OGF 21: October 15-19 2007, Seattle, WA Attendees: A Stephen McGough, Vesselin Novov Main groups of focus: RUS, GSA-RG, JSDL, Joint Session on Information Modelling for Computing Resources, RSS, GridNet2 Workshop, OGSA Workflow, OGSA f2f, HPC Profile, SAGA, OGF tutorial on Vulnerability Assessment and Secure Coding Practices for Middleware The RUS group as outlined previously is re-working its specification that was published previously based on implementation experiences. These include dealing with large records and the ability to request sub-parts of the whole Usage Record sets along with the rendering of RUS both in WS-I and WS-RF. As we have previously implemented the old specification for RUS we are keen to feed-back our experiences on this work. The Grid Scheduling Architecture group are focusing on Scheduler interoperation. In their architecture they see BES instances as being the end points for execution and JSDL as the language to describe the jobs sent to these resources. For this work they seek to profile JSDL in terms of which attributes within the document are relevant for the scheduler and how they should be interpreted. This again shows further groups within the wider Grid community adopting JSDL and BES. The work of GSA is not to use the entire of JSDL but to partner it with a scheduling language for those things relevant to scheduling but not to JSDL. We were able to take an active role within this session as both a group that was developing scheduling above BES/JSDL through GridSAM but also as part of the JSDL group with our knowledge of what was currently in JSDL and what would likely be there in the future. For JSDL at this OGF we held three sessions. The first session was a general session for outlining our work and current status. We iterated over the JSDL 1.0 errata document for which we have now resolved 25 of the 32 issues – extra issues have been raised since the last F2F meeting. We are also now collecting experience reports from those groups that have implemented JSDL. At present we have four responses with more promised. We also had a presentation from GridWay for their use of JSDL thus we now have twelve known implementations of JSDL. The second session was focused on the XQuery extension for JSDL. This focused on the new idea to use capabilities and requirements in a JSDL document and how you could define the requirements as XQuery requests over the (to be matched) resource description document. This session looked into how XQuery could be used to achieve this goal and weather XPath, as a simpler mechanism, could be used to achieve the same result. The conclusion arose that if all we wished to achieve was extracting parts of the other document then XPath would be sufficient. However, if we wanted to perform more complex operations on the other document, such as the processor speed is twice as important as the memory, then we would need the full XQuery. The third session was more like a BoF session. We were proposing to start some new work on an Activity schema. This document, for which JSDL would be a part, would contain the whole activity lifecycle of the job. Information on resource usage, scheduling, job status would all be placed into this document. Though the contents of these sections would not be defined by the specification, mealy refer off to the appropriate specification. Presentations were made by GridSAM, UDAP, GLUE and NEREGI as to what they have done already in this area. This session allowed us to illustrate GridSAM and place it in prime light within the community. The Joint Session on Information Modelling for Computing Resources session brought together most of the parties that had discussed information models at the last OGSA F2F meeting in a wider context. The discussion was on how GLUE relates to other specifications such as BES and JSDL. This allowed the groups to fill in the details of how we are all integrating our work together. For us the use of resource descriptions into the JSDL document from those in GLUE. The Resource Selection Service (RSS) is now submitted to the OGF editor. This specification defines an interface for how to select a set of potential resources for executing a job. As such it relies heavily on JSDL as a job description language and for terms defining what type of resources are appropriate. Again showing uptake of JSDL (and BES) thought the OGF. During this OGF we had organised a GridNet2 Workshop in collaboration with Omer Rana. This has allowed all participants funded through GridNet2 to highlight the work they have achieved. For this we were able to present work on GridSAM, GRIDCC and the development of the ICENI II architecture as areas which have both influenced the development of the OGF standards, but also benefited from these emerging standards. As much of the scope for the OGSA Workflow group has changed since the last OGF meeting we presented an overview of its progression and metamorphism into its new form. Out now is the plan to develop our own workflow language or to add simple workflow constructs onto JSDL. Instead we are now focusing on surveying what is currently being used within the wider community. As such initial results of our survey were presented. This lead to discussion on what else we should be asking and how to solicit responses from other groups. It was also discovered at this stage that the WFM group are currently performing a similar survey and it was decided to combine the effort. At the end of the OGF meeting a one day OGSA f2f meeting was held. An EMS session was held to discuss the revising of the EMS scenarios document. The plan is now to swap out BES for HPCP where appropriate in the document and downplay CDDLM as it is now “resting”. Other work such as RSS needs to be reviewed. It was decided that as there was not yet enough community support for the Service Level Terms work and as there was not enough consensus for what to do in this area it was not a prime area for standardisation. As such the work will be left for now until we have more knowledge on what can be done and what people want to do. In the area of workflow it was decided that rather than duplicating the work of the WFM group’s survey OGSA would collaborate with them on their survey and once completed determine how best for OGSA to progress with workflow. The Data Movement Interface is an interesting piece of work which helps abstract the user away from how data is actually transported around the Grid. Rather than specifying protocols and resources the user specifies known endpoints one for where the data is currently and one for where you wish the data to end up. The DMI is then responsible for identifying the most appropriate method for transferring the data. This has major implications for both JSDL and GridSAM. DMI could be used to describe the data staging endpoints and GridSAM could use DMI to source and send the files. This is highly relevant work. Much of the work in the HPC Profile WG session focused on the upcoming Interoperability fest scheduled to take place at the SuperComputing convention in November'07. In particular, much of the discussions covered the profile extensions which were to be demonstrated as the new additions to the specification. The Data Staging extensions was based in large part on the existing Data Staging element in the JSDL standard but the extension document restricted the set of options otherwise acceptable under the original spec. The text of the extension explicitly listed the type of data transfer protocols, which support would be expected from compliant implementations - http, ftp(both GridSAM supported) and scp. The Activity Credential element would provide a security context to the processing of a submitted computational job. The inclusion of this element was to account for the missing but otherwise necessary security left out of the scope of the JSDL standard. At the session we agreed on the general framework, sequence and logistical details of the interoperability demonstration which in the most part would replicate the successful interop fest at the SuperComputing in the previous year. The functionality provided by comparatively more mature GridSAM already offered staging of input/output files as well as a particular service configuration option allowing the user to supply security credentials necessary for data transfers and/or the launching of the executables. Simple API for Grid Applications (SAGA) a C++ an Java implementation of the GWDR.90 standard specification. The efforts of the SAGA group are focused on the development of a high level framework for use in creating grid aware applications. The interest in the activities of this group come from the fact that this API presents to the user as a simplified way to interact with grid-middleware products, with one such being GridSAM. In architecting and implementing the recommendations in GWE-R.90, the group addresses the inherit difficulties and issues of heterogeneity and complexity of grid applications. Therefore the SAGA is envisioned to provide common grid functionality at the correct level of abstraction along with the ability to hide varying underlying semantics. SAGA can make important contribution to interoperability at the Application level (an abstract layer above the one where such middleware as GridSAM functions), especially given that the SAGA software environment has matured sufficiently, thanks to advances at several levels; the SAGA engines (C++ and Java); the adaptors for different middleware distributions(Condor, GridSAM); and the maturity of the core SAGA specification. During the group session the discussion revolved much around the current status of the two programme language bindings, the road-map, deliverables as well as the specific ways SAGA contributes to interoperability efforts among grid users, application and middleware developers and resource provides. The OGF tutorial on Vulnerability Assessment and Secure Coding Practices for Middleware gave us valuable insight into the low-level requirements for a secure software development. Secure programming would be essential for our OMII Authorization application. The tutorial was relevant to anyone attempting to assess software for security flaws and for developers trying to minimize security flaws in software they write. Part of the tutorial covered a process for actively discovering vulnerabilities – the gathering of information about a targeted system, the use of the collected information to direct the search for vulnerabilities and the integration of these two steps into the development cycle. The tutorial pointed out specific coding practices for prevention of vulnerabilities, exposed many types of vulnerabilities with examples of how they would commonly arise, and showed some simple techniques for avoiding them. At the OMII-UK workshop we were presented of the general, overall direction of the organization's current activities. OMII-UK's mission had been to provide the UK eScience community and their international collaborators with the software solutions and support they needed to enable a sustained future. During the session we were given demonstrations on some of the ways to help research communities benefit from the increasing number of resource providers - from national grid infrastructures, such as the NGS, to campus grids and departmental clusters. The demonstrations also included examples of using stable APIs (e.g. SAGA) and showing how common scenarios can be solved by integrating components and providing new access mechanisms such as portals, virtual research environments, and desktop grids. We also discussed how projects could build on top of efforts such as GIN and OMII-Europe to bring together Grids through open standards and interoperable implementations. The future of the OMII-UK and new developments in the organization's efforts are of a particular importance to our team in London e-Science Centre. We have had a very productive relationship with OMII-UK resulting in a number of projects, some of them that have already proved successful, in delivering grid middleware application to the wider scientific community: GridSAM, GridRUS, OMII-AuthZ and GridBS.