Atomic topical segments detection for instructional videos
Ying Li, Youngja Park, et al.
MM 2006
The open, dynamic and ever-changing natures of Internet require the just-in-time scalability of middleware for Internetware. As the middleware for Internetware becomes much more powerful and complicated than ever, it is challenging to auto-scale middleware dynamically and efficiently under the Internetware environment. In this paper, a profile-based approach is proposed to automate the dynamic scaling of Internetware middleware. Profiles formalize (a) states that are keys to consistent middleware being scaled;(b) reflected information required to trigger and fulfill dynamic scaling; and (c) the best-practice of scaling domain-specific middleware. The auto-scaling tool based on this approach is capable of provisioning a scalable middleware and scale it dynamically according to the needs of Internetware. The provisioning and dynamic scaling of a scalable Mongrel cluster exemplifies the feasibility and efficiency of this approach.
Ying Li, Youngja Park, et al.
MM 2006
Ying Li, Shrikanth Narayanan, et al.
IEEE TCSVT
Jing Luo, Ying Li, et al.
ICWS 2008
Geetika T. Lakshmanan, Ying Li, et al.
DEBS 2010