Microsoft TerraServer: A Spatial Data Warehouse

1
Microsoft TerraServer A Spatial Data Warehouse

• Analyzing a research paper
• CS 8715-Spatial Databases
• Instructor-Shashi Shakhar
• Group - G6
• David Moretz and Chandana Gangodagamage

2
Problem statement

• Designing a database model for spatial data to be
  accessed by thousands of simultaneous users via
  standard web browser is a demanding and long
  term aspiration of internet users.
• The web is inherently html base environment and
  in general, spatial data is internet unfriendly
  since its complexity in modeling through a
  database system.
• Delivering terabytes of spatial (aerial images,
  satellite images, and topographic maps) and
  attribute information at low internet connection
  speed need ways and means of reducing
  dimensionality.
• Technical issues of possibility of delivering
  spatial information over the internet through
  relational database system without using object
  oriented or object-relational system.

3
Major contribution of the paper

• Microsoft Terra server is the first client/server
  imagery database application which use SQL based
  relational database.
• Terra server can be accesses by millions of
  users, quarries only examine relative few records
  to deliver the answer, and records can contain
  ten time more than the traditional system.
• Users connect at lower speed(32kbs) can access
  all high resolution USGS Digital
  Ortho-Quadrangles (DOQ),USGS Digital Raster
  Graphics (DRG),Aerial images SPIN-2, and Encarta
  Shaded Relief Data at different scales.
• Introduction of terra cutter and terra scale
  components into terra server system where terra
  cutter reformat the imagery received from data
  source and terra scale compute lower resolution
  image-pyramid tiles.

4
Key concepts

• Terra server is a image database tiling system
  and it vary for online mapping websites in that
  the latter project the data to match the user
  request while the former displays the data in
  the projection chosen by the user.
• Terra server is a collection of seamless scenes
  where it used UTM and NAD ellipsoid to project
  the data in to planner surface and create a
  mosaic of user interest area.
• Terra server introduces a grid system which
  consist of themes resolution, scenes, and
  location to enable zoom and pan across the mosaic
  of tiles.
• Each tile is attached a six fields Resolution,
  Theme, SceneID (Individual scene Identifier),
  Scale (at scale 1m z10), X (tiles relative
  position on X axis), and Y (Tiles relative
  position on Y axis)
• E.g.. Digital Orthoquadrangles-Theme1-T1 UTM zone
  10(s10) at scale 1m(z10) with X2766 and
  Y20913.
• All these Image tiles and metadata are stored in
  an SQL database.
• The Gazetteer database schema is introduced in
  database design and system architecture is shown
  in the figure.
• Terra Cutter which breaks a large image up into
  tiles and terra scale which resample the tiles to
  create lower resolution tiles were introduces to
  terra server as load management concepts.

5
Example of Theme and UTM zone, scale, X, and Y
Gazetteer Scheme
http//terraserver.microsoft.com/tile.ashx?t1s1
1x1203y12450z15 200200 pixels UofM EE
building http//terraserver.microsoft.com/tile.ash
x?t2s11x1203y12450z15 http//terraserver.
microsoft.com/tile.ashx?t4s11x1203y12450z
15
Figure 1 Imagery database scheme
Figure 2Gazetteer Database Scheme
6
Validation methodology

• Microsoft Terra server project validation may be
  different from other research project. We
  identified following validation criteria.
• User Input18 000 mails massages were received
  from users and most of them are constrictive
  criticism or praise. The most common complain was
  the missing images.The second common complain was
  alignment of images against each other. The third
  common complain was the java applet which does
  not run on certain machines.
• Network Traffic When terra server site launched
  in 1998 it received 35 millions web hits and 8
  million page views and millions more rejected.
  So microsoft team quickly grew the site to 10
  servers.
• Terra server has reach steady state of five to
  six million web hits and five to eight million
  database stored procedure execution, and 50 GB of
  image tile down load per day.
• System availability The system went out of
  service for three hours for software upgrade, 2.5
  hours to move the server within the centers, 35
  minutes due to software bugs. The overall system
  availability is 99.97 for the period through
  july1998 through July1999.

7
Strengths and weaknesses

• Strengths
• Terra server dramatically reduced the access
  complexity and cost for simple application in
  remote sensing and aerial photagrammetry and it
  makes the availability of high resolution images
  for whole internet community.
• Using relational database to pyramids of image
  file format separate the storage management from
  image presentation and tile design allow rapid
  pan and zoom to any part of the database.
• Weakness
• Aerial image source is not artho-rectified and
  hence mismatch with other datasets.
• Client application can identify the overlapping
  tiles from separate themes since terra server
  tiling algorithms cuts image in to separate
  tiles.
• The coverage of image database limited even in US
  and very less coverage in other parts of the
  world.
• No proper research validation has carried out and
  mainly oriented on Microsoft product promotion
  aspect.

8
Assumptions

• Major assumption of this work is that the
  relational database can fully support to large
  scale image repositories and standard web
  browsers are good geo-spatial image viewers.
• User interface to terra server can function
  adequately over low-speed (28.8kbps) connections.
• Terra server can handle steady state network
  traffic.

9
Rewriting the paper

• Even though terra server was introduced in 1998,
  there were many aspects which can be preserved
  even today terra server grid system ,Imagery
  database scheme, Gazatteer database scheme, terra
  cutter concept, and terrace scale.
• Terra cutter concept can be incorporate with
  wavelet decomposition technique to represent to
  images at different scales.
• Terrace server use the relational database
  concept which were brave step at 10 years before
  , object oriented or object relational database
  systems works well with spatial data schemes

10
Rewriting the paper

• Microsoft terra server does not integrate with
  any of the vector layer. Vector layer can used as
  intelligent objects size of the vector layers are
  small and provide easy access to end users.
• We definitely integrate 3D viewing capability,
  capability of user to load their own dataset to
  the viewer, various GIS operation like buffer,
  distance calculation etc.

11
Justification

• It is a great step taken by Microsoft at exactly
  a decade before to integrate spatial data with
  relational databases.
• During last 10 years technology has change and by
  learning lessons from those in the past today
  Google Earth has introduced excellent technology
  for spatial data integration paradigm.