Grid Classification

Posted By Jacob Bensimon in Blog

The term Grid is not clearly defined as such and is hence applied to a wide variety of different things. Commonly they are distinguished as cluster grids , science grids and business grids . Alternatively, we may categorize Computational Grids or Data Grids according to their focus on provision of computing cycles or data solutions. They also occasionally are grouped according to the user community they address, such as Health Grids, Earth Observation Grids, Finance Grids.

Common to all these approaches is that they have a high overlap and a very blurred boundary. For the anticipated classification of a wide range of different solutions as needed for BEinGRID a new approach is required. This holds true in particular, when the categorization not only aims at grouping the experiments, but also at providing an assessment about which reference components are required for which type of Grids.

The basic assumption was that it is impossible to find one particular term or group to categorize the type of Grid. Similar properties for Grids can be found in different existing categorizations such as eScience and Business but also in different application domains from Health, Finance or Engineering.

Six major domains have been chosen organised into categories. In all domains the categories had been chosen to allow a clear allocation of an experiment to one of them. The drawback of not allowing a continuum might result in choosing a best fit category. Some domains do intentionally allow one experiment to be placed in multiple categories as for example end user might be integrated in several ways in the frame of one single experiment. The chosen domains are as follows:

  1. Conglomeration based: The term conglomeration refers here to the resources or services provided by the participant in a virtual organisation. While in some cases the conglomeration is simply data or compute resources, in other cases the provided service is a complex conglomeration realised by using a large number of local or remote resources that are actively managed.
  2. End­ User integration based: In this domain a Grid solution might apply to more than one category as different end users might access resources in different ways: An experiment developer might use a generic interface to design a set of standard workflows and parameter studies. A scientist might then use the pre defined workflows or experiments using a portal or application integration solution.
  3. Collaboration type based: This classification looks at the nature of the collaboration of different Grid nodes. As the dynamicity of the changes in partnership have a high impact on the nature of the underlying middleware and the complexity of many aspects, it is expected that this category will be able to act as a clear differentiator between technologies. Solutions specifically designed for long term collaboration Grids will not be able to meet the requirements for ad hoc collaboration Grids and vice versa.
  4. Hierarchy based: The relationship between nodes in a Grid depends directly upon how interaction between these nodes takes place and how responsibility for service provisioning is distributed. This may be regarded as the hierarchy of a Grid infrastructure, where higher nodes contain more structure related information than lower nodes, comparable to the hierarchy of a tree diagram.
  5. Grid Management based: Depending on the goals pursued by the infrastructure, most Grids will require some kind of management of the participating nodes. Management as understood here involves (1) maintaining membership information, (2) maintaining, issuing and updating configuration information, (3) managing relationships and (4) coordinating execution. It may involve additional issues, such as related to legal aspects, encapsulation etc., which will not be detailed here as they may vary from case to case and do not have a direct impact on the management type.
  6. Application Domain: Besides all the other categories listed above, a Grid solution may be allocated in a certain application domain or community such as Medicine or Engineering. For the purpose of the BEinGRID project it is proposed to use the domains identified for each of the Business Experiments as a descriptive category.