Edge computing extends the cloud computing capabilities to the edge of the network to facilitate processing of the data in the close proximity of its generation. It augments the deployment of several applications in the Internet of Things (IoT) domain which demand low latency and near real-time response for their reliable operation. However, the existing approaches that help accomplish this are inherently static and suffice only for scenarios offering a fixed service at the edge. Thus, rendering them infeasible for a large scale IoT scenario with several heterogeneous services and end users such as a Smart City. Among the several challenges in realizing edge computing solutions for such a scenario, the impact of incorporating key aspects of the interactions of the IoT devices and the clients with the edge, remains to be explored. These interactions translate to the ease of use which is vital to determining the degree of adoption of an edge computing infrastructure by the end users and subsequently its profitability for the service provider. This thesis presents EDIRO, which is an edge-driven distributed orchestration framework for edge computing that enables the edge to drive the workload orchestration through the collaboration of multiple constituent edge nodes. It takes into account the existence of on-demand and ephemeral nature of workloads that require an input, i.e., an IoT resource such as an image or sensor data for their execution. These IoT resources are sourced from the end users and the IoT devices in the vicinity. Recent studies that highlight the idea of collaborative processing and the simultaneous presence of data producers and consumers in an IoT ecosystem, support this vision. The underlying concept in EDIRO is the utilization of such IoT resources that are contributed by the end users in the vicinity, to carry out service orchestration for the client requests. To the best of author’s knowledge, this is the first work in the edge computing domain to conceptualize the idea of edge-driven distributed orchestration and implement a proof of concept followed by its evaluation and practical feasibility analysis. The main contribution of this thesis is the edge computing orchestration framework EDIRO, which is developed in Golang and released as open source to encourage collaboration with the community. Experiments are conducted on three different types of computing devices emulating edge nodes in a field scenario to determine the practical feasibility of EDIRO. The measurements include the time to serve a client request, the overhead due to the distributed orchestration approach and the computing resource utilization under bursty and normal traffic scenarios. The evaluation suggests that EDIRO is feasible for practical IoT use cases and provides a reasonable trade-off in terms of the benefits offered by this edge-driven approach and the overhead incurred. This thesis shares valuable insights into the ways in which this work opens up the scope of further research in this domain along with the key findings from the system development and experimentation phase.