1. Introduction¶

The goodies of database technology for building flexible, large-scale information systems are well known:

information integration bringing along better consistency and much easier administration
flexibility by replacing static APIs by dynamic query languages
scalability achieved through advanced storage and processing techniques, in particular: query optimization which is a powerful performance booster
maturity of decades of development and functionality richness, as opposed to 1.0 versions of reinvented wheels (load balancing, indexing, transaction handling, catalog metadata management, …)

Unfortunately, these advantages until now can be reaped only for alphanumeric (and, more recently, also vectorial) data types. Raster data (also called gridded data, sampled data, etc.) do not benefit, since traditional databases do not support the information category of large, multidimensional arrays.

This gap is closed by the rasdaman technology which offers distinct array management features. Its conceptual model supports arrays of any number of dimensions and over virtually any cell (“pixel”, “voxel”) type. The rasdaman query language, rasql, is crafted along standard SQL and gives high-level, declarative access to any kind of raster data. Its architecture principle of tile stream processing, together with highly effective optimizations, has proven scalable into multi-Terabyte object sizes. Rasdaman has been developed since 1996 and since has reached a high level of maturity itself being in operational use since many years.

The rasdaman project strongly commits itself to open standards, in particular those of the geo service community. We actively participate in the development and maintenance of the Open Geospatial Consortium open geo raster standards. Among other activities, we have developed specification and reference implementation of OGC WCS 2.0, WCPS, and WPS.

The free and open-source rasdaman community version is available for free download; generally, rasdaman is available in a dual [wiki:License license model]. If the scientific background of rasdaman is of interest, then check out our Publications …and cite our papers!

Finally, for the legalese see Imprint and Disclaimer.

1.1. Features¶

Technically, rasdaman is a domain independent Array DBMS, which makes it suitable for all applications where raster data management is an issue. The petascope component of rasdaman adds on geo semantics for example, with full support for the OGC standard interfaces WCS, WCPS, WCS-T, and WMS; see matrix below for details and EarthLook for a kaleidoscope of hands-on interactive demos.

Historically, rasdaman has pioneered the field of Array Databases, being the first system of this kind. The rasdaman technology has been developed over a series of EU funded projects and then marketed by rasdaman GmbH, a research spin-off dedicated to its commercial support, since 2003. In 2008/2009, the company has teamed up with Jacobs University for a code split to establish rasdaman community (encompassing a complete Array DBMS) as an open-source project. The original rasdaman code remains as rasdaman enterprise. Both are kept in sync at any time, and both rasdaman GmbH and Jacobs University contribute actively to the open-source project. The features covered here concern the open-source community version; a summary of rasdaman community versus rasdaman enterprise was presented here earlier, but has been meanwhile abandoned as OSGeo frowned on this.

Contributions to the rasdaman community code come from a worldwide team of collaborators. Notably, a significant extent of the fixes and new functionality is coming from rasdaman GmbH (such as WMS recently). All community contributions submitted are made available in rasdaman community immediately after checking them for correctness and coherence (eg, with the code guide); no contribution whatsoever goes into rasdaman enterprise first. The only action the company undertakes is to keep both rasdaman variants in sync by merging the rasdaman community tree into rasdaman enterprise, which typically occurs upon release of versions so as to keep both in sync. Aside from that, rasdaman enterprise is developed exclusively by the company and does not contain any community code that rasdaman community does not contain. So rest assured that your valuable contributions are to the benefit of the worldwide user community.

1.1.1. Array data model¶

Arrays are determined by their extent (“domain”) and their cell (“pixel”, “voxel”). Extents are given by a lower and upper bound taken from the integer domain (so negative boundaries are possible as long as the lower bound remains below the upper bound). For the cells, all base and composite data types allowed in languages like C/C++ (except for pointers and arrays) can be defined as cell types, including nested structs.

Over such typed arrays, collections (ie, tables - ODMG style, again) are built. Collections have two columns (attributes), a system-maintained object identifier (OID) and the array itself. This allows to conveniently embed arrays into relational modeling: foreign keys in conventional tables allow to reference particular array objects, in connection with a domain specification even parts of arrays.

As such, rasdaman is prepared for the forthcoming ISO SQL/MDA (“Multi-Dimensional Arrays”) standard, which actually is crafted along rasdaman array model and query language. This standard will define arrays as new attribute types, following an “array-as-an-attribute” approach for optimal integration with relations (as opposed to an “attribute-as-table” approach - as pursued, e.g., by SciDB and SciQL - which has some remarkable shortcomings in practice).

1.1.2. Query language¶

The rasdaman query language, rasql, offers raster processing formulated through expressions over raster operations in the style of SQL. Consider the following query: “The difference of red and green channel from all images from collection LandsatImages where somewhere in the red channel intensity exceeds 127”. In rasql, it is expressed as

select ls.red - ls.green
from LandsatImages as ls
where max_cells( ls.red ) > 127

Rasql is a full query language, supporting select, insert, update, and delete. Additionally, the concept of a partial update is introduced which allows to selectively update parts of an array. In view of the potentially large size of arrays this is a practically very relevant feature, e.g., for updating satellite image maps with new incoming imagery.

Query formulation is done in a declarative style (queries express what the result should look like, not how to compute it). This allows for extensive optimization on server side. Further, rasql is safe in evaluation: every valid query is guaranteed to terminate in finite time.

1.1.3. C++ and Java API¶

Client development is supported by the C++ API, raslib, and the Java API, rasj; both adhere to the ODMG standard. Communication with a rasdaman database is simple: open a connection, send the query string, receive the result set. Iterators allow convenient acecss to query results.

Once installed, go into the share/rasdaman/examples subdirectory to find sample code.

1.1.4. Tiled storage¶

On server side, arrays are stored inside a standard database. To this end, arrays are partitioned into subarrays called tiles; each such tile goes into a BLOB (binary large object) in a relational table. This allows conventional relational database systems to maintain arrays of unlimited size.

A spatial index allows to quickly locate the tiles required for determining the tile set addressed by a query.

The partitioning scheme is open - any kind of tiling can be specified during array instantiation. A set of tiling strategies is provided to ease administrators in picking the most efficient tiling.

1.1.5. Tile streaming¶

Query evaluation in the server follows the principle of tile streaming. Each operator node processes a set of incoming tiles and generates an output tile stream itself. In many cases this allows to keep only one database tile at a time in main memory. Query processing becomes very efficient even on low-end server machines.

1.1.6. Server multiplexing¶

A rasdaman server installation can consist of an arbitrary number of rasdaman server processes. A dynamic scheduler, rasmgr, receives incoming connection requests and assigns a free server process. This server process then is dedicated to the particular client until the connection is closed. This allows for highly concurrent access and, at the same time, increases overall safety as clients are isolated against each other.

1.2. Rasdaman Application Domains¶

Its features make rasdaman suitable for all applications where raster data management is an issue, such as:

earth sciences

1-D sensor time series; 2-D airborne/satellite image maps; 3-D satellite image time series; 3-D geo tomograms; 4-D climate and ocean data; … At EarthLook there is a demonstration of services on 1-D to 4-D geo raster objects. The workhorse of the service stack is rasdaman, running on top of PostgreSQL.

space sciences

2-D visibility maps; x/y/frequency observation data cubes; 4-D cosmological simulation data; …

life sciences

3-D brain activation maps; 3-D/4-D gene expression maps; …

engineering

1-D measurement time series; 3-D/4-D simulation result data; …

multimedia

1-D audio; 2-D imagery; 3-D movies; …

See the publication list for descriptions of a variety of projects where rasdaman has been successfully used.

1.3. OGC geo standards support¶

While rasdaman itself is domain agnostic and supports any array application, the petascope servlet, as part of rasdaman, adds in geo semantics, such as dealing with geo coordinates. To this end, rasdaman implements the Open Geospatial Consortium standards for gridded coverages, i.e., multi-dimensional raster data. The OGC service interfaces supported are - Web Coverage Service: a versatile, modular suite for accessing and server-side processing of coverages, - Web Coverage Processing Service: OGC’s Big Datacube Analytics language, - Web Map Service: for rendering coverage data into maps which can be displayed with a wide range of open-source and commercial clients.

The Princial Architect of rasdaman, Peter Baumann, is chair of the OGC WCS Standards Working Group (WCS.SWG) and editor of coverage model (GMLCOV), WCPS, and most of the WCS specifications, rasdaman naturally has become Reference Implementation for several of these standards and usually implements them first and way ahead of other systems, even before final adoption. Likewise, any changes to coverage-related specifications usually are verified in rasdaman first and, hence, become available early. The same holds for the OGC conformance testing of coverage services where rasdaman code contributors have a lead. In summary, rasdaman can be considered the most comprehensive and best tested implementation of the OGC coverage standards.

1.4. How to Contribute¶

There are lots of ways to get involved and help out with the rasdaman project:

Help us spot & fix bugs.

Which software is perfect? We know there are some bugs in rasdaman, see the open tickets (or the low complexity tickets for beginners. Whether you add a ticket or provide a fix, all is most welcome.

Write documentation.

Users can always benefit from better documentation. Currently the documentation is in reStructuredText format, and HTML/PDF is automatically generated. We’re eager for any documentation contributions.

Contribute to the Wiki.

Of course you can also contribute to the wiki, for example by adding HowTos and FAQs. Send a message with a change request to patch in the domain rasdaman.org.

Help plan and design the next version.

Browse this section of the website, we use “Feature” tickets to hold ideas for new features; add your own and/or discuss a topic on the dev list.

1.5. Reporting problems¶

Reporting problems should be done by sending an email to the rasdaman-users mailing list. Your email should include a report with relevant information about the issue, either prepared with help of the prepare_issue_report.sh script, or manually specified as bellow:

OS distribution and version (see /etc/os-release)
Rasdaman version

Ubuntu: apt-cache show rasdaman

CentOS: yum info rasdaman

Concise description of your activity that led to the problematic behavior: queries, package management commands, etc.
Properties of the data operated on, including (but not limited to) data format, pixel type, coordinate reference system, dimension, along with data ingestion details (ingredients files, scripts); include a small resized data sample if possible, output of WCS DescribeCoverage of affected coverages, dbinfo on the underlying collections and spatial domain:
```
coverageId=???
ows_endpoint="http://localhost:8080/rasdaman/ows"
describe_cov_req="service=WCS&version=2.0.1&request=DescribeCoverage&coverageId=${coverageId}"

curl "${ows_endpoint}?${describe_cov_req}" > DescribeCoverage.xml
rasql -q "select dbinfo(c) from $coverageId as c" --out string > dbinfo.json
rasql -q "select sdom(c) from $coverageId as c" --out string > sdom.txt
tar cfz /tmp/data_details.tar.gz DescribeCoverage.xml dbinfo.json sdom.txt
```
Attach /tmp/data_details.tar.gz to the report, along with the ingredient files. Sample data should be uploaded elsewhere, e.g. Google Drive, if it is larger than 20 MB.
Relevant log files in /opt/rasdaman/log and /var/log/tomcat*/; you can compress the last 20 log files as follows (but try to execute the problematic query / operation last, just before the step below):
```
cd /opt/rasdaman/log
petascope_log="$(sudo find /var/log/ -name petascope.log)"
tar cfz /tmp/rasdaman_logs.tar.gz $(ls -t | head -n 20) $petascope_log
```
Attach /tmp/rasdaman_logs.tar.gz to the report.

Prior to sending your request, you should inspect the log files, they may already provide a clue that helps you resolve the issue.

1.5.1. Script for issue reporting¶

Rasdaman distributes with a prepare_issue_report.sh script in /opt/rasdaman/bin, which helps prepare a report for an issue encountered while operating rasdaman. Running the script will open an editor where you can enter a description of how the issue got triggered.

Various options can be specified to control what additional information is included in order to help developers in understanding and reproducing the issue. Following the options, you can specify files to include in the report, e.g. screenshots, ingredient files for importing data, sample (downsized if possible) data, etc.

Everything is compressed into a single archive in the current working directory from which the script is executed, and the path to it is printed at the end.

By default the script will try to include config files, latest 200 log files, petascopedb, and RASBASE, as long as the resulting archive is not larger than 20 MB to make it suitable for sending by email. Parts which are too large will be left out, in reverse order of priority (first RASBASE, then petascopedb, etc). The limit can be changed with –limit-size <N>. As soon as a particular –include-* option is specified, the default behavior is no longer in effect and exclusively the specified options are considered.

Check prepare_issue_report.sh --help for a list of all available options.

Examples

Describe the issue, including config files and 100 most recent log files, as well as a screenshot illustrating the problem:

$ prepare_issue_report.sh --include-recent-logs 100 -f screenshot.png \
                          --no-coverage-id

Describe the issue, include config files, all log files, petascopedb and RASBASE, as well as sample data and ingredients:

$ prepare_issue_report.sh --include-all-logs --include-petascopedb \
                          --include-rasbase -f sample_data.tar.gz \
                          -f ingredients.json --no-coverage-id

Like the first example, but also include information about coverage TestCov:

$ prepare_issue_report.sh --include-recent-logs 100 --coverage-id TestCov \
                          -f screenshot.png

Provide a screenshot and include details up to a maximum archive size of 20 MB (default behavior):
```
$ prepare_issue_report.sh -f screenshot.png --no-coverage-id
```