InstallingHiveWeb

Installing HIVE Services

This page describes how to install and initialize the HIVE services, including:

• HIVE Web Service
• Importing vocabularies into HIVE

System Requirements

• Java 1.6
• Java servlet container/application server (Tomcat 6.x recommended)
• See memory requirements below

Memory Requirements

When using the KEA++ indexer with the H2-based vocabulary and all vocabularies loaded, HIVE requires ~512MB of memory.

To increase the maximum JVM heap size, use the -Xmx option. To limit the proportion of VM spent in garbage collection before an OutOfMemoryError is thrown, use -XX:UseGCOverheadLimit. In Tomcat, this can be set using the CATALINA_OPTS variable, as in catalina.sh:

CATALINA_OPTS="-Xmx512m -XX:-UseGCOverheadLimit"

For KEA++ memory requirements using the in-memory indexer, see the HIVE memory usage page.

Installing Tomcat

The HIVE services can be installed using any Java servlet container or application server. For the purpose of this document, Tomcat 6.0 is assumed.

• Download the latest Tomcat 6.x release
• Follow the Tomcat install instructions
• Optional: Modify the catalina.sh to set CATALINA_OPTS for increased heap allocation (see memory requirements above).

Simple install steps:

• Download Tomcat zip or .tar.gz file
• Extract archive (unzip *.zip or tar xfz *.tar.gz)
• cd <tomcat_install>/bin
• bin/startup.sh
• Goto http://localhost:8080/, confirm Tomcat startup page displays

Installing the HIVE Web Service

• Download or build the hiveweb.war
• Note: Due to a limitation in the version of GWT used by HIVE, the HIVE Web application must be installed as the ROOT webapp.
• Extract the contents of the hiveweb.war into the ROOT webapp directory

  cd /path/to/tomcat6/webapps/ROOT
  unzip /path/to/hiveweb.war

• Modify the WEB-INF/conf/hive.properties file. This file contains the list of vocabularies configured for this HIVE Web instance and the "tagger" used for automatic indexing.

  # Configured vocabularies
  hive.vocabulary = # vocabulary name, e.g., "lcsh"

  # Selected tagger
  hive.tagger = # "kea" or "dummy"

• Create a hive.vocabulary row for each configured vocabulary.
• Set the hive.tagger value to either "kea" for KEA++ indexing or "dummy" for basic indexer.
• Create a file <vocabulary>.properties for each vocabulary listed in the hive.properties file.

HIVE Sample Data

To get started quickly, you can also download the AGROVOC sample data. This sample data includes pre-initialized indexes and KEA++ model.

1. Download and extract hive-agrovoc-sample.zip from the Downloads page
2. Move the hive-data directory to the desired location (default is /usr/local/hive/hive-data)
3. Copy agrovoc.properties to the WEB-INF/conf directory
4. Edit agrovoc.properties, update paths to location of hive-data directory.

Importing Vocabularies into HIVE

The HIVE services require the initialization of several indexes from one orm ore vocabularies in SKOS RDF/XML format. You will not be able to start the HIVE Web or Rest services until this initialization process is complete. This process creates a Sesame store, Lucene index, alphabetic and top-concept indexes from the source RDF. If KEA++ indexing is enabled, the import process also creates and trains the KEA++ mode.

The import process is detailed in the ImportingVocabularies section. A brief summary of the requirements include:

1. A vocabulary in SKOS RDF/XML format
2. A HIVE vocabulary configuration file ((e.g., <vocabulary>.properties
3. A hive.properties file
4. If "kea" indexing is enabled, a set of training documents