A%20Domain-Specific%20Modeling%20Language%20for%20Scientific%20Data%20Composition%20and%20Interoperability

1
A Domain-Specific Modeling Language
forScientific Data Composition and
Interoperability
Hyun Cho University of Alabama at Birmingham
Jeff Gray University of Alabama
2
File Formats Image Files

• Organize and store digital images that are
  composed of either pixel or vector (geometric)
  data
• Bitmap-based
• Created by scanner and digital camera
• TIF, JPG, BMP
• Vector-based
• Geometric description Bitmap
• Resolution Independent Infinitely scalable
• Font, DRW, CGM

3
File Formats Music and Audio Files

• Storing audio data that are produced by
  audio-to-digital converters
• Key Parameters
• Sample Rate, Resolution, Number of channels
• Uncompressed formats
• WAV, AIFF and AU
• Lossless compression Formats
• FLAC, Lossless Windows Media Audio (WMA)
• Lossy compression Formats
• MP3, Lossy Windows Media Audio (WMA)

4
File Formats Text Files

• File formats that are structured as plain text,
  representing a sequence of lines
• ASCII, TXT

5
File Formats Compound File Formats

• Used to structure the contents of a document in
  the file
• Contain a number of independent data streams that
  are organized in a hierarchy
• Stream files in a file system
• Storage sub-directories in a file system
• MS Office, OpenOffice

6
Characteristics of Generic File Formats

• Can handle one or two data types
• Numeric data or alphanumeric data
• May have a limitation of the file size
• Mostly limited to a maximum file size of 2GB
• May increase file I/O time linearly as the file
  size grows

An In-Depth Examination of Java I/O Performance
and Possible Tuning Strategies http//pages.cs.wis
c.edu/remzi/Classes/736/Fall2000/Project-Writeups
/KaiHongfei.html
7
Characteristics of Generic File Formats

• Can handle one or two data type
• Numeric data or alphanumeric data
• May have a limitation of the file size
• Mostly limited to a maximum file size of 2GB
• May increase file I/O time linearly as the file
  size is grew

These generic file formats are not appropriate
for storing and retrieving scientific data
because the files were not designed to maintain
high volume of complex scientific data, such as
high resolution images, massive numerical data,
and graphs.
An In-Depth Examination of Java I/O Performance
and Possible Tuning Strategies http//pages.cs.wis
c.edu/remzi/Classes/736/Fall2000/Project-Writeups
/KaiHongfei.html
8
Scientific Data Format NetCDF3

• Network Common Data Format
• Machine-independent file format
• Support a wide variety of platformsincluding
  Linux, MacOS, Windows
• Representing multi-dimensional arrayswith
  ancillary data

Time 1
Time n
9
Scientific Data Format HDF5

• Hierarchical Data Format
• File format for managing any kind of data
• Support high volume and/or complex data
• Platform-independent
• Flexible, efficient storageand I/O

10
Characteristics of the Scientific Data File
Formats

• Self-Descriptive
• Contain metadata to inform the contained data
  type and their organization
• Directly Accessible
• Can access arbitrary data through APIs
• Concurrently Accessible
• Multiple threads or processes can access data
  simultaneously
• Enable high performance computing and speedier
  access
• Archivable
• Have their own archiving mechanism to backup and
  restore a high volume of data

11
Challenges in Using the Scientific Data File
Formats

• Use different representations to organize the
  file structure
• Each file format needs its own data visualization
  and composition
• It is difficult to exchange data between two or
  more scientific data formats
• Manage the evolution of APIs
• Challenging to verify that APIs are evolved in
  accordance with the evolution of file
  specification
• Maintain stability of existing applications from
  API evolution
• User applications are subject to change of APIs
• Limited support for data integration among
  heterogeneous scientific data formats

12
Framework for Scientific Data File Management
13
NEW SLIDES NEEDED HERE TO INTRODUCE DSM!
14
Model-Driven Engineering (MDE) and
Domain-Specific Modeling (DSM)

• MDE specifies and generates software systems
  based on high-level models
• Domain-Specific Modeling (DSM) a paradigm of MDE
  that uses notations and rules from an application
  domain
• Metamodel defines a Domain-specific Modeling
  language (DSML) by specifying the entities and
  their relationships in an application domain
• Model an instance of the metamodel
• Model Transformation a process that converts one
  or more models to various levels of software
  artifacts (e.g., other models, source code)

15
Unifying the representation of file structure
organization
Analyze data model of each scientific file format

• Adapt a DSML to build a tool for visualizing
  composing the scientific file format in a unified
  way

Common Data Model
Feature Model
Variable Data Model
Define DSML from Feature Model
Grammar Syntax
Implement DSML
DSML Tool
16
Unifying the representation of file structure
organization

• Feature Model for Scientific File Format
• Describe some highlights here
• And here

17
Unifying the representation of file structure
organization

• Content Composer
• DSML Modeling tool for scientific data file
• Implemented by using GEMS

18
API Abstraction Layer

• Help to protect user applications from the
  evolution of APIs

Abstraction
createFile( const char path, FileCreationProperty fileCreationProperty)
NetCDF HDF5
int nc_create ( const char path, int cmode, int ncidp) H5File ( const char name, unsigned int flags)
19
Integrating data among heterogeneous data formats

• Content Mapper
• Define rules how to map data from a scientific
  data format to another
• Content Verifier
• Verify the correctness of the file composition
• Verify the correctness of mapping rule

20
Summary

• From the prototype of the framework
• A DSML can help to build a graphical tool to
  compose and support interoperability across
  scientific file structures
• Adoption of the layered architecture in the
  framework can help to maintain the independence
  of each layer
• Both the API abstraction layer and the layered
  architecture are essential to develop and
  maintain user applications
• Further works
• Create metamodels that include full specification
  of each scientific file
• Categorizing APIs in accordance to their intended
  use for API abstraction layer
• Develop metamodels for managing API evolution

21
Thank you!
22
Example of Scientific Data Format OPeNDAP

• Client-server protocol for scientific data access
• Targeted oceanographic data management