Name
Abstract | ||
|---|---|---|
The emerging active network concepts, particularly those focusing on the application level, will provide the users with the flexibility of deployment of customised services on the operator's infrastructure. This will be achieved through deployment of user customised control code in virtual execution environments residing on the network nodes such as routers and servers. In order to achieve efficient network operation, the resource consumption on both element and network levels must be managed. In an active networking environment resource requirements come in three categories: bandwidth, memory and processing. Based on the active networking architecture proposed in the IST project ANDROID, we develop resource models and metrics that take into account these three resource categories. The resource models are further used to define the necessary classes of policies in order to achieve efficient resource management. Policies are expressed in XML, and an example policy is given. Finally, we present an experiment comparing the use of alternative resource allocation policies. We also briefly discuss some security issues in the application level active networks. Application level active services are based on programs supplied by the users of the network. Those programs will run on equipment owned by the operators, enabling users to have access to custom services that will be managed by them without the operators' intervention. This will give flexibility to the user making the network even more attractive and the services more versatile. Effectively, these active network concepts are an example of grid computing (GridComp) philosophy for telecommunications. Various active network architectures have been proposed but none of them has more than a very basic management system. The IST project ANDROID (Active Network Distributed Open Infrastructure Development) is based on application level active networking (ALAN) (Fry99). In the ANDROID scenario, users supply service components as Java applications that run on specially designed execution environments on active nodes. Management responsibility is delegated to system components that make autonomous decisions in response to changes in their local operating environment. In an application level active network, resources come into three categories: network, computation and storage. Quality of service in conventional networks is provided by protocols and algorithms that manage network resources. Bandwidth management, priorities of queues in routers and traffic classification are some of the mechanisms used to offer the desired quality of service (Blak98) (Brad94) (Brun98). In an active services network, computation intensive processes offer the final services to the user community. Thus, the available network bandwidth is not the only parameter that influences the overall performance of a service. The total delay that data packets suffer in an end to end service is the sum of the network delay and the time needed to process data in active nodes. Moreover, during processing in active node data and code need to be stored in memory or disk making necessary to manage those resources as well. In this paper, we first present the active services architecture. Next, we introduce the candidate resource model for the active servers. Then, we discuss resource management policies and resource mechanisms. Moreover, we present the experimental results comparing the alternative resource allocation policies. In this context, an example XML policy used in given. In the conclusive sections, we also tackle some security issues in this environment. |
Year | Venue | Keywords |
|---|---|---|
2001 | LANOMS | xml.,resource model,application level active network,policies,resource management,management system,network delay,active network,resource allocation,xml,traffic classification,resource manager,bandwidth management,quality of service,service architecture,grid computing |
Field | DocType | Citations |
Resource management,Computer science,Element management system,Computer network,Human resource management system,Active networking,Resource allocation,Network management application,Network traffic control,Network management station,Distributed computing | Conference | 10 |
PageRank | References | Authors |
0.89 | 7 | 3 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Ioannis Liabotis | 1 | 49 | 7.12 |
| Ognjen Prnjat | 2 | 54 | 10.72 |
| Lionel Sacks | 3 | 99 | 15.71 |