ADMIRE D 2 . 9 – Final report on the ADMIRE architecture , with an assessment and proposals for its development

language patterns These are required to allow the use, and re-use, of common and complex data integration and mining patterns, such as all-meets-all, k-fold crossvalidation and decision-tree building. They allow us to evaluate how well the architecture can handle the abstractions introduced by DISPEL, and in turn how well the separation of concerns is facilitated by the architecture. A Architecture Evaluation Critieria 41 ADMIRE – Final report on the ADMIRE architecture Public Use by domain experts This will be primarily facilitated and demonstrated by the creation of domain-specific portals that provide a quick and easy interface for domain experts to effect their common data integration and mining tasks. Steps in workflow This is the number of processing elements in the longest path from source data to output results in a DISPEL graph, with a number greater than 20 indicating that the capacity of the DIPSEL enactment system is being challenged. A Architecture Evaluation Critieria 42 ADMIRE – Final report on the ADMIRE architecture Public B Completed Use Case Questionnaires On the following page is the front page of the questionnaire sent to use case owners. An appendix to the questionnaire contained a description of the evaluation criteria defined in D2.7, and Table 1 of D2.7 that mapped criteria to use cases. In the following completed questionnaires the front page and the appendix have been removed. The only other changes made to the questionnaires below, are the removal of unnecessary white space, the removal of page numbers, and the occasional inclusion of a use case name when this was omitted by the person filling in the form. B Completed Use Case Questionnaires 43 ADMIRE Prototype Architecture Evaluation Questionnaire for Use Case Owners D2.7 stated: “The main goals of the ADMIRE Architecture are to provide a distributed system for data mining and integration that • is scalable in terms of data volume, number of dimensions and number of sources; and processes running concurrently, on heterogeneous data sources, distributed sites and in real-time; • is easy to use in terms of developing and submitting DMI requests that are sophisticated in terms of the data flow graph; and facilitate development of sophisticated abstract patterns; and enable use by domain experts; and contain a large number of steps in the data flow graph” This questionnaire is therefore mainly based around the ʻscalabilityʼ and ʻease of useʼ evaluation criteria defined in D2.7, with a few additional questions to tease out more detail relating to user experience of working with the architecture. Questions below related to criteria defined in D2.7 are preceded by an ʻ*ʼ. A description of these criteria is given in pp 4-6 of D2.7, but is also included at the end of this questionnaire for your convenience. Note that Table 1 pg 7 (copied from D2.7) has estimates for these criteria for each use case (except SVP). You should input in the questionnaire below the actual figures of the implemented use case (in Actual column), and the original estimated figure as in Table 1 (in Planned). You may use the Notes column to explain any discrepancies – these may well be related to the architecture itself and in which case we need to know about it as it could well feed into ʻfuture workʼ. Or, you can use the Notes column to provide any other details you think relevant to the evaluation of the architecture. Sometimes a ʻquestionʼ is really a set of questions – you do not have to answer each and every question in a set – they are there to give you as clear an idea as possible as to the sort of information that we are looking for. Feel free to add any other details you think are relevant. ADMIRE – Final report on the ADMIRE architecture Public B Completed Use Case Questionnaires 44