Meeting 31.05.2015

From RDF Stream Processing Community Group

Logistics

Location

Attendees (Please list your name here)

  • Jean-Paul Calbimonte (JPC)
  • Adrian Paschke (AP)
  • Alessandra Mileo (AM)
  • Alejandro Llaves (AL)
  • Minh Dao-Tran (MD)
  • Robin Keskisärkkä (RK)
  • David Monks (DM)
  • Javier Fernandenz (JF)
  • Josiane Xavier Parreira (JP)
  • Peter Wetz (PW)
  • Maxim Kolchin (MK)

Agenda 31.05.

Minutes of breakout groups

Query language and semantics

Group members

JPC, DM, MD, PW, AG, Antoine, Jürgen

Material

Query Model

Minutes

AG: We have eliminated the definition of timestamped triples. We also managed to get through 70% of the RSP-QL data model page [1]. We agreed on using xsd:datetime when implementing the timestamps. We then discussed, what a timestamped graph should be exactly. tg is defined as a triple (g,p,t), g = graph, t = timestamp, p = predicate connecting g and t (cann be either systime, or apptime, maybe capture also an interval (starttime, endtime)). Multiple graphs can have same t. Same graph can have different p (for intervals). We also identified some todos (e.g. identify a vocab for p). We also defined a stream, but it's still under discussion whether it it a sequence or a set. We introduce "bounded substreams" (= extract of a window with upper bound, lower bound). We did not call it "window (data)" to not confuse it with the "window (operator)"


ACTION: AG to translate flipchart content to sentences/formulations on the github data model Done

  • Picture of flip chart summarising discussion of bounded substreams

    Picture of flip chart summarising discussion of bounded substreams


AP: Can the data model reference subgraphs when streaming? Are subgraphs streamed?

AG: I would say that depends on the implementation.

AP: The definition says it is a graph, but it could be a compound structure (referencing to other graphs) which are not necessarily streamed.

AG: It can be the case that there are referenced graphs which are not streamed.

AP: So should it be made explicit in the data model?

AG: I think it does not need to be made explicit, because it would make the basic data model (which we want to keep simple) more complex.

Semantic Complex Event Processing (SCEP)

Group members

AP, AM, JP, RK, Thu

Slides

Here

Minutes

AM: We want to leverage RSP-QL efforts to support "some kind of cep in rsp". Main questions are: What are key aspects that should be handled? Where is the cut-off point?

Benchmarking & Evaluation

Group members

AL, JF, MK, Ali, Marcin

Slides

Here

Minutes

AL: The idea of the group was to brainstorm metrics for benchmarking & evaluating RSP systems.

AL shows and discusses table with 3 cols: queries, streams, rsp systems (see slides) query throughput: how many queries can be handled by the system? data throughput: "messages per second"

Retrieved from "https://www.w3.org/community/rsp/wiki/index.php?title=Meeting_31.05.2015&oldid=584"