Taken by Andre Luckow of Potsdam University, Germany. Big thanks! SAGA General/Working Session Notes Tuesday, June 3rd, 2008, 11-12.30 am, Session Chairs: Shantenu Jha, Thilo Kielmann 1.) Introduction (S. Jha): Shantenu welcomed all attendees with a short overview of the agenda. 2.) SAGA in a Nutshell (T. Kielmann, Vrije Universiteit, Netherlands): Thilo Kielmann gave an overview of the motivation of SAGA, the functionality and structure of the SAGA API. The idea of SAGA is to create a simple, integrated, stable, uniform and high-level (i.e. at the right level for applications) API. The API aims to abstract the complexity of current Grid middleware platforms, which is mainly caused by the Web service environment. Based on a file copy and job submission example Thilo clearly showed the simplicity and power of the SAGA API. SAGA is structured into two set of packages: the functional and the look-and-feel packages. The functional packages provide a simple access to Grid middleware services, such as file transfer, job submission, replica managements etc. Look-and-feel packages address non-functional aspects, such as authentication, authorization, and monitoring. At the same time, the API provides an abstraction layer to different Grid middleware platforms. The C++ as well as the Java reference implementation provide an adaptor mechanisms, which is used to encapsulate middleware specific code. With this functionality applications can use the same API independent from the underlying middleware. The SAGA API enables the developer to easily integrate their application with Grid infrastructures: applications can easily spawn tasks or move files using SAGA. 3.) JSAGA: Using SAGA for Uniform Access to Heterogeneous Grid Infrastructures (Sylvain Reynaud, IN2P3, France) The main challenge addressed by this talk is the integration of various heterogeneous Grid resources from EGEE, OpenPlat, and the OpenScience Grid in different project of the CC-IN2P3. JSAGA is an open source implementation of the JAVA SAGA bindings, which was developed to provide a unified access to these resources. The main focus of the implementation is the execution management of computational jobs and file transfer. Other functional packages are currently not implemented. JSAGA provides the possibility to introduce plugins at different layers of the architectures. All plugins are orchestrated by the JSAGA engine. To meet all use cases it was necessary several extension to SAGA have been necessary. For example, JSAGA supports several job description language (in addition to the job_description class of SAGA), such as JSDL, RSL and its own proprietary language. The author concludes that SAGA is suitable for most of the need, but to support all use cases, e.g. the automatic staging of files, some extensions in form of additional methods, attributes and flags have been necessary. Another issue addressed by the presenter is the state model for SAGA jobs: JSAGA had to introduced the state queuing to solve its use case. JSAGA is successfully used in several tools at IN2P3 and is available under LGPL at the following location: http://grid.in2p3.fr/jsaga/. 4.) Implementation status, future direction, Discussion (S. Jha) The session was concluded with a short discussion about the usage and requirements of other session participants. In general, SAGA is very interesting for most participants. Several questions regarding the availability and functionality of the different Grid middleware adaptors have been addressed. Shantenu concluded the session with a short project status and timeline. TODO: - Issue with URL class in SAGA Java bindings should be resolved in current release candidate (Sylvain, Thilo) - To increase the interoperability between the Java SAGA implementations and enable the reuse of adaptors it was decided to schedule an SAGA Adaptor API standardization session at OGF24 (Thilo, Sylvain). -- Thilo Kielmann http://www.cs.vu.nl/~kielmann/