Since its introduction, the Grid computing paradigm has been widely adopted both in scientific and also in industrial areas. The main advantage of the Grid computing paradigm is the ability to enable, in a transparent way, the sharing and the coordination of several heterogeneous and large-scale distributed resources belonging to different institutional domains. One of its limitation is the lack of facilities for executing services. In fact, Grid computing has been traditionally used and improved for running computationalintensive or data-intensive applications. A service differs from this kind of applications in that it usually waits for requests from clients and replies with useful information; moreover, a service is typically subjected to some predefined constraints, called Service Level Agreement (SLA), including both temporal and performance restrictions. In this paper we present the TAAROA middleware, a software system that tries to extend the traditional target of the Grid computing paradigm to include the service concept. It attempts to accomplish its goal by using the virtualization technology. By abstracting the hardware and software resources of a computer, virtualization brings to TAAROA two important benefits: (1) the encapsulation of the service runtime environment, and (2) the possibility, through the migration facility, to move a service from the computer where it is running to another one that hopefully reduces the risk of violating some of the SLA constraints. In the current version of TAAROA middleware there is no explicit mechanism for achieving the level of a service as defined by the related SLA; this means that actually TAAROA is only able to provide a best-effort service.