Advantages of using the Apache TIKA parser library for indexing

Starting TeamForge 7.0, the underlying parser library for indexing has been changed from Stellent to Apache TIKA.

The Apache TIKA parser library has the following advantages over the Stellent parser library:
Issue Stellent Apache TIKA
Stale process issue Parsing of corrupt or unrecognized files by the Stellent parser libraries often result in stale processes that consume swap space and add to the load on the system, which may at times lead to site outage. To manage such processes, you may choose to create and deploy stale process monitors and the stale processes, when detected, must be removed manually to prevent site outage. Parsing of unrecognized or corrupt files by Apache TIKA libraries is robust and needs no manual intervention as there are no stale process issues.
Search queue processing speed It takes five minutes to timeout when the Stellent parser library encounters a corrupt or unrecognized file that it knows not how to parse. If there are more such corrupt or unrecognized files, more time is wasted by the indexer waiting for a response (or a timeout) from the Stellent parser, which in turn adversely impacts the search queue processing speed. The Apache TIKA parser library is capable of determining whether a file it encounters can be parsed or not. As no time is wasted by the indexer waiting for a response (or a timeout) from the parser, the search queue processing speed is better with the Apache TIKA.
Multiple processes Vs Single JVM For parsing files, the Stellent parser library spawns one subprocess per file. Meaning, the number of subprocesses is equal to the number of files to be parsed and it is possible that we may end up with the stale process issue as discussed earlier. As a result, if the Stellent processes consume more resources, other processes and applications are left with scarce resources. The Apache TIKA, being a Java-based parser library, works within the JVM and makes the external resource pool available exclusively for other processes and applications. As the search JVM, where the Apache TIKA library lives, can also be separated starting TeamForge 7.0, it can be managed better.