User:Janhoy/FAST Enterprise Search Platform (ESP)

FAST ESP (Enterprise Search Platform) is Microsoft's flagship Enterprise Search solution for "Internet Business", i.e. needs not using Sharepoint.

ESP, being Fast Search & Transfer's main product, is a generic search engine capable of indexing structured as well as unstructured content from a multitude of sources and content repositories including the web.

Main Features

 * Full-text search
 * Structured search of Strings, Numbers, Dates
 * Faceted search (Navigators)
 * Boolean query logic
 * Range query
 * Hit Highlighting
 * Sorting of results, including multi level sort
 * Document Processing Pipeline
 * Lemmatization in >40 languages
 * Spell Checking (Did You Mean)
 * Entity extraction
 * Automatic language detection
 * Document conversion for >220 document types

Competitive edge
ESP's main strength has been its fast sub-second search response time as well as its linear scalability capabilities. When growing in content volume, new columns of servers are added. When growing in query volume, new rows of servers are added. In addition the strong linguistics capabilities with i.e. automatic language detection for >80 languages, as well as advanced entity extraction has often been a competitive edge.

The scalability and query response advantages are however not unique anymore, as even open source engines like Apache Solr are able to match these.

Technology
ESP is written mainly in C++, Java and Python programming languages. The Document Processing Pipeline can be extended by users, writing Python scripts. Users can also interact with ESP using Web Services.

Platform support in version 5.3 includes Windows, Linux and Solaris-X86. Starting from 2010, Microsoft will continue development for the Windows platform only, abandoning Linux and Unix.

The ESP index is an Inverted index, using tf-idf as its main relevancy formula. Relevancy is tunable, and other factors in relevancy can be:
 * Static document boost
 * Term boost
 * Freshness boost
 * Proximity boost
 * Phrase boost
 * Boolean operator boost
 * GEO distance
 * Query based document boost
 * Fixed placement of top-10 per query

Licensing model
ESP has been licensed for base system, with additional license options being available for extra features (such as high availability, extra languages or alerting) as well as extra capacity (QPS or GB storage).

Microsoft may plan to align licensing to its other products.

ESP in MS Office Sharepoint
After Microsoft's acquisition of FAST in 2008, they started incorporating the ESP technology into Office Sharepoint. The first release with ESP included is expected in Sharepoint 2010. The ESP search will be an add-on for advanced customers.