Final Report Summary - CONN-BRO (Connectivity Brokerage for Collaborative Optimization of Heterogeneous Wireless Networks)
The CONN-BRO project focuses on the design of novel distributed collaborative optimization algorithms for wireless networks and software architectures for their real life implementations. The CONN-BRO architecture has evolved to leverage the software defined networking (SDN) paradigm, where different networks of air interfaces coexist (Figure 1). Devices can access multiple of these interfaces, and the goal is to allocate the resources and other parameters of these networks jointly and collaboratively to satisfy the needs of several applications running on these networks.
For this purpose, we designed and implemented a distributed SDN controller. Hazelcast library is leveraged for the inter-controller communication. For the controllers to be able to change these parameters remotely, a library for the 802.11 interface is developed and it is accepted as a part of the OpenWRT project. Then the architecture is extended for the control and collaborative optimization of the parameters of the wireless interfaces.
Since multiple controllers are needed for the collaboration of multiple interfaces, effect of multi-domain control planes with partial network map sharing in flow-level network performance is investigated. It is found via extensive emulations that partitioning has gains in the flow setup time and has little effect on throughput.
Then optimization of heterogeneous wireless interfaces has been considered. We first studied an 802.11ac network coexisting with legacy 802.11n nodes. In realistic simulations, we analyzed the effect of collisions from 802.11n nodes when multi-user MIMO beamforming is performed at the 802.11ac access point, and observed benefits of a protection mechanism. We proposed a boosting mechanism for fair time allocation between two networks. We also studied the overhead-information trade-off of the channel sounding for multi-user MIMO.
Using a comprehensive, joint Markov chain model, we analyzed the feasibility of a cognitive radio (CR) network where the primary network (PN) with aggregate on-off traffic employs retransmissions and queues, and the secondary network (SN) is infinitely backlogged and performs imperfect sensing. Then we considered energy harvesting CR networks, where the only source of energy for the SN is the RF signals of the PN and there is a trade-off between spectrum availability and energy availability. It has been observed that the secondary utilization increases with increasing PN channel utilization and burstier PN traffic, contrary to regular CR networks. We then extended the analysis to a case where SN is heterogeneous, PN performs retransmissions, SN traffic is on-off, and both networks employ finite queues. We also designed optimal energy sensing algorithms for cognitive radio networks.
We built an experimental test-bed to implement and test our architectures and algorithms. The test-bed consists of desktop PCs, laptops and netbooks, USRP software defined radios and programmable 802.11 access points. There are also high performance servers available. A real-life demonstration on the CONN-BRO testbed has been performed; where an 802.11 mesh network is constructed across the OzU Engineering building and the performance of real-time video delivery under various conditions have been studied.
The CONN-BRO project has concluded with achieving all of its envisioned goals. Its outcomes are expected to have high impact in many dimensions. Two masters and one PhD student were funded substantially by the CONN-BRO project, one of whom graduated in summer 2013, and the other two are expected to graduate by the end of 2014. The project has been instrumental in establishing the investigator’s career, his integration to academia and securing him a long-term position at OzU. Activities in this project have resulted in many publications and other dissemination/outreach activities, and many more are submissions to high impact journals are planned. The outcomes of this project are also strategically important for Europe, helping to the goal of positioning Europe in a leading role in the areas of SDN and wireless communications. The results will also have socio-economical outcomes, by increasing the ease of access to digital content and the level of quality of experience, while lowering the infrastructure and equipment cost.
The project web site can be found at https://faculty.ozyegin.edu.tr/aliercan/?page_id=307
For this purpose, we designed and implemented a distributed SDN controller. Hazelcast library is leveraged for the inter-controller communication. For the controllers to be able to change these parameters remotely, a library for the 802.11 interface is developed and it is accepted as a part of the OpenWRT project. Then the architecture is extended for the control and collaborative optimization of the parameters of the wireless interfaces.
Since multiple controllers are needed for the collaboration of multiple interfaces, effect of multi-domain control planes with partial network map sharing in flow-level network performance is investigated. It is found via extensive emulations that partitioning has gains in the flow setup time and has little effect on throughput.
Then optimization of heterogeneous wireless interfaces has been considered. We first studied an 802.11ac network coexisting with legacy 802.11n nodes. In realistic simulations, we analyzed the effect of collisions from 802.11n nodes when multi-user MIMO beamforming is performed at the 802.11ac access point, and observed benefits of a protection mechanism. We proposed a boosting mechanism for fair time allocation between two networks. We also studied the overhead-information trade-off of the channel sounding for multi-user MIMO.
Using a comprehensive, joint Markov chain model, we analyzed the feasibility of a cognitive radio (CR) network where the primary network (PN) with aggregate on-off traffic employs retransmissions and queues, and the secondary network (SN) is infinitely backlogged and performs imperfect sensing. Then we considered energy harvesting CR networks, where the only source of energy for the SN is the RF signals of the PN and there is a trade-off between spectrum availability and energy availability. It has been observed that the secondary utilization increases with increasing PN channel utilization and burstier PN traffic, contrary to regular CR networks. We then extended the analysis to a case where SN is heterogeneous, PN performs retransmissions, SN traffic is on-off, and both networks employ finite queues. We also designed optimal energy sensing algorithms for cognitive radio networks.
We built an experimental test-bed to implement and test our architectures and algorithms. The test-bed consists of desktop PCs, laptops and netbooks, USRP software defined radios and programmable 802.11 access points. There are also high performance servers available. A real-life demonstration on the CONN-BRO testbed has been performed; where an 802.11 mesh network is constructed across the OzU Engineering building and the performance of real-time video delivery under various conditions have been studied.
The CONN-BRO project has concluded with achieving all of its envisioned goals. Its outcomes are expected to have high impact in many dimensions. Two masters and one PhD student were funded substantially by the CONN-BRO project, one of whom graduated in summer 2013, and the other two are expected to graduate by the end of 2014. The project has been instrumental in establishing the investigator’s career, his integration to academia and securing him a long-term position at OzU. Activities in this project have resulted in many publications and other dissemination/outreach activities, and many more are submissions to high impact journals are planned. The outcomes of this project are also strategically important for Europe, helping to the goal of positioning Europe in a leading role in the areas of SDN and wireless communications. The results will also have socio-economical outcomes, by increasing the ease of access to digital content and the level of quality of experience, while lowering the infrastructure and equipment cost.
The project web site can be found at https://faculty.ozyegin.edu.tr/aliercan/?page_id=307
