StreamOS: distributed operating system for correlated multi-modal streaming applications

Rivas Toledano, Raoul

Permalink

https://hdl.handle.net/2142/46635 Copy

Description

Title

StreamOS: distributed operating system for correlated multi-modal streaming applications

Author(s)

Rivas Toledano, Raoul

Issue Date

2014-01-16T17:56:57Z

Director of Research (if dissertation) or Advisor (if thesis)

Nahrstedt, Klara

Doctoral Committee Chair(s)

Nahrstedt, Klara

Committee Member(s)

• Campbell, Roy H.
• Gupta, Indranil
• Xu, Dongyan

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Teleimersion
• Real-Time Systems
• Operating Systems
• Multimedia Systems
• Correlated Streaming
• Activity Detection

Abstract

3D Teleimmersive Systems (3D TI System) are geographically distributed systems that enable remote collaborative activities. 3D TI Systems are composed of cameras, sensors and microphones to capture the video, audio and sensory data from each site at different geographical locations. These data streams are then disseminated to the local and remote sites. At each remote site the audio, video and sensory information is aggregated with the information from the local site and rendered at the displays, haptic devices and speakers to recreate a fusion of the remote and local spaces where local and remote participants can interact and participate in collaborative activities. Resource and device management in 3D TI Systems poses several challenges: Large-Scale of Distributed I/O Devices, Time and Space Correlated Streams, Diversity of Interactive Activities and Non-Standard Heterogeneous Multimodal Interfaces. These challenges leave a significant burden to the user-space application implementing the 3D TI System. Despite this high-burden, resource and device management in 3D TI Systems is implemented as a user-space application. The main reason is that current approaches to resource and device management do not consider all the challenges in 3D TI systems. In this dissertation, we address the problem of designing Distributed Operating System services that address the device and resource management challenges in 3D TI. Our Distributed Operating System, StreamOS, is composed of 6 layered subsystems: Cyberphysical Activity Layer: Provides a model for activities and stream processing in 3D TI Systems. Kratos: An Activity Management and Detection Layer provides an activity detection system based on a Support Vector Machine for 3D TI System metadata. This layer provides an initial basis to address some of the challenges that arise due to the Diversity of Interactive Activities by providing activity information to all other layers of StreamOS. Decima: A Device I/O Management Layer addresses the device management challenges that arise from Non-Standard Heterogeneous I/O devices being plugged and unplugged due to the Diversity of Interactive Activities. It also addresses the rapid changes in hardware of these Non-Standard Heterogeneous I/O. This layer addresses the problem of providing contextual support in terms of location and identification for time and space correlated groups of interactive streams that arise from the Distributed Nature of I/O devices. Prometheus: A Streaming as a Service Layer that provides end-to-end data delivery for I/O streaming devices. This layer addresses the challenges that arise from Distributed Correlated Streaming I/O devices by providing streaming of Bundle of Streams across geographically distributed TI sites. It also solves the challenge of interfacing with Non-Standard Heterogeneous I/O devices by providing a universal interface that can be accessed by a large range of multimodal devices. Finally, this layer also provides user-defined processing functions to Bundle of Streams to address the challenges in terms of activity-driven stream processing due to the Diversity of Interactive Activities and the time and space Correlated nature of Streaming I/O devices. Zeus: A Real-time Stream Scheduler Layer that provides CPU Quality of Service guarantees to groups of correlated interactive streams (i.e., Bundle or Streams). As part of this layer, we provide a Process Calculus for analysis of dependencies and concurrencies in time and spave correlated streams, and novel algorithms that provide scheduling for concurrent and codependent streams based on multi-core Earliest Deadline First (EDF) policy. This layer addresses the dependencies at the CPU Task Scheduling level. The dependencies arise from from Correlated Streaming I/O devices and from the variability in the demand of CPU resources that due to the Diversity of Interactive Activities. Hera: An Activity Based QoS Estimator that provides offline estimation of the QoS requirements in terms of bandwidth and CPU utilization of the 3D TI System. As part of our solution, we provide a QoS Model that estimates the QoS requirements of a 3D TI Session based on the activity and the type of devices used during the session. This layer addresses the contingency and variability of QoS requirements caused by the Diversity of Interactive Activities.

Graduation Semester

2013-12

Permalink

http://hdl.handle.net/2142/46635

Copyright and License Information

Copyright 2013 Raoul Rivas Toledano

Owning Collections