Design and Specification of a Prototype Service / Resource

This deliverable presents the current state of the design and specifications of the XtreemOS component called “Service/Resource Discovery System (SRDS)”. The SRDS is a key component of the highly available and scalable infrastructure described in the deliverable D3.2.1 (Design of an Infrastructure for Highly Available and Scalable Grid Services) under the responsibility of WP3.2. The SRDS will offer to other XtreemOS components the capability of discovering and selecting services and resources. The system is based on two main components: the Resource Selection Service (RSS) and the Application Directory Service (ADS). The specifications of the interfaces of these two components and the interactions between them (intra-SRDS interactions) are described in this document. The document presents a first specification of the interactions among the SRDS (as composed by the RSS and the ADS) and other XtreemOS components, developed by other project WPs. In particular, current specification was produced by using the requirements gathered from WP3.3 (Application Execution Management) [7] and WP3.4 (Data Management) [8], and taking into account the design constraints imposed by security and VO management requirements [9, 5, 6]. In order to describe the services/resource discovery and selection process in the document we use the “machete and bistoury” metaphor, where the “machete” is the rough cutting tool used to open the path in a wild and dense bush, whereas the “bistoury” is the fine and precise cutting tool used in surgery. In our metaphor the “wild and dense bush” represents “a large-scale, multi-domain, distributed systems” (a.k.a. Grid) composed of a large number of heterogeneous resources/services. The services/resource discovery and selection process is a two-step process. The first step involves RSS that will act as a “machete” handling the first-level of resource/service selection in a Virtual Organization (VO) by leveraging an overlay network which will hold node information, and answering to multidimensional range queries based on static attributes of the resources/services, and returning to the ADS a list of node identifiers (NodeIDs). In order to meet the project requirements of large VO management, the RSS will implement an efficient and highly scalable distributed algorithm handling range queries. The second step of the discovery and selection process is performed by the ADS. The ADS will behave as a “bistoury”, handling the second-level of resource and service selection, and answering queries expressed as predicates over the dynamic attributes of the resources/services. The ADS can also create an applicationspecific “directory service”, exploiting the NodeID list received from the RSS. To provide scalability and reliability, and to support dynamically changing attributes and complex queries, state-of-the-art improvements to DHT techniques are being evaluated and will be exploited. The document describes the interaction between