QoS-Aware Service Management for Internet-Scale Distributed Applications

Jin, Jingwen

Permalink

https://hdl.handle.net/2142/11141 Copy

Description

Title

QoS-Aware Service Management for Internet-Scale Distributed Applications

Author(s)

Jin, Jingwen

Issue Date

2005-12

Keyword(s)

• distributed applications
• computer science

Abstract

Component-based software development has evolved from tightly coupled object-oriented style to loosely coupled service-oriented style in the past few years. The new service-oriented paradigm will eventually allow heterogeneous component-based systems to interoperate in large, open networks like the Internet, and allow applications to be dynamically aggregated in a plug-and-play manner. The paradigm shift incurs more complex management issues, as services (both their functional and QoS aspects) should be managed not only as individuals, but also as aggregated entities based on their logical compositional structure required by the applications. The ultimate goal of service management is to ensure that an application, seen as a composition of multiple component services, is instantiated on top of a resource-sufficient network infrastructure satisfying specific QoS requirements, and runs adaptively according to runtime resource fluctuations and robustly despite runtime resource failures. Service management must run across three stages: (1) state collection stage that deals with collection of functional and QoS states of individual services; (2) composition stage that sets up an application by properly selecting the best QoS-satisfied service instances and establishing the path; and (3) maintenance stage that takes care of resource adaptation and failure recovery at application runtime. This dissertation focuses on providing scalable service management solutions, in which scalability is interpreted in two dimensions: network size and application size. In terms of network size - when the network becomes too large such that centralized service management relying on global knowledge of the network becomes impractical, we investigate two scalable approaches: (1) a hierarchical approach based on semi-global knowledge (by means of network topology clustering and QoS state clustering), with divide-and-conquer service management, and (2) a distributed approach based on local knowledge, with hop-by-hop decision at the service composition stage, local resource adaptation, and distributed failure detection, reporting, and recovery at the maintenance stage. In terms of application size - when an application involves one-to-many group communication, we explore resource sharing, both in network bandwidth and computational resources, by proposing a special type of multicasting - service-added multicasting. Two service tree construction algorithms, Optimal-Service-Paths Tree (OSPT) and Longest-Prefix Tree (LPT), will be proposed and analyzed. Resource sharing is further maximized as we integrate traditional data multicasting into service multicasting. We call the integrated solution hybrid multicasting. Results presented in this dissertation have been mostly implemented in ns-2, and extensive simulation tests have been conducted to verify their performances.

Type of Resource

text

Permalink

http://hdl.handle.net/2142/11141

Copyright and License Information

You are granted permission for the non-commercial reproduction, distribution, display, and performance of this technical report in any format, BUT this permission is only for a period of 45 (forty-five) days from the most recent time that you verified that this technical report is still available from the University of Illinois at Urbana-Champaign Computer Science Department under terms that include this permission. All other rights are reserved by the author(s).

Owning Collections