One of the important criteria at the introduction of the Internet was that it be able to operate without being dependent on a single node, system or institution and thus maintain and ensure its continuity. That is why the Internet was devised in a distributed structure. Blocks that harbor IP addresses were assigned to institutions and units in accordance to their needs. In this manner METU was provided with the usage rights of IP addresses between 144.122.0.0 - 144.122.255.255. The inevitable result of this distributed structure was having different IP address blocks residing at different regions. The way to operate such a network which is not structured centrally but in distributed IP blocks was for the router devices, maintaining connection of the networks to the exterior, to interact with each other and share the information of the networks they provide access to. Gateway protocols were developed to achieve this interaction. Thus, without having to inquire to a central point, any router around the world has become capable of knowing which IP block is assigned to which location.

Two of the most commonly employed gateway protocols are ospf and bg. Even though, the OSPF (Open Shortest Path First) is a protocol generally utilized by the service providers in their internal networks, it is widely used for institutional networks as well. The BGP (Border Gateway Protocol), however, is used by service providers worldwide, wide institutional networks, and specifically by huge campus networks in order to mutually interconnect.

There are ten thousands of discrete networks on the Internet. The packet transmission on the Internet for the side routers can only be possible over a common protocol in order to convey and share information. The logic behind BGP is founded upon the idea that each router being able to convey to its neighboring routers the information about which networks they can access to. With the BGP system, each side router announces its own network, that is, the network it gives access to the Internet to, to the side routers surrounding it. The surrounding side routers that obtain this information pass it on to other side routers they are connected to. As a result, each side router around the world will come to know one or more paths to reach an IP block.

The operating mechanism of BGP is similar to counting the traffic lights on different routes to reach from one point to another and choosing the route with the minimum number of traffic lights. When we think of each routing device as a traffic light, the traffic light closest to the destination informs all the other traffic lights, in the vicinity, of the number of traffic lights to be covered in order to reach the destination via itself. Assuming that this number for the first traffic light is "0", the nearby traffic light which gets this information increases this value by "1" and conveys it to the neighboring traffic lights. These will increment the value again by "1" and send it to the others as "2". In this manner this number is raised to 3... 4... 5... and conveyed to all the other traffic lights. The main point here is that information from many paths will arrive at a traffic light. Let us assume while "4" comes from one light "5" comes from another. In this case, the path to the traffic light that says "4" will be shorter than the one that says "5". The BGP operates with the same logic. It directs the packet to the right route by analyzing the information arriving from various resources that give the number of hops to reach the designated IP block.

OK, what if there is a break on the route known to be the shortest and the destination cannot be reached.
To be aware of such a situation a mechanism called "announcement" is used with the BGP. Each BGP router device sends "I am up, on duty" message to its neighbors which it is connected to. A neighboring device that does not receive this message for a certain time clears out the received information from the related router and reconsiders the routing calculations.

It is possible to adjust various settings with the BGP system. If the establishment has more than one main network connection, one of these may be set to show itself as distant and the usage of the other line may be encouraged or work load balancing between the lines may be performed, a certain line may be provided to access an IP block, filters may be scripted for certain IP blocks etc. In this way, each main router device incorporating the BGP routing protocol has the information of which IP block around the world could be accessed via which neighboring routing device the shortest or the preferred path within a table that it keeps. To simplify, by checking its own table, it conveys the packet to the neighbor that is the nearest or preferred, the neighbor repeats the same procedure checking its own table and this procedure carries on until the packet reaches its destination. The network device that transfers the package to its neighbor is never concerned with whether the packet reaches its destination or not. It has fulfilled its task.

Currently, the server incorporating BGP is a PC router device. On it, routing tables are formed by an open source code free software. METU has four "BGP neighbors". These are ULAKNET, METROETHERNET, TRNET and BİLKENT. While the BGP tables of Ulaknet ve Metroethernet are used for the Internet access of the campus, the TR-NET and Bilkent tables are used for a system called "peering" which is used for the mutual direct connections of the users of the institutes. As of September 2008, the campus main router keeps tables of the IP blocks that can be accessed from approximately 265.000 Ulaknet lines, 263.000 metroethernet lines , 40 TR-Net lines , and 1 Bilkent line and directs the traffic in accordance with the BGP.

A. Onur Yurtsever - Hüsnü Demir