
http://www.paper.edu.cn
- 1 -
A Distributed and Autonomic Resource Management
Mechanism in Network Virtualization
Wenfeng Zhao, Xuesong Qiu
**
5
(Beijing University of Posts and Telecommunications)
Brief author introduction:赵文峰(1985-),男,硕士研究生,网络管理与通信软件
Correspondance author: 邱雪松,男,教授,网络管理与通信软件. E-mail: xsqiu@bupt.edu.cn
Abstract: In recent years, network virtualization has been proposed as a long-term solution to the
gradual ossification problem faced by the current Internet and a harbinger of future network. By
allowing multiple heterogeneous network architectures to cohabit on a shared physical substrate,
network virtualization provides flexibility, increased manageability, promotes diversity and promises 10
security. One major challenge on network virtualization is the resource management of substrate
networks. The exiting research are mostly focus on design algorithms for mapping virtual networks to a
substrate networks, but neglect maximizing the resource utilization dynamically in a constantly
changing substrate network environment. In this paper, we propose a distributed autonomic resource
management mechanism to optimize the substrate network resources. In this mechanism, cooperative 15
work agent running in each substrate node can manage resource in the substrate network dynamically.
The simulation shows that our mechanism brings considerable benefit in resource utilization.
Key words: Resource Management;Network Virtualization;Resource Reallocate;Distribute;
Autonomic
20
0 Introduction
Recently the concept of network virtualization has attracted significant attention in the debate
on how to model the next-generation networking paradigm that can replace the existing Internet.
[1]The Network Virtualization concept has been proposed as a promising way for diversifying the
Future Internet architecture into separate logical networking environment called Virtual Networks 25
(VN). Multiple network architectures, experiments and services can be simultaneously supported
by the VNs over a shared substrate network (SN) [2,3].These VNs can be constructed through
deploying virtual nodes in substrate node (e.g. virtual/substrate routers) and virtual links in
substrate path. The substrate resource
One of the most important challenges in network virtualization is an efficient utilization of 30
SN resources [4]. It will help to maximize the number of coexisting of VNs as well as increase the
resource utilization and revenue. Existing research include resource allocation and resource
scheduling. Resource allocating is to mapping virtual networks into a shared substrate network
[5,6,7], those approach can find the optimal VN mapping solution when a VN request comes.
However, resource allocation approaches usually shows low responsiveness to network 35
changes .This create a need for resource schedule, especially in a highly dynamic and changing
substrate network environment. A autonomic resource schedule approach has presents in [8]. But
in this approach there is no mechanism to find a best schedule solution. Besides, it only considers
the virtual node reallocating, while virtual link reallocating can also bring great benefits in
resource utilization. 40
In this paper we propose a distributed autonomic resource management mechanism to
schedule and reallocate substrate resource in a network virtualization environment. We
decentralize the substrate control and management and build a multi-agent framework, since
distribute approach shows more flexible and manageable in a dynamic environment. Then we
design a schedule mechanism that intelligent and self-organized agent running on each substrate 45
node manages substrate resource by local information and information exchanged neighbors.