemented in a network, depends in many parameters such as network hardware, features, cost, scalability, few bandwidth wastage etc. Later on this paper we will present a survey and a performance test of most used IGP protocol, OSPF.
T his type of protocols calculate the distance and the direction (vector) to any network and it is knows as Bellman Ford algorithm. This algorithm collects all needed information about the distance to each network. The distance is measured by the routers located until the destination’s path or in number of hops. Routers with distance vector protocols implemented are not allowed by the algorithm to know the entire network’s topology but only the hops count to determine a destination’s best route, with a maximum number of hops used as value to consider a network unreachable. In distance vector protocols, the routing tables include information about the IP address of route’s first router for every network listed in its routing table and the cost of each route which is defined by its metric. Finally, in these type of protocols there are regular routing table updates between the routers.
L ink state routing protocols uses the Dijkstra’s routing algorithm that is also known as “Shortest Path First Algorithm”. SPF maintains a quite complex database which is called SPF DB and contains the exact network topology and full information about the remote users. A routing device that is configured with a link-state algorithm creates information about itself, the connected links on it and their status. The information is transmitted between the routers and every routing device creates a copy of it without changing it. The link-state packets (LSP), also called link-state advertisements transmit these data to every neighboring routing device. These algorithms main goal is to provide exact information to each router about the network connections, so every routing device will be able to calculate the best routes to a network (Black, 2000).
O pen-Shortest-Path-First (OSPF) is developed for the Internet Protocol networks (IP) by the Interior Gateway Group (IGP) group of IETF that was formed in 1988 and it is based on the Shortest Path First algorithm. The OSPF was created as the Routing Information Protocol (RIP) was not capable to serve large heterogeneous networks. OSPF now, is one of the most widely Interior Getaway Protocols for the reason that is a non-proprietary routing protocol and can be used by the most vendors in contrast of it biggest rival, the Enhanced Gateway Routing Protocol (EIGRP) which is Cisco proprietary (Gedler et al, 2005). OSPF is a link-state protocol that generates routing updates when changes occur in the topology. In OSPF process the device that has detected a change such as a link state, generates a link-state advertisement (LSA) that concerns the specific link and sends it to the neighboring routers using a dedicated Multicast address. The neighboring routers receive the LSAs copy, update their own link-state database (LSDB) and pass the LSAs to their neighboring routers (Wollman et al, 1995).
L Link-state protocols are based on Dijkstra’s algorithm or SPF and send updates only when incremental changes have presented since the last routing table update. In the time of this incremental update, every router sends just the portion of its routing table that indicate the status of its own links instead of the complete routing table. In link-state protocols it required from the routers to send routing updates periodically to its neighbor routers in the internetwork. Additionally, the link-state protocols have fast convergence of their routing updates over the internetwork in contrast of others such as distance vector protocols which also makes them less prone to routing loops. Nevertheless, link-state protocols require more memory and CPU resources as they are based on the concept of map distribution, which means that all routers in the internetwork have a copy of the network map that is frequently updated (Cherkassky et al, 2005). Furthermore, factors such as the size of the routing table, the adjacencies between the routers and the number of the routers in the Area affects the router’s memory and CPU usage in link-state protocols are commonly seeing in asynchronous transfer mode (ATM) networks where some routers have more than fifty adjacencies and perform poorly (Wu et al, 2003).
The main OSPF protocol operation has three consecutive stages and leads the internetwork to convergence. Those stages are, compilation of LSDB, Shortest Path First tree and routing table entries creation. OSPF router link-state advertisements (LSAs), external routes LSAs and summary LSAs are stored in LSDB that is compiled by the LSA exchanging between the synchronizing neighbor routers. By this, all routers will now have the appropriate entries in their LSDB when the autonomous system (AS) has converged. The LSDB creation, it is demanded by every router to receive a valid LSA from the other routers in the AS. This procedure is called LSA flooding and is being initialized when routers send out LSAs that contain their own configuration. When a router has receive an LSA from another router, it propagate it to its neighbor routers.
This way, a givens router LSA is flooded in the AS so that other routers will now contain that router’s LSA. This procedure appears to cause a large amount of traffic while LSAs are flooded across the AS but OSPF is very efficient in LSA information propagation. When LSDB is compiled every OSPF router in AS performs a least cost calculation using the Dijkstra Algorithm on the LSDB’s information to create a shortest path tree to the network and the other routers with themselves as the root. As this SPF tree contains cost path information by each router with their selves as root, it is different SPF tree for every router in the AS. From the time the SPF Tree is ready, then the OSPF routing table entries are created and a Single entry for every network inside the AS is produced (Moy, 2008).
O OSPF convergence procedure includes routing calculation, failure detection, LSA flooding, RIF and FIB update. Every procedure has its own timers for limiting protocol overhead. All these timers introduce great delay and are suggested to be configured, although it is advised not to be configured as it is possible to damage protocol’s stability (Basu et al, 2001).
Failure detection on OSPF protocol is done by HELLO protocol. This protocol enables a periodically hello packets exchange for establishing adjacency. The frequency is determined by the hello interval and in case that the router has not received the packet during the dead interval timer then the adjacency is declared as “down”. Next, the LSAs are flooding the network to detect the failure by the router. HelloInterval’s suggested timer is 10 seconds, thus the failure detection can take up to 40 seconds after the occurrence has happened. Making the failure detection faster, it is obvious that convergence can be accelerated significantly. Nevertheless, reducing the timers of Hello Interval may occur false alarms in case of CPU overload or link congestion and these chances are getting bigger as the timer is decreasing (Zhao et al, 2013).
Network status, changes detection New LSA generation reflecting new changes Flooding of new LSA into the OSPF network Each router performs SPF calculations after the arrival of new LSA Each router updates RIB/FIB\
O SPF metric, also called cost of an interface that participates in OSPF, is the overhead indication that is required to send packets across an interface. The cost of the interface is inverse to its bandwidth, in other words the higher the interface bandwidth the lower cost which normally gives the Fiber Distributed Data Interface (FDDI) a 1 as a metric. An example of the inversely proportional bandwidth is, a 56k serial line that has more overhead (increased cost) than a 10Mbit Ethernet line and the calculation formula is: cost= 10000 0000/bandwidth in bit per second.
The cost of an interface by default is calculated based on bandwidth, but it can be forced to change with the appropriate command. The cost in the OSPF protocol identifies the best route to a destination. The summary of the metrics for every hop on a path are used by the OSPF to compute the cost. Finally, OSPF metric is flat 16bit range that is from 1 to 65536 (Parkhurst, 2004).
I n large scale networks that may have many devices and therefore many adjacencies will probably produce heavy control packet traffic flooding across the network, already known as LSAs. OSPF uses designated routers on all Multi-access networks to alleviate the potential problems caused by the traffic. The designated router stands for receiving all routing devices LSAs instead of them to broadcast their LSAs to all their neighbors in AS. Designated router is present in every Multi-access network and its two main functions are to originate the LSAs on behalf of the network and establish adjacencies with the routers on the AS, thus it participates in the synchronization of link-state databases. Designated router’s elections takes place when OSPF is initially established. By the time the OSPF links are active the router with the highest identifier which is defined by the router id is elected as designated router and the second router that has the next highest router id is elected as backup designated router in case of designated router’s failure or connectivity loss. When backup designated router take place then assumes its role and a new election is takes place for a new backup designated router between the routing devices in the OSPF AS. The router identifier is used by the OSPF not only to elect the designated router but also to identify the router from which a packet has been originated.
At the election of designated router, router priorities are firstly evaluated, but if the routers tie then the routing device with highest router identifier which is usually the IP address will be elected as designated router. In case of the routing device has no set an identifier the IP address of the first online interface will be set as router id and this is usually its loopback address interface. Differently, the first hardware interface with an IP assigned will be used. OSPF Routers by default have a priority set to 128. Furthermore, setting up a priority to “0” then routing device is marked as ineligible to be elected as DR. Finally, setting up the priority to “1” it means that the device has the least chance to be elected as DR and the value of “255” means that the specific device will always be elected as a DR in the AS. This OSPF concept is considered a big step above other routing protocols such as RIP. While RIP suffers from the frequent and sometimes not necessary updates that will probably slow down the network, OSPF not only updates a small part of the routing table if it is need but uses the designated router to reduce the traffic even more (Halabi, 1996).
R unning OSPF in small network, the links and the routing devices number are relatively small and the best routes to the destinations are easy to be deduced. Nevertheless, describing a larger network with many links and routing devices becomes a lot complex. Calculating the all possible routes easily can turn to a time consuming and complex calculation for routing device that runs SPF.
Partitioning OSPF routing domain into smaller Areas is a method for reducing the complexity and the size of the LSDB information, also this method helps to reduce the SPF execution time but all OSPF routing devices must have identical routing table entries in their databases. Routing devices inside an Area exchanges detailed link state information; nevertheless, the transmitted information between the Areas contain only summaries of the LSDB entries excluding the originate Area’s topology details. These summarized link state advertisements from the other Area are injected directly inside the routing table without making the routing device to run SPF again.
OSPFs hierarchy is divided into a two layer Areas, the Backbone Area or Area 0 and the non-backbone Area. The Backbone Area is the physical and logical structure of the AS and Multiple Areas are attached to it. The backbone Area redistributes the routing information to all non-backbone Areas and has to be contiguous, however it does not need to be contiguous in physical form. In general, there are not any end users found inside the backbone Area and is not allowed to be split. The non-backbone Area stands to connect the end users to resources and is set up according to geographic or functional groupings. The traffic between non-backbone Areas must always pass through the backbone Area to reach each other. Finally, a non-backbone Area does not need to have a physical connection to backbone Area (Moy, 1997).
i. Area Border Router
The Area Border Routers (ABRs) are placed in Multiple Areas of OSPF which means that can be Multiple ABRs within the OSPF network and because of this ABRs have Multiple LSDB instances. ABRs own a summarized for every Area that is presented to the backbone for redistributing these routes to the other Areas.
ii. Autonomous System Boundary Routers
ASBRs connect Multiple AS and exchanges information with routers between them. ASBRs also advertise the exchanged external routing information outside its own AS. Finally, ASBRs can run simultaneously OSPF and another routing protocol such as BGP or RIP and have to dwell inside a non-stub Area.
iii. Backbone Routers
The BRs are the routers that have their physical interfaces connected to the backbone Area and not to another OSPF Area, if they did have one of their interfaces connected to another OSPF Area then they would be considered as ABRs (Abdullahi, 2014).
O SPF is designed by the IETF to be efficient, quick converging and scalable routing protocol. For the reason that IETF designed OSPF it is an open standard routing protocol and can perform on a wide variety of vendors, as a result in a large scale infrastructure an organization can use Multiple vendor routing devices running the same routing protocol which is a great advantage offering easier configuration and troubleshooting. Another one benefit of OSPF is the convergence which is based on bandwidth. Unlike other routing protocols, OSPF sends its routing updates when a change has occurred in the network infrastructure to the directly connected routers. This process dramatically decreases the bandwidth use for routing information sharing and it is more accurate. Further, route selection on OSPF is prioritized by cost which means that packets are routed in the network using the lower cost routes that are calculated based on the bandwidth formula (Black et al, 2000). Bandwidth’s formula is 108 / Bandwidth (10 to the 8th divided by the bandwidth). Bandwidth is configurable by the network administrator on every individual interface and based on bandwidth the OSPF will choose the best path for the route. Another great advantage of OSPF is the load balancing between Multiple links as it monitors these links and routes the traffic over both while it monitors saturation levels and their speed (Chuang et al, 2001).
On the other hand, OSPF has its cons also. OSPF requires more memory resources to hold the adjacencies (neighbor lists), the link state databases that contain the routers and their routes and finally the routing tables. Also OSPF requires increased CPU processing when the SPF algorithm runs and has complex configuration. Last but not least, OSPF cannot support un-equal load balancing as it picks the links with the smallest metrics to the destination in contrast of other routing protocols that would support by configuration un-equal path load balancing (Thaler et al, 2000).
O SPF link-state advertisements are divided in 11 types and for each one there is a 20-byte header that includes the link-state ID. Every routing device generates route link advertisements for the Area to which it belongs.
Type 1: The routing device link advertisements describe the router’s links state and are flooded only to a specific Area (the Area which they belong). The link-state ID is for this type of LSA is the originating router ID.
Type 2: That type LSAs are generated by the DRs for Multi-access networks to describe the attached routers to a specific Multi-access network and floods them in the Area that contains the network. Type 2 LSA ID is the DR’s interface IP address (Moy, 1991).
Type 3: These are generated by the ABR and are summaries LSAs which are flooded outside the backbone Area to other ABR’s. Type 3 LSA ID is the IP address of destination network. Furthermore, these entries are not flooded into totally stubby or not so stubby Areas.
Type 4: This LSA is generated when an ASBR is present in the Area by the ABR and describes the routes to ASBRs. Mainly, type 4 LSAs are used make ASBRs reachable from other Areas. The link id is the ASBRs router id and like type 3 LSAs are not flooded into totally stubby or not so stubby Areas (Lindem et al, 2007).
Type 5: Are generated by the ASBR and advertise redistributed routes from outside the AS. Type 5 LSAs are flooded everywhere expect the stub Areas and its id is not changed outside the Area. Furthermore, a type 4 LSA is required to locate the ASBR and by default the routes it advertises are not summarized. These kinds of routes are described in the routing table as E1 and E2 routes.
Type 6: These specialized type of LSAs is strictly used OSPF Multicast applications.
Type 7: Routing devices in not so stubby Areas (NSSA) do not receive LSAs from external ABRs but are authorized to send external routing information for redistribution. ABR routing devices after they get the routing information they turn Type 7 LSAs to type 5 and floods the rest OSPF network with them.
Type 8: Are also specialized LSAs used in OSPF internetwork and BGP.
Type 9, 10, and 11: These types are designated for future OSPF upgrades (Berger et al, 2008).
he most basic form of route summarization is the representation of multiple route entries to a smaller one. Route summarization helps on reducing routing device memory resulting improved performance. In order to be able to use effectively the route summarization a hierarchical addressing scheme must be implemented which also will have positive impact on scalability and performance of the OSPF network. A great advantage of summarizing the routing entries is not only the performance improvement cause by the decreased amount of used memory but it also reduces the spending time of looking for a route in its routing table. This fewer routing entries existence also makes the SPF algorithm to perform faster as the LSDBs are now smaller and greatly helps on troubleshooting since it is easier to isolate the network part which might have issues. Routing summarization is normally done in at the boundaries of OSPF AS and more specific on the ABR/ABRs and the best practice is to summarize in the backbone’s direction that has as result the address aggregation and the injection of them in other already summarized Areas (Shaikh et al, 2002). OSPF generates type 5 LSAs when the routes are summarized in an Area to represent them which leads to reduced SPF calculation and more optimal routing. Furthermore, the external routes summarization makes easier the redistribution and the control of what is advertised. Also, this allows the representation of contiguous networks with a Single entry in the AS. This happens with the Type 4 LSAs that OSPF generates which are usually used to replace the type 5 and 7 LSAs. Every route is individually advertised in an external LSA when other protocol routes are redistributed into OSPF. Finally, the external routes injection to OSPF is done by the ASBRs via redistribution and OSPF can be configured to advertise a Single routing entry for all routes that are redistributed
that are covered by a particular network (Aggarwal et al, 2012).
n many networks a default route is enough to reach an uplink routing device the has connectivity to the outside networks, also default routes saves routing device’s resources as there no any need to run the SPF algorithm when occurs an external network change. OSPF has different ways for generating and advertising the default routes (0.0.0.0) based on the Area type that the route is going to be injected into. The OSPF Areas that will be covered are: normal Area, stub/totally stub Areas and not-so-stubby Areas (NSSAs). Normal Areas can be either standard or backbone Areas and can accept inter-Area, intra-Area and external routes. The backbone Area is the Area that all other OSPF Areas are connected to. Stub Areas are not accepting routes that belong to external AS, although they have intra-Area and inter-Area routes. Intra-Area routes are referring to updates that are going to be passed between different Areas and the external routes to updates that are passed from a different protocol to the OSPF domain. Furthermore, routing devices in the stub Area uses an injected default route by the ABR in order to reach outside networks. Typically, stub Areas are configured in cases that some branches does not need to know all routes to other branches of offices, for this situation a default route to the central branch and get to the other branches from them could be used. This type of Area reduces the memory requirements of the lead node routing devices and so the OSPF LSDB size. Totally stub Areas allow only intra-Area and default routes within the Area to be propagated. The default route is injected into the Area by the ABR and the rest routing devices that belong to that Area use this default route for sending any traffic outside the Area. In other words, totally stub Area’s main purpose is to block LSAs type 3,4 and 5 advertisements. No-So-Stubby-Area (NSSA) provides the flexibility to import some external routes into the Area but also retains the characteristic of a stub Area. Assuming that a routing device in a stub Area has connectivity to an external AS that is not running OSPF protocol, it will become an ASBR and now the Area cannot be called stub Area. In that case if the Area is set as an NSSA the ASBR will generate external type 7 LSAs that will be flooded throughout this Area. Finally, these type 7 LSAs will be converted to type 5 LSAs at the NSSAs ABR and will be flooded out of the OSPF domain (Lammle et al, 2001). Translating type 7 LSAs to type 5 LSAs is controlled using
Area range statements by the ABR. As stub Areas, an NSSA can be identified with a setting in OSPF options field (an N bit), intending to be the assurance that all routers in NSSA agree in the Area’s type which means that if two routers do not agree in Area’s type, they will not form an adjacency.
With OPNET the user can simulate the modeled network’s behavior, collect statistics such as link utilization, response time of running applications and display graphs. Finally, OPNET can also be used to design computer networks, troubleshoot or validate a network setup or evaluate a proposed network upgrade (Sood, 2007).
IT Guru greatest strength, is the usability as it has an integrated user interface (GUI) that speeds up the creation and visualization of networks. It can easily group larger networks parts into subnets allowing visually the classification of the nodes. IT Guru also has a wide range of different simulation devices and the new devices can be included by the device creator tool, allowing a big variety in the network simulation. These devices configuration includes excessive capabilities that almost can reproduce them identical to the real devices. Another great advantage is the speed. It is known that emulation needs time, but in fact almost all emulators do not run in real time. This software is able to simulate the behavior of a network over 24 hours in minutes, always depending on the simulated infrastructure. This ability facilitates in a short period a long term testing before the user implement any changes. Furthermore, a trial and error approach is allowed which changes small things on every simulation for achieving a goal. Another feature of IT Guru is the Virtual Command Line Interface (VCLI), which gives access to the user to a shell on the simulated routers. The VCLI accepts certain CISCO commands like auto summary and synchronization which makes it very useful, although it does not include in its instruction set other important commands. Finally, OPNET IT Guru, has a sophisticated recovery function on that scenarios and projects which due to internal or external influences have been destroyed or corrupted, can often be restored. Although this function exists it is not perfect and for this reason it is always recommended to back up work regularly (Probst, 2007).
O PNET is a C and C++ built high level UI that comes with a very wide library for its functions and has a hierarchical structure that is divided in 3 domains.
In the network model configuration, physical connection and interconnection can be included and it represents the overall system such as geographical map, network and sub network to be simulated.
Node domain is the network’s domain internal infrastructure and can be workstations, switches, routers, satellite etc.
The process domain specifies the processor’s attribute model by using C++ source code that is inside the nodes. In more detail, process domain all modules which are programmable by the user and executes tasks or processes.
he protocol used in this thesis is Open Shortest-Path-First routing protocol. The proposed routing protocol is evaluated and compared based on
network converge duration, packet delay variation, CPU utilization and voice jitter in Single and Multi-Area environments. Furthermore, OSPF’s performance will be test through scaling effects with real traffic such as voice over IP and video conference in the whole network.
Two network topologies have been generated in order to observe the performance of OSPF Single and Multi-Area, based on four quantitative metrics which are, network convergence duration, packet delay variation, CPU utilization and voice jitter.
Analytically, these scenarios are shown as follow:
1) Single Area OSPF in a simple topology (OSPF Single Area 1)
2) Multi Area OSPF in a simple topology (OSPF Multi Area 1)
3) Single Area OSPF in a scaled topology (OSPF Single Area 2)
4) Multi Area OSPF in a scaled topology (OSPF Multi Area 2)
wo different network topologies will be created, where OSPF Single Area and OSPF Multi Area will be applied in four unique scenarios. The simple topology consists
of 7 routers and the scaled topology contains 11 routers. It must be noted that OSPF Multi Area is divided in three Areas which are Area 0, Area 10 and Area 20, where Area 0 is the backbone one and includes two Area border routers. The previous mentioned Areas
are attached to the Area border routers. These two different topologies created in order to evaluate the scaling effects as well as the impact of using OSPF in Single and Multi-Area. Both network topologies designed in a campus layout 20×20 with voice and
Multimedia server as well as Multimedia users. The duration that the simulation requires to run in order to give accurate results is set to five minutes and every scenario run separately providing average values.
The simple topology is created over PPP DS1 Duplex Link lines. A link failure has been estimated to occur between R2 which is the Area border router and R1 router inside the backbone Area in order to investigate the OSPF Single Area as well as
OSPF Multi Area performance in a simple topology.
The scaled topology grows the number of routers acquiring more complex structure, however, this topology likewise simple topology is generated over PPP DS1 Duplex Link lines and the link failure is achieved between R2 and R1 routers inside the backbone Area so as to evaluate OSPF Single Area and OSPF Multi Area behaviors. The link failure that occurs between the Single and the scaled topology, starts at 120 seconds and recovers at 180 seconds, reflecting the OSPF dynamic routing protocol performance based on the quantitative metrics that will be examined. Both above mentioned topologies contain: IP routers, Ethernet Servers, Ethernet Switches, PPP DS1 links, Ethernet 10BaseT links, Ethernet 100BaseT Switch LAN, Application configuration, Profile configuration and failure recovery configuration. The simple and the scaled topologies contain the Profile Definition Object and the Application Definition Object, specifically, the Profile Object Definition which is called “users”, holds Multimedia clients that generate video conference and voice over IP traffic. Moreover, the Application Definition Object is situated to receive video conference and VoIP traffic that Multimedia users originate.
As it is observed in figure 6, the OSPF Multi Area in Single network topology has higher average network convergence duration compared to the OSPF Single Area probably due to the fact that the OSPF Multi Area is divided in sub Areas and the Area border router (ABR)
require more time to operate LSAs flooding between these Areas.
The specific figure depicts the scaling effect of the OSPF Single Area in both simple as well as scaled topology. Particularly, the figure 7 shows from the results that OSPF Single Area in simple topology presents slow average convergence compared
to scaled topology, possibly due to the fact that more routing devices have entered in the topology and the LSAs flooding is more difficult to be achieved in a short time.
In general, the figure demonstrates a better behavior of OSPF Multi Area in scaled topology, probably because OSPF Single Area is not able to manage LSAs flooding in the entire network. Α possible answer for that especially after the link failure
could be that OSPF Single Area has to recalculate the link database and flood it in the whole network compared to the OSPF Multi Area which is divided in subareas, where the internal routers have only the knowledge of the LSAs that belong within the Area.
Figure 9 depicts the average network convergence duration of the OSPF Multi Area in simple topology which appears slightly higher and requires further investigation.
In this figure (10) the video conference average delay variation of the OSPF Multi Area in simple topology presents worse behavior than the OSPF Single Area, probably due to the fact that the OPSF Multi Area conducts more processes in order to
send the packets not to mention a respective occupation of the lines and the machines occurs because of the LSAs that ABRs have to flood in the Areas.
The specific Figure (12) depicts that the average CPU utilization for both OSPF Single Area and Multi Area in simple topology is more or less the same. This measurement gives this result possibly because of the small size of the network.
Scaling the network, Figure 13 presents that the average CPU usage of the OSPF Single Area is significantly greater in the expanded topology compared to simple topology, probably due to the LSAs flooding where more processes needed.
O SPF is a prominent IGP routing protocol widely used by ISP and enterprise networks. OSPF detects changes in topology such as link failures and with its loop free algorithm convergences within seconds. An OSPF network topology may be structured or divided with Areas for optimizing traffic, resource allocation and simpler administration. This thesis attempted to compare one Single topology and one scaled topology in order to evaluate the OSPF Single Area and the OSPF Multi Area performance. It must be noticed that the divergence of scaling effects in both network topologies was not as sufficient as required in order the weaknesses of Single Area to be pointed out. However, the hypothesis of this thesis came true based on the results. Specifically, the Single Area OSPF performs better on simple topologies compared to Multi Area OSPF while in scaled topologies the Multi Area OSPF seems to behave more constant in comparison to Single Area OSPF. In terms of network convergence duration, the OSPF Multi Area in Single network topology presents higher average network convergence duration compared to the OSPF Single Area. However, comparing the OSPF Single and Multi-Area in scaled topology, is observed that the OSPF Multi Area performs better. In order to indicate the scaling effects both Single Area and Multi Area were compared in simple and scaled topology. The simulation results showed that the OSPF Single Area in scaled topology appears worse performance compared to the OSPF Single Area in simple topology however the OSPF Multi Area in simple topology shows somewhat better behavior opposed to the OSPF Multi Area in scaled topology. These variations in the protocol’s performance between the Single & Multi Area both in simple and scaled topology generated mainly due to the nature of each topology, the infrastructure and the LSAs flooding.
Concerning the video conference were heavy traffic passes through the network, is observed that the video conference average delay variation of the OSPF Multi Area in simple topology presented worse behavior than the OSPF Single Area, because more processes needed in order to send the packets not to mention a respective occupation of the lines and the machines occurred to support the LSAs that ABRs have to flood within the Areas. Regarding the average voice jitter, the results indicated that the average voice jitter of the OSPF Single Area in scaled topology deteriorated compared to the OSPF Single Area in simple topology due to the fact that the resources of the network (links and routers) were occupied more. After the link failure, the CPU Utilization of router R2 was measured in order to depict the performance between OSPF Single and Multi-Area in the simple topology where more or less both of them appeared to be the same. Furthermore, the OSPF Single Area was compared in both Single and scaled topologies in order to measure the scaling effect. The results showed that the average CPU usage in the expanded topology was significantly greater because more processes needed to conduct the LSAs flooding. After the conduction of the simulation testing and collection of the OSPF dynamic routing protocol performance results, the hypothesis appears to be valid. Specifically, the literature review as well as the critical evaluation of the overall outcomes, led to the conclusion that the OSPF Single Area in simple network infrastructures performs better compared to OSPF Multi Area, however in scaling network topologies, Multi Area presents more advantageous behavior in comparison to OSPF Single Area. Future work is required in the specific issue that this thesis examined. Further experiments in larger network topologies should be achieved with the appropriate simulation tools.
F rom this thesis, I achieved to study the OSPF dynamic routing protocol in depth. Many features and characteristics were unknown to me until the conduction of this dissertation thesis. However, I faced some problems in the begging of the usage of the simulation tool due to its limitations. The simulation tool that I used was Riverbed Academic edition 17.5 which provide a workstation only for educational purposes. This prevented me from designing more complex network topologies which was the first plan of this thesis.
. Wollman, W. V., & Barsoum, Y. (1995, November). Overview of open shortest path first, version 2 (OSPF V2) routing in the tactical environment. In Military Communications Conference, 1995. MILCOM’95, Conference Record, IEEE(Vol. 3, pp. 925-930). IEEE.
. Wu, J., Dai, F., Lin, X., Cao, J., & Jia, W. (2003). An extended fault-tolerant link-state routing protocol in the internet. Computers, IEEE Transactions on,52(10), 1298-1311.
. Berger, L., Zinin, A., & Bryskin, I. (2008). The OSPF opaque LSA option.
. Moy, J. (1991). OSPF protocol analysis.
. Lindem, A., Shen, N., Aggarwal, R., Shaffer, S., & Vasseur, J. P. (2007). Extensions to OSPF for Advertising Optional Router Capabilities. Work in Progress.
. Gredler, H., & Goralski, W. (2005). Introduction, Motivation and Historical Background. The Complete IS-IS Routing Protocol, 1-9.
. Cherkassky, B. V., Goldberg, A. V., & Radzik, T. (1996). Shortest paths algorithms: Theory and experimental evaluation. Mathematical programming,73(2), 129-174.
. Moy, J. T. (2008). OSPF complete implementation. Pearson Education.
. Parkhurst, W. R. (2004). Routing first-step. Cisco Press. ISO 690
. Halabi, S. (1996). OSPF design guide. Cisco Systems Network Supported Accounts.
. Shaikh, A., Isett, C., Greenberg, A., Roughan, M., & Gottlieb, J. (2002, November). A case study of OSPF behavior in a large enterprise network. In Proceedings of the 2nd ACM SIGCOMM Workshop on Internet measurment(pp. 217-230). ACM.
. Aggarwal, A., & Khera, S. (2012). Combat Resources Shortages by making Stub Areas and Route Summarization in OSPF. International Journal of Scientific and Research Publications, 2(8).
. Lammle, T., Odom, S., & Wallace, K. (2001). CCNP: Routing Study Guide. Sybex.
. Moy, J. (1997). OSPF version 2.
. Abdullahi, R. (2014). Modeling and Simulation of an OSPF router (Doctoral dissertation).
. Black, U. D. (2000). IP routing protocols: RIP, OSPF, BGP, PNNI, and Cisco routing protocols. Prentice Hall Professional.
. Chuang, J. C. I., & Sirbu, M. A. (2001). Pricing Multicast communication: A cost-based approach. Telecommunication Systems, 17(3), 281-297.
. Thaler, D., & Hopps, C. (2000). Multipath issues in unicast and Multicast next-hop selection (pp. 1-8). RFC 2991, November.
. Basu, A., & Riecke, J. (2001). Stability issues in OSPF routing. ACM SIGCOMM Computer Communication Review, 31(4), 225-236.
. Zhao, D., Hu, X., & Wu, C. (2013). A Study on the Impact of Multiple Failures on OSPF Convergence. International Journal of Hybrid Information Technology, 6(3).
. Sood, A. (2007). Network Design by Using OPNET™ IT Guru Academic Edition Software. Rivier Academic Journal, 3(1).
. Mirzahossein, K., Nguyen, M., & Elmasry, S. (2013). FINAL PROJECT: Analysis of RIP, OSPF, and EIGRP Routing Protocols using OPNET.
A. Shaikh, C. Isett, A. Greenberg, M. Roughan, and J. Gottlieb, “A Case Study of OSPF Behavior in a Large Enterprise Network”. In Procedings of Internet Measurement Workshop, Marseille (France). November 6-8, 2002.
. Goyal, M., Soperi, M., Baccelli, E., Choudhury, G., Shaikh, A., Hosseini, H., & Trivedi, K. (2012). Improving convergence speed and scalability in OSPF: a survey. Communications Surveys & Tutorials, IEEE, 14(2), 443-463.
. Jun, J., Sichitiu, M. L., Flores, H. D., & Eidenbenz, S. J. (2007, June). The Optimum Number of OSPF Areas for MANETs. In Sensor, Mesh and Ad Hoc Communications and Networks, 2007. SECON’07. 4th Annual IEEE Communications Society Conference on (pp. 451-460). IEEE.
. Probst, A. (2007). Simulating the BGP with OPNET Guru 10.5.
. Black, U. D. (2000). IP routing protocols: RIP, OSPF, BGP, PNNI, and Cisco routing protocols. Prentice Hall Professional.
. Sendra, S., Fernández, P. A., Quilez, M. A., & Lloret, J. (2011). Study and performance of interior gateway IP routing protocols. Network Protocols and Algorithms, 2(4), 88-117.
. Pereira, T. B., & Ling, L. L. (2002, September). Network Performance Analysis of an Adaptive OSPF Routing Strategy–Effective Bandwidth Estimation. In International Telecommunications Symposium.
. Zhao, D., Hu, X., & Wu, C. (2013). On Understanding OSPF Convergence Dynamics in Presence of Multiple Failures. In ISA Proceedings, The 7th International Conference on Information Security and Assurance (pp. 48-51).
. Bojewar, S., & Dhangar, D. (2013, March). OSPF with Deterministic Routing. In International Journal of Engineering Research and Technology (Vol. 2, No. 10 (October-2013)). ESRSA Publications.