Portal currently provides two search frameworks: search collections and search service. The one you use depends on whether you value performance or ease of maintenance.
Using search collections, the Portal Search Engine (PSE) which is a text search engine, crawls and indexes IBM Connections content and stores it in an Operating System files directory (also called collection). PSE internally uses the Seedlist Framework to index content. The indexing process happens at a regular interval updating the changes made in the Connections server to the collections directory.
The advantage of this approach is that it removes extra network traffic to the Connections server during the search process, making information retrieval fast. One disadvantage is that the crawler must run frequently to synchronize local content with all the latest content on the Connections server. It can do a better relevance ranking of search results, as search results (including Portal and Connections) are served by the Portal engine only.
Search service uses the Connections API which internally uses REST framework to search content directly from a Connections server. Connections and Portal integration uses the "Remote Content Server Search Service Type" (referred to as RCSS type), using the ATOM/REST APIs exposed by Connections. It is a federated approach in which the search is federated between Portal and Connections instances.
The advantage using this approach is that the maintenance is easy as there is no Collections to maintain. Since it is a live search, the disadvantage of Connections is overload during search process. Restriction:
Currently, you must implement the search integration of Connections Content Manager (CCM) documents using the Search Service approach. The search collections method is not supported at this time.