1st TCom Meeting: 16th March 2012 Discussions and Notes

From IMarine Wiki

Jump to: navigation, search

Contents

Meeting Agenda

Meeting Participants

Ecosystem Operation

Andrea Manzi (CERN)

Slides: pptx file

FARM should be equipped with the Hadoop cluster resources.

The goal of the Runtime Resource is to promote resource discovery rather than configuration aspects.

From next release on, all the services that needs such a kind of resource should rely on the Runtime Resource facility.

gCube Portal Bundle

Massimiliano Assante (CNR)

Slides: pptx file

Re: portlets dependencies, so far all portlet dependencies are in a sort of root classpath. In the forthcoming release, dependencies should be in the Web archive (WAR), this way we obtain hot deployment.

The revised version of the Workspace Portlet has been brilliantly demonstrated with a WOW effect from the folks.

Parallel session: Array DB, WP11, Resource Model

Array DBs and WCPS Aggregator infrastructure (Parallel Session)

George Kakaletris (NKUA), P. Koltsida (NKUA)

Slides: pptx file, Rasmadan tutorial

NKUA has been involved in extending the WCPS query language. The actual requirement is to bring the WCPS query language close to the XQuery one.

The OGC approach is based on catalogues for "discovery". Unfortunately, there is limited support for "coverage" discovery. The getCapabilities is just for extensive / complete browsing of "products".

GeoNetwork implements this discovery approach by relying on the getCapabilities method and then stores the data retrieved in a Lucene instance. This is the way to discovery.

The proposed approach relies on the availability of resources representing the WCS service instances available.

There is only one implementation of WCPS (Peter Baumann and its Rasdaman DB).

  • Fabrice reported on Terradue experience with Baumann.

Currently there are few adopters of this technology, e.g. Earthlook. Thus the real goal is to publish data through this approach.

In Rasdaman, an RDBMS is used to store information that others use to store in files.

Environmental data are continuously produced and tend to 'Big Data'. Thus data can not be harvested, rather it is expected to be produced on demand.

This approach might be used for statistical analysis also.

Fabrice: one topic Terradue is interested is in relying on R for processing Geospatial data.

  • The community should be involved/informed of this.

Discussion should continue through emails.

WP11

Rena Tsantouli (NKUA)

Slides: []

In order to give priority to refactoring activities, the libraries that have most chances to be reused should have top priority.

Besides the production of the framework, it should be planned to revision of the 'clients" of this framework.

The plan should be revised to include feedback from others.


Parallel Session: Usage Tracker Service Integration

Slides: PPTX file


Presenter: Andrea Manzi (CERN)

Participants:

  • M. Assante (ISTI-CNR)
  • G. Giammatteo (ENG)
  • P. Fabriani (ENG)
  • E. Travaglino (ENG)
  • L.Bonelli (ENG)
  • C. Formisano (ENG)
  • G. Farantatos (NKUA)

Overview of the solution implemented by Venus-C for resource accounting

The software is going to be released in gCube 2.9.0

NKUA as responsible for Workflog engine is going to integrate the Usage Tracker right after the release.

The first step will be the definition of the UsageRecords which they might be adapted from the existing records published trough the System Accounting Library.

After the TCOM Eng/CERN to circulate an email to the TCOM with the specification of the service/records in order to start discussing integration also with the others developers

One big issue is related to the information related to the Service Callers , needed to track user usage of the resources, which are missing

While Security is going to provide a solution to propagate caller info, the records now will include info on the scope in order to aggregate Usage Information by scope

Parallel Session: Resource Management Portlet Evolution

Slides: PPTX File

Presenter: Andrea Manzi (CERN)

Participants: M. Assante (ISTI-CNR) G. Giammatteo (ENG) P. Fabriani (ENG) E. Travaglino (ENG)

Minutes:

The edit functionality for Runtime Resources is missing, CNR to add this feature by clicking on the selected runtime resource from RMP and redirect to RR Portlet for editing;

CERN suggests to add the copy and paste feature for Runtime and Generic Resources (both in the same scope or cross-scope);

CERN explains the missing functionalities from its point view: among the others the following is a priority: Remove From Scope for all resources;

CERN proposes to add the GUI for Data Transfer Extension (To be implemented by them);

Parallel Session: New Resource Model

Slides: pdf file (use the discussion topics slide only)

Re non-service resources, e.g. Runtime Resources:

  • is there any value resulting from publishing the resource from where it is?
    • if the management is limited to publication only, then no real added value seems to be there;
    • if it is possible to eg start/stop the resource then it should be analysed;
  • services can create them, e.g. VREs, Generic Resources, gHNs;
    • it seems that there are two approaches, either we continue with resources typologies other than SW or we add facets to existing resources;
    • if the direction is going toward one resource typology only, then it is meaningless to call this "service", it might be just a facets container;
  • some labels can be added to add a sort of type to the broad classes of resources, as to make it possible to rely on this common information to discover the resource to which start attaching a facet;
  • there are cases where the facet or part of it is expected to be used by components belonging to diverse "groups"
    • an approach might be to 'standardise' facets thus API characterising a "group";
  • the current model has some strange behavior / model with the way collection is represented, the collection profile and the potential collections;
    • should these be merged / unified;
    • it should be decided how to capture the current information with the new model;
      • the new approach retain the Resource Property Document for capturing the service status;
      • it might be possible to have two different facets, one for the "resource profile" and one for service state;
        • one of the two (the service state) is very dynamic, it depends from the service lifetime;
      • another aspect is the identification of all the facets characterizing a resource, this is expected to be dynamic;
        • this way the dynamic part modeled as facet depends from the current state of the infrastructure and is obtained via a query;
  • re on resources allocation, ie partners involvement;
    • since this activity is not blocking the rest (i) we need an 'incremental strategy' and (ii) there are few resources allocated; 'zero-dependency management" requires the new resource model;
    • NKUA is willing to contribute,

TCom wrap-up

Pasquale Pagano (CNR)

Slides: []

Re Geospatial Cluster, the proposed approach should be intergated in the context of the infrastructure, i.e. some of the pieces should be revised to 'join' the infrastructure;

  • for example, WPS-hadoop should be revised to be informed on the available Hadoop clusters by getting this information from the IS; Where the P_i are expected to be;
  • Re SPREAD, it should be defined how to integrate it with the overall (functional) architecture defined (geospatial repositories, data discovery, data access, data processing and data visualtization and related technologies);
  • A meeting is scheduled for April 6th at 10 AM;

Re Biodiversity Cluster

  • (Edward) hosting OBIS database is feasible (including images). This should be verified;
  • a cluster meeting should be scheduled; a tentative date is 5th April at 4 PM;

Re Statistical Cluster

  • the overall goal is clear / shared, i.e. the delivery of something that is called 'CodeList Manager' and 'CodeList Mapper';
  • the plan and the requirements are very fuzzy; it is needed to go further;
  • forthcoming meeting with DG-Mare is expected to improve the 'specification' phase;
  • the proposal for a December delivery should be judged / agreed only once the requirements are really specified;
  • a cluster meeting should be scheduled;

Re Semantic Cluster

  • a number of potentialities has been presented in terms of use cases. No concrete plan for their implementation has been defined / agreed. Input from the board is expected;

Y. Tzitzikas: the clusters are not disjoint. There are a number of opportunities for re-using stuffs implemented in the context of one cluster to serve needs stemming in other clusters;

  • P. Pagano: this might happen, however no concrete plan has been defined to promote this 'sharing';

Re Next TCom:

  • Location: Greece - candidates are Rodhes and Corfu
Personal tools