Last June I had the pleasure to present a RINA paper at the 9th International Conference on Wired/Wireless Internet communications (WWIC 2011), held in Vilanova i la Geltrú and organized by the Avdanced Network Architecture Labs of the UPC. The paper, written by Eleni, myself, John Day, Lubomir Chitkushev, Ibrahim Matta, Steve Bunch, Miguel Ponce de Leon, Patsy Phelan and Xavier Hesselbach explained why the current Internet architecture cannot handle heterogeneous networks efficiently, and provided a quick overview on the added value that RINA provides on this issue.
There are two main issues that prevent TCP/IP from handling transport over heterogeneous networks efficiently:
- The first one is that TCP can only act end-to-end (host-to-host), on a global scope, therefore carrying the implicit assumptions that the mechanisms within TCP will work efficiently over all the underlying networks. Obviously, if networks have different characteristics, such as wired and wireless, TCP behaviour will not be optimum.
- The second one is the fact that the TCP and the IP layers are not independent. TCP uses data of the IP header (including the source and destination IP addresses) to calculate the TCP header checksum. Therefore, when either the source or destination IP address change, the TCP checksum fails and the connection breaks. This is what happens when a mobile node goes from an IP subnetwork to another, for exeample.
As seen in previous posts, RINA provides a better framework for dealing with heterogeneous networks:
- As each DIF’s transport protocol is configurable, each DIF’s policies can adapt to their operative conditions: a DIF that operates on top of a wireless media can be optimized for the wireless environment, a DIF operating on top of a wired environment will have different policies. This way the overall network performance can be maximized.
- When nodes change the point of attachment and move through different subnetworks the transport connection is not interrupted, therefore providing seamless mobility.
The figure above shows the different DIFs involved in a scenario where a mobile terminal does a vertical handover moving between two different wireless technologies belonging to different subnetworks. DIFs named 1,2, 3, 4 and 5 are the 0-level DIFs (directly on top of the physical media) and DIF 7 floats over DIFs 1,2,3,5 and 4 to provide end-to-end transport services. The circles within each DIF represent the IPC processes (nodes), with its address (A, B, C, …). The big, transparent squares represent the systems (physical boxes). The picture also shows the forwarding table of each IPC process in the purple DIF. The forwarding table tells through wich underlying DIF a PDU has to be forwarded in order to reach a certain destination IPC Process address. Node mobility just results in routing updates that cause the forwarding tables to be recalculated.
For a more detailed overview of the topic, please check the full paper “Transport over heterogeneous networks using the RINA architecture”, available online at the Pouzin Society website or as part as the conference proceedings published by Springer.