abarry@instra.gsfc.nasa.gov
The World-Wide Web (WWW) offers new solutions to growing problems in managing information generated by large technical development efforts. At NASA's Goddard Space Flight Center (GSFC), coordination and exchange of program information has, in the past, been inefficient and time-consuming. Problems become increasingly difficult managing international space projects. A data management system based on the WWW solves these problems. The initial implementation is currently being used at GSFC to manage collaborative space exploration projects like the Cassini/Composite Infrared Spectrometer.
Managing engineering information generated in designing and building spacecraft or space instrumentation has proven difficult for a number of reasons. These problems are not limited to the space industry but are common to any multi- disciplinary effort managing large amounts of data of different types, particularly when data is generated at multiple sites. The level of difficulty in handling these problems is in large part determined by a project's size and complexity.
Collecting and standardizing information is the most basic problem encountered in managing data for any project. This can be made difficult and costly as a result of the many data types generated during a project's life cycle. Computer-aided design (CAD) drawings, bit-mapped images, text documents, schedules, budgets and spreadsheets as well as specific outputs from various thermal, mechanical, optical and electrical analysis packages are among the many data types that a design project must contend with. In addition, these materials typically reside on hardware platforms that are often incompatible.
Second, the lack of centralization and standardization impedes data access. Users, both local and remote find it cumbersome to obtain information generated by others. For instance, the originator of a mechanical drawing stores it locally on hard drivein its native format. Until a formal release or revision has been made, others can only get the current drawing through the originator. Moreover, the data may not be in a format that is convenient to use. If the originator happens to be halfway around the globe, another engineer may find it extremely difficult to get a recent copy for reference.
Difficulties in data access are compounded by a third problem: configuration management (CM). Much time and effort is spent collecting and archiving data properly to ensure that the latest approved version of a drawing or document is available and distributed in a timely manner. Currently, it may take hours or even days to get copies of controlled data depending on how records are kept and what methods are used to reproduce it. Tracking records of drawings are kept in a separate database. For information not under configuration control, the project relies on the organizational skills of the originator to ensure that users are provided with the most current copy. This has often proved disastrous. In addition, whether the data is controlled by the individual or bythe CM group, it is not only difficult but differs from project to project thus preventing standardization not only of data within a project but of data managed across project boundaries.
CM databases have also posed problems in project data management. Typically, they catalog only a fraction of the data and no provision is made for data storage within the same database framework. The database keeps tedious tracking records but the overhead of data entry and maintenance on such a system is high. Data access is through database queries and is available only through a project's CM representative. Database response time is slow. In addition, database inadequacies cannot be addressed due to a lack of flexibility in the database design.
Many program management tools solve some of these problems but provide only for the management of program support data such as schedules and budgets and not of engineering or other technical information. Moreover, they do not provide solutions to the additional complications of managing data generated bygroups that are not co-located. The complexity of these management tools also require proper training and expertise of a staff which may not be of a technical nature.
The Project Information Management System (PIMS)
The system provides a central home page with links to introductory pages for each project. Subsequent HTML pages provide background information and links to specific project data areas on disk. The top-level project directory is broken into sub-directories, each containing information relating to a specific area of project development usually based on data type (documents, drawings, schedules etc.) thus mapping into a tree-like structure.
The uniqueness of this system is in the scheme used to manage the data. Custom CGI scripts have been written to provide a dynamic, graphical representation of the data trees. For instance, an engineering drawing tree would have sub-assemblies and individual drawings. These map nicely to directories and files. Each sub-assembly becomes a directory and the drawings are files within. They can be represented graphically as separate icons that are hyper-linked to the next assembly level or to the actual drawing stored on disk. Each time a user clicks on an icon to take them to a new assembly, the script dynamically regenerates the HTML display page. This guarentees that the most recent information is presented as new drawing files are stored under their associated sub-assemblies. Generating the tree dynamically also saves maintenance overhead over one implemented statically.
Each archived file has an associated header file in HTML format. When a user clicks on the icon for a drawing, the CGI script brings up the header file. The file contains drawing information previously stored in the database (e.g. size, originator, title, comments, keywords, release date, associated changes, references etc.). Header information may be static or may again be hyper-linked to other pages that give more detail. For example, any changes that are written against a drawing have to be documented and signed off. In PIMS, the change orders are associated to the drawing by the file naming convention and appear as hyper-links in the header file. When a user clicks on a hyper-link for a particular change order, the order is brought up on the screen as an ASCII text file thereby giving the user immediate access to the original data, reference data and the associated database information. Having the data available in this way eliminates much of the clerical bureaucracy involved with keeping hardcopies and distributing them through conventional mail systems.
Custom CGI scripts perform simple to complex level searches for information within the tree, to generate on-line configurable reports and to perform some rather simple database functions. For example, searching the tree or any of its branches for a particular drawing that contains user specified keywords.
Some data previously embedded in the database is also viewed directly in the file name. In our case, we use the drawing or document number followed by a page number (used when a drawing has multiple pages or document has multiple volumes) and a version number thus making the most pertinent information immediately obvious and available. The file type is indicated in the suffix. Files are stored both in their native format, a standard viewable format (GIF) and a printable format (text, postscript or plot) so users may view information even if they do not have the specific application package used to generate it.
Under this system, configuration management is simpler and more efficient. All CM operations can be performed on-line through the browser via CGI scripts. This includes the ability to upgrade a revision for a particular file, to move files from one directory to another, purge files, create reports and modify header file information. Because each CM operation is mapped to a separate CGI script, the system is more flexible and allows for modifications more readily than its database predecessors. Having these procedures available on-line eliminates the need for special database servers and dedicated terminals. The procedures also insulate the user from operating system level details so that no special training or computer skills are required to use the system.
The Project Information Management System has solved the problems surrounding large-scale project data management. All project data, regardless of type or point of origin, is centralized on the PIMS serve and access is direct rather than through complicated database searches through the CM group. Being Web-based implies that access is limited only by a user's inability to connect to the Net. It is cost-effective, easy to use and runs on all commonly used hardware platforms. Sophisticated custom functions have been developed to provide a totally integrated archival and management environment. The graphical nature of the interface allows images to be imbedded, giving remote users quick access to detailed information that was previously unavailable and may have taken days or weeks to obtain. Multiple projects can now co-exist and be managed using the same system, thus controlled information is now immediately available and consistent, even between projects.