Studying On Performance Optimization And Energy Consumption Of Embedded Based NAT-PT Gateway

To address the limitations of IPv4, the IETF has developed a suite of protocols and standards known as IPv6. The transition from IPv4 to IPv6 protocol transitions will take years and stage by stage. In the early stage of transition, a lot of IPv6 islands are separated by the IPv4 oceans. In order for the IPv6 islands can connect each other, the information must be tunneled through the IPv4 Internet; In the middle and later periods of the transition, the scale of IPv6 network catches up with and exceeds IPv4 network gradually, the subject matter at this moment is how to make IPv4 network and IPv6 network can be connected, the solution that relatively admits to this problem in the world at present is NAT-PT technology. The NAT-PT system is being researched and developed on Windows, FreeBSD, and Linux, etc.The key problems that the NAT-PT gateway involves are as follows, Network Address Translation, Protocol Translation, and Application Layer Gateway. This paper launched to study on these problems. We described the limitations of IPv4 and advantages of IPv6 at first, and simply introduced the studying actuality of the NAT-PT and power dissipation model of the embedded system. On the basis of comparing the difference between IPv4 and IPv6, we described the principle of NAT-PT in detail, and analysed every key techniques which must be considered when NAT-PT being designed and implementated. We provided an environment and an implementation method of NAT-PT system, then carry on the interaction tests between the IPv4 and IPv6 networks.We also studied the NAT-PT architecture and the source code of every module in detailed, presented the functions call relation of the system, and tested the system bandwidth and latency. After using NAT-PT, the network performance is reduced greatly, the bandwidth is 5%, and the latency is 5-10 times which does not use the gateway. By using the dynamic profiling techniques, we found the systematic bottleneck is that there is a lot of searching in linear linked list. So the paper proposed various kinds of optimization schemes to the searching algorithm, and carried on overall test to executeing time and power consumptions for each optimized scheme. The finally conclusions are: In order to improve the systematic performance, as to few host's networks, such as residential network, can adopting chain addressing Hash or sorted sequential searching algorithm; And to the large-scale networks, such as campus network, can consider adopting linear probling Hash searching algorithms, of course, can adopting chain addressing Hash or sorted sequential searching algorithm too; If does not want to modify the original system extensive, then the sorted sequential algorithm is a wise choice.For similar researchs and the network equipments manufacturer, the studying approachs which proposed in the paper have general adaptability.