Up to now, search engines/products required a separate solution to be integrated, developed and administered when extracting information, for example using a full-text search or recommendation function, In response to this inconvenience, PFI developed “Sedue” which provides full-text searching, recommendation and other functions in a single product.
Sedue stores data in tabular form similar to a relational database management system (RDBMS). By doing so, it can flexibly perform operations such as narrow, sort, and drill down on search results according to stored field values.
For example, if the dates when documents are published are stored, it is possible to extract search results of documents that were published after a certain date or year.
Sedue offers high performance and scalability using a cluster system which allows a large-scale and highly efficient search service to be easily configured and operated. Since there is no single point of failure (SPOF) in the search system, high availability is attained. In addition, documents can be added and deleted instantly.
In the above figure, documents are imported from several data sources (database, file repository, etc.) and the system is configured to search several machines. With this configuration, a search system with a QPS (Queries Per Second) of a few hundred can be configured against a data size of roughly 150 GB.
If there are multiple data sources, a search can be performed only on a particular data source or across all data sources. For example, if a website is operating numerous media sites, it is possible to set up a system where searches can be performed within certain media sites or across all media sites.
If performance, capacity and reliability are not a concern, it is possible to operate a search system on a single machine.