Architecture of Interconnection between networks

[MacKenze]

The operating mode (datagram or virtual circuits) determines the network architecture.

• Operating mode connection - when using this type of operation (usually virtual circuits) each intermediate system connecting two subnets. To pass information from a sender to a receiver, both systems provide a logic circuit through a series of intermediate systems. These intermediate systems are the same and unique for each connection of two computers connected. For users sending and receiving, it seems that the connection is point to point. To make this possible, the network layer of the transmitter, receiver and intermediate systems must provide similar functions.
• Operating mode offline - offline operation (usually in datagrams) the transmitter sends a block to each intermediate system network and repeats the block to route it to the final system. Thus, it is possible that it reaches its destination block several times and by different routes. In each routing unit decides the best way forward for each block, whether they belong to the same issuer and the same fate. For this it is necessary that all systems transmitter, receiver and intermediate has a similar network protocol (IP).
• Approach using bridges, via bridges, is the MAC layer (under Network) responsible for the retransmission of the blocks. For this, start and end systems must share the network and transport layer. Furthermore, all networks must use the same protocol in the link layer.

[Abraham.J]

Requirements for interconnection between networks:

1. Provide a link between networks.
2. Provide routing and delivery of data between processes on different networks.
3. Maintain an accountability mechanism and state of networking and referrals.
4. To provide these services without having to change the network architecture.

For this, the systems must accommodate the differences between the networks:

• Different addressing schemes.
• Different maximum block size.
• Different mechanisms for access to the network.
• Different values of expiration timers.
• error recovery.
• Status Reports.
• routing techniques.
• Access control to the user.
• connection, offline.

[garfield1]

Example, A and B are two systems belonging to two different networks connected through another network WAN. The WAN network is packet switching. Systems A and B must have the same IP network protocol and identical upper protocols (transport and application). Routing devices should only implement the network and lower layers. The IP protocol to receive data blocks and adds a header global network address (network address of station B). Thus, a datagram is constructed. This datagram is passed to the network and is received by the first routing system that reads the IP header and puts the header needed to be read by the WAN. WAN receives it and passes it to the routing system will guide you to the final station. This system removes the routing header and put the WAN IP to send the system will reach its end where IP (and IP header will be passed without its top coat). Under the IP protocol is the LLC, MAC and physical. Each of these protocols is adding its own header to be removed and start again for each of the routing systems. The final system does the same. When a routing device reads the IP header of the datagram that has to be routed and do not know where to send, it returns a datagram with the error information.

[Neil'o]

When a routing device reads the IP header of the datagram that has to be routed and do not know where to send, it returns a datagram with the error information.

Each new unit of data is queued in its bottom layer until it is their turn to be sent. If two networks connected by a routing system, it can discard datagrams on its tail so as not to damage the faster network datagrams waiting for the slowest. IP does not guarantee that the data reaches its destination and in order, is TCP which is responsible for this. IP, by failing to ensure order and delivery data, work with any type of network because the data can take multiple paths before reaching their destination. This allows you to also change routes when there is congestion or some sort of compatibility.

[Solaris]

It may be necessary packages, going from one network to another, must be cleaved by needs of such networks. You can let the system turns to the final assembly, but this means there is too much work for it and besides, there may be intermediate subnets that can work with larger blocks than those supplied by the previous network, so that efficiency is lost . But the advantages of this system is that final assembly routing devices do not have to keep in memory the successive bits of the datagram and also allows dynamic routing (as successive pieces need not take the same routing). IP is in final assembly. The final system must have sufficient memory for storing the pieces go to join them when they reach everyone. As IP does not guarantee the arrival of all data, you should use a timing system (either by using a proper time since the arrival of the first portion of the datagram or by using the timing data included in the datagram header).