Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Part II: Database Design and AccessBilgenBrief Recap - Intro●Goal: Simplify FEA by utilizing DB●File-based vs. DB-based–byte stream / data subset●DB gives what FEA wants: strongly typed, self-defining semantic info + data independence + support for arbitrary read/writesBrief Recap - FEA with DB●Represent meshes with element-vertex relations and element attributes.Brief Recap - FEA with DB●Read and Write (DB)–Bulk copy commands bcp DB.dbo.T in C:\F.dat –N -T–ASCII files, Binary files, SQL tables●ETL problem–Data loading requirements (monitoring, debugging, execution..)–SSIS●Simulation Data–Inputs (partitions of meshes)–Outputs (support for fast analysis)Comutational Geometry Problems●Locating points and interpolating field values at those points●Issues –Search structures (code complexity)–Parallelism (code complexity)–Memory limitationsThe Point-in-Cell Query for Unstructured Meshes●Finding the mesh tetrahedron containing a given point●Directed Local Search method–Select a candidate tetrahedron by rank comparison on a Hilbert space-filling curve–Containment Test and Tetrahedron-connectivity-graph traversalDLS in SQL Server 2005●Calculate H code for each tetrahedron and create a clustered index on itDLS in SQL Server 2005●Point-in-Tet Test ●Intersection facets ●GetTetFaceNeighborPerformance●Points per second●20,000 points in each table T (created from a center-radius pair)●SELECT COUNT(DISTINCT dbo.fnGetTet4Point(x, y, z)) FROM
View Full Document