Efficient Mechanisms for Network State Management And Data Capture in MultiNetworks
- Author(s): Raj, Ranga
- Advisor(s): Venkatasubramanian, Nalini
- et al.
Multinetwork INformation Architecture(MINA) is a reflective (self-observing and adapting) middleware approach to manage such dynamic and heterogeneous multi-networks in pervasive environments.
MINA depends upon a strong foundation for maintaining the network state at all times. It needs to have a robust platform that performs its functions from small or large topologies alike.
This thesis describes the extensions made to the database access layer of the MINA. Modifications were made to the database schema as well as the database capture layer to capture these periodic or intermittent changes of the network state.
We tested the modifications so that MINA scales to handle a large topology.
We have also enhanced the runtime component and show that an in-memory representation of the multi-network provides very predictable responses to network path queries. MINA can now use the runtime enhancements made and provide accurate predictions of the network bottlenecks and accurately predict other what-if scenarios.
In order to perform a more realistic test on the code changes, we used publicly available datasets. Node mobility information was captured at single-second intervals observed from an access point. We successfully tested the capability to handle large volumes of changes to the network.
We then generated data for larger topologies as well as large batches of changes to network state to see how the modifications handle such stresses. We measured the time taken to persist such large volumes of changes.This gives us an idea of the frequency of changes that MINA can accommodate.
We then show one possible extension of MINA. We propose a suitable deployment for emergency response in a shanty town. We assume that we instrument the entire area of the shanty town with low-cost sensors capable of capturing local parameters like air quality, smoke, temperature etc.
Additional network components like mobile nodes and fixed access points or routers perform the task of capturing any relevant sensor data and routing it to a central location (like an Incident Response Center) for further action.
We discuss an extension where the mobile nodes are assigned a path on a road network thus directing the mobility. We present efficient techniques to assign a path to these nodes so that the time taken to capture the data from sensor nodes is minimized. We formulate the task as a Dynamic Vehicle Routing Problem with Time Windows constraints which is NP-hard.
We present heuristic based algorithms to generate an efficient path for such mobile nodes.
We show how such potential extensions can be incorporated into the MINA design framework to solve real-world problems.