SOLID Experiment by GRIDNET SOLID is an experiment that aims to demonstrate virtualized Intelligent Multi-Access User Bearer Control. The multiple access technologies that are intended to be used in the experiment are Wi-Fi and 4G, LTE-A, although in practice others could also be used. Many experimental approaches have been tried to affect the goal of Multi-Access User Bearer Control, such as MP-TCP, SIPTO, and LWA. However, all of these techniques have significant drawbacks or are a generation away as they require significant modifications. The main objective of SOLID is to leverage the technologies of SDN and NFV provided by the SoftFIRE facilities in order to build a sophisticated offloading framework for heterogeneous networks (LTE/LTE-A & Wi-Fi) that is driven by the end-user perceived QoS. The SOLID experiment aimed to provide a demonstration of virtualized Intelligent Multi-Access User Bearer Control by operating a tunneling protocol TUP between User Equipment(s) (UE) over both LTE and Wi-Fi towards a new entity called an EPC-Bridge that can make intelligent routing decisions in coordination with each UE in order to optimize usage of the dynamic performance of each available Radio Bearer (Wi-Fi and/or LTE-A). The EPC-Bridge function is virtualized and connected on the north side of the LTE PGW but before the internet. A secondary virtualized function is also deployed which is a Controller for the EPC-Bridge. All required functionalities have been built as new Network Functions (NFs) that are fully interoperable with the SoftFIRE framework and are also left to the community towards realizing new experiments by SoftFIRE users. SDN functionalities helped in implementing seamless offloading maintaining IP connectivity. All the needed mechanisms were integrated into a practical system implementation that was evaluated under specifically designed scenarios to showcase the offered advantages. The framework referred to a set of experiments with two main scenarios: Initially, small experiments performed in order to assess the functionality of the developed NFs and the offloading mechanism. Once the framework was ready for large-scale experimentation, a scenario with multiple users and SLAs was conducted. This involved a set of user flows that utilized the LTE network simultaneously each one with a guaranteed SLA by the provider. Once the LTE network capacity was not able to serve all the users without violating the SLAs, the offloading mechanism decided which user and traffic flow to offload to the Wi-Fi network. The offloading mechanism detected in time the possible SLA violations and undertook the necessary decisions to offload traffic, based on the output of the dedicated NF, which estimates the network performance of the LTE and Wi-Fi networks. A secondary experimentation scenario involved traffic flows of different applications, in order to evaluate the offloading mechanism in conjunction with the flow classification NF. The flows included voice and video traffic, web-browsing, file-transfer and more. The proposed framework served voice and video over LTE since it is more stable than Wi-Fi in terms of delay and jitter, whereas large volume traffic could be easily offloaded to the alternative Wi-Fi network when LTE is congested. The realization of the experiment in a real setup enabled the SOLID team to verify that their theoretical assumptions regarding the offloading mechanism could be technically achieved in a real LTE network. Even though the duration of the experiment was limited, they managed to extract valuable knowledge regarding the adaptations needed from their mechanisms in order to operate in a real LTE network. There are things left for further investigation and experimentation, in order to find different ways of making the offloading mechanism as seamless as possible for the client and the operator.
