Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 1 W. Fredrick Limp University Professor Robert Harris DBA Center for Advanced Spatial Technologies University of Arkansas Spatially Enabling the DBMS.

Similar presentations


Presentation on theme: "1 1 W. Fredrick Limp University Professor Robert Harris DBA Center for Advanced Spatial Technologies University of Arkansas Spatially Enabling the DBMS."— Presentation transcript:

1 1 1 W. Fredrick Limp University Professor Robert Harris DBA Center for Advanced Spatial Technologies University of Arkansas Spatially Enabling the DBMS Design, implementation, operation and implications for the future

2 2 2 Outline n Review of geospatial data structures and their relationship to “traditional” DBMS data types n Feature and feature-class based spatial data models n Data storage and indexing in enterprise geospatial systems n OpenGIS and enterprise systems n SQL operations in 10G n Examples of on-line systems

3 3 3 Initial spatial data model n Paradigm was a map n Map layer theme file n Data model corresponded to technology driver n Developed in response to “slow” display hardware/software performance and need to be responsive

4 4 4 The attributed graphics model n Graphic files were linked to attributes in a separate file and later DBMS table n Allowed user to conduct searches on and get graphic element(s) that was associated Graphic Info element 1 Road 2 water line 3 valve 1 2

5 5 5 Replaced by - blob-based entity spatial model n Software developers take advantage of a special type of record in database - the blob – a blob is a chunk of disk space that the DBMS keeps track of but otherwise just knows “something is there” n Changes view of basic “thing” recorded n Spatial entities become the “element” – a particular road segment, a particular parcel

6 6 6 Database storage in this model Lot 2 Lot 1 Tract layer Tract 1population an attribute Geometric stuff another attribute 94102 94103 234 567 100 blob Columbia Jefferson All geometric stuff in a “blob”

7 7 7 BLOB problems (1) n BLOB contents are not managed by DBMS –BLOB contents are unknown to DBMS –Managed by 3-rd party software –All operations external to DBMS n BLOBs do not reflect known data types

8 8 8 BLOB problems (2) n Definition of a thing – This is now clear within a single theme – but still problem in confusion between themes District 1 The City Smith Road What’s this ??? What happens if the road is moved?

9 9 9 Theme and element confusion n What if the “road” changes - and it is also the legal definition of the city limit and voting district boundary? n If the road is an entity then perhaps the road department updates it – then the city limit and voting district boundaries are now out-of-date n BUT maybe that’s correct - the road changes but the district boundary doesn’t!

10 10 Spatial operations in this model n How do you find a geographic thing – find all old street segments within Alderman Smith’s district n old is an attribute n “district” is a geographic area n “IN”, “BESIDE” etc. are viewed as topological relationships n Requires specialized “GIS” operations and possibly specialized data structure n DBMS simply the data store –All operations in client or middle-ware

11 11 Enterprise systems and interoperability n Spatial data has (to date) been stored in many different internal formats n Common that each software vendor has a different spatial data format n Different formats made exchange of data difficult (if not impossible) n How to solve this in an enterprise context?

12 12 Solutions n Old solution – Everybody got the same software from the same vendor OR – ORGANIZATIONAL RULE was passed – Problems with this n one size doesn't fit all n limits technological innovation n New solution – Open GIS (TM) consortium and interoperability specifications – Multiple vendors = competition – Lower cost and increased capacity – Flexibility by units

13 13 Definition of OpenGIS ® Open and interoperable geoprocessing: The ability to share heterogeneous geodata and geoprocessing resources transparently in a network environment.

14 14 OpenGIS Specification is an Interface Standard (not a data or transfer) standard OpenGIS-compliant clients can be understood by OpenGIS-compliant servers: – Set or get geodetic datum and projection – Create, modify, copy, delete features in the database – Select features from the database by query – Execute GIS commands –And many others... –Adopted by ISO (TC211)

15 15 Simple features model n With development led, in large part, by DBMS vendors, a new spatial data model is created n Requires some new algorithms n But otherwise can utilize “normal” relational and/or object DBMS

16 16 Basic Data Model

17 17 Base Elements n Basic building blocks of a geometry n Element types –Point (1) –Line (2) –Polygon (3) n Constructed using coordinates n No limit on number of coordinates

18 18 Extended Elements Points Line/Circular Arc Strings Polygons Rectangle optimized for space Circles Compound elements without area Compound elements with area

19 19 Geometry n Represents a spatial feature n Consists of an ordered set of primitive elements n Uniquely identified with a geometry identifier Geometry 1 Florida, USA Geometry 2 Hawaii, USA

20 20 Layers n Consists of geometries that share a common set of attributes n Stored using a set of tables State layer

21 21 Conversion from map tile to feature - element in ORDBMS Map tile by “theme”Enterprise system’s feature classes

22 22 Database properties n Allows storage of individual element – each segment of road is potentially accessible – a geometric element serve multiple purposes n road segment n city limit segment n voting district segment n “Automatic” updates of all associated “themes” because they are simply DBMS linkages

23 23 Potential problems with this approach n Zillions of elements n How do you keep track of where they live – rapid access to a record in a DBMS provided by INDEX – index means that all records do not need to be sequentially accessed and checked n How to index spatial elements – attributes are X, Y (and maybe Z)

24 24 Quad tree indexing n A full tessellation of space

25 25 Morton Codes (HHcode) for Tiles

26 26 Use quad trees to spatially index _SDOINDEX SDO_GIDSDO_CODESDO_MAXCODE 1013 T2 T2FFFF…F 1013 T320 T302 T320FF…F T302FF…F

27 27 Spatial indexing implications n Spatial joins (a VERY BIG DEAL!) – IF same extent is used to create a quad tree index then the same index number corresponds to the same place on the earth Address layer Other layer Geocode returns Index location

28 28 R-tree index n Commercially introduced by Informix Spatial Cartridge –From Illustra technology –Data (not space)is partitioned n Bounding boxes (MBR) around each element are created n MBR encompassing lower groups n Index of MBR “values”

29 29 R-trees

30 30 R-trees n Variety of forms n Are balanced index structures n Appear to be much faster n Probably not effective as quad-tree in a very dynamic data environment n More reading –http://www.geog.ubc.ca/courses/klink/gis. notes/ncgia/u37.html

31 31 Spatial queries n Two stages –Stage 1 - Index n Based on spatial index n Rapid n Can dramatically reduce data size –Stage 2 - Exact n Uses spatial operators n Performs calculation

32 32 Query Model Layer Data Primary Filter INDEX Reduced Data Set Secondary Filter Query Functions Exact Result Set Table where coordinates are stored Index retrieves area of interest (window) Procedures that determine exact relationship

33 33 Spatial Queries in SQL n Based on Egenhofer Relationships Between Geometries –Inside –Contains –Covers –Covered By –Touch –Overlap Boundaries Intersect –Overlap Boundaries Disjoint –Equal –Disjoint –AnyInteract

34 34 INSIDE AND CONTAINS B INSIDE A A CONTAINS B A B

35 35 COVERS AND COVERED BY B COVEREDBY A A COVERS B A B

36 36 TOUCH A TOUCHES B B TOUCHES A A B

37 37 OVERLAP BOUNDARY INTERSECT A OVERLAPBDYINTERSECT B B OVERLAPBDYINTERSECT A A B

38 38 OVERLAP BOUNDARY DISJOINT A OVERLAPBDYDISJOINT B B OVERLAPBDYDISJOINT A A B

39 39 EQUAL A EQUALS B B EQUALS A A blue B green

40 40 DISJOINT A DISJOINT B B DISJOINT A A B

41 41 ANYINTERACT AND DETERMINE n ANYINTERACT - Given two geometries, returns true if the two geometries are not disjoint n DETERMINE - Given two geometries, returns their relationship to each other.

42 42 Define an Area Of Interest Area Of Interest

43 43 Features Inside Area of Interest INSIDE

44 44 Client or server queries n Many applications that connect with enterprise systems have option for query –Server side –Client side n Will return to this question later

45 45 Queries and indexing n Performance –Retrieval of ALL values –Retrieval of selected set –Data set size n Under 100,00 features n 100,000 – 1,000,000 features n More than 1,000,000 features

46 46 Relationships n Formal “connections” between elements in two tables in a database n Examples –Transformer is “on” power pole n Implies that geographic location of transformer is dependant on location of pole – move pole move transformer –Water meter is at end of water line –Bridge is connected to street segment –City limit is boundary of voting district

47 47 What happens in relationships n Messaging and/or “triggers” n User performs an action on a feature that has a relationship n Software recognizes existence of relationship and executes defined operation n May require user input may be automatic n Extensive use of relationships has performance implications n Use has design complexities

48 48 Transactions and versioning n Enterprise systems provide methods to allow features to be updated while under use –User 1 locks individual records are locked for edit/update purposes –User 2 can access data but nor change –If multiple users are updating rules define how multiple changes are applied n Versioning allows data configuration to be date defined

49 49 Metadata n Spatial metadata vs applications metadata n Spatial metadata –E.g. n ISO TC 211 19115 n FGDC –Originally file oriented n Date n Source –Enterprise systems are dynamic feature/element structures n Date = date of last update n Movement to element based, dynamic metadata


Download ppt "1 1 W. Fredrick Limp University Professor Robert Harris DBA Center for Advanced Spatial Technologies University of Arkansas Spatially Enabling the DBMS."

Similar presentations


Ads by Google