Enabling Interactive Video Streaming for Public Safety Monitoring through Batch Scheduling
Matin Hosseini, Mohsen Amini Salehi, Raju N. Gottumukkala · 2017
Public safety officials want to have maximum situational-awareness though real-time information, such as video content, for natural disaster management. The video content can be generated by surveillance cameras or crowd-sourced (e.g., using smart-phones) and live-streamed to the Incident Commander. Such video contents need to be processed to adapt the characteristics of the specialized multi-view display devices. When a disaster occurs, there is a surge in the number of videos streamed to the Incident Commander that oversubscribes the processing servers and the network load. Incident Commanders, however, need a smooth and uninterrupted viewing experience specifically for the important events of interest that can change over time. The challenge is how to enable the Incident Commander to interactively prioritize important video streams to receive them uninterruptedly while the system is oversubscribed. In such a system, normal video streams (i.e., non-prioritized ones) should not be interrupted at the expense of prioritization. To address this challenge, in this research, we propose a stream-priority aware resource allocation mechanism to enable interactive video prioritization without a major impact on the flow of non-prioritized video streams. The mechanism includes a method to select appropriate tasks from the arriving ones and a method to map the selected task to the appropriate video server. Our simulation results express that the percentage of normal and prioritized video streaming tasks that have completed on-time are improved, when compared with baseline scheduling methods.