There is a large number of connected devices in the Internet of Things (IoT) networks that are typically several orders of magnitude bigger than enterprise networks and campus networks. The exponential increase in the number of interconnected smart devices is expected to exceed 60 billion smart objects shortly. The requirements needed for IoT networks are scalability, low power consumption, and simplified routing and security protocols. IoT networks are also heterogeneous, composing different types of networks together. Legacy network protocols like IPv4 has deemed to be inefficient for IoT networks. As the number of IPv4 addresses is almost consumed with regular network devices, we propose the use of IPv6 addressing for IoT-connected devices as IPv4 cannot accommodate the scalability requirement of IoT. IPv6 provides extended address space and enhanced mobility which are very essential for IoT networks. In this research, we apply IPv6 to IoT networks to avoid the scalability bottleneck of the IPv4 subnet. IPv6 accommodates a large number of connected devices and solves issues resulting from the heterogeneous nature and access methods of IoT devices. However, IPv6 is a large protocol that does not suit itself well in the IoT world. The Maximum Transmission Unit (MTU) permitted for IEEE 802.15.4 MAC data frames with their encapsulated IPv6 packet is limited to 127 bytes. We need 40 bytes for the uncompressed IPv6 header and 8 bytes are needed for the uncompressed UDP header. As a result, there are either 54 bytes left for the payload when security is not considered or 33 bytes when security is considered. We investigate throughput improvement for IoT networks by applying adaptation to IPv6 through header compression with UDP header compression. We also apply fragmentation for MAC frames that exceed the 127 bytes MTU limit. Simulation results showed that IPv6 compression with or without fragmentation serves toward adapting IPv6 packets to IoT networks. In the case of applying fragmentation, the technique of fragment forwarding greatly enhances the performance. Network traffic within the same network can be compressed to 2 bytes. On leaving the 802.15.4 network, the header increases to 12 bytes if the network prefix is known or to 20 bytes if the network prefix is unknown. Two evaluation metrics, namely, Compression Gain and Packet Delivery Ratio were applied to our proposed implementation method to prove the validity and the efficiency of our proposal.