Information systems have moved from centralized systems to client-server environments. Our DLs therefore may follow that same path; scalability makes it almost obligatory.

Decentralized: Petaplex

The Intelligence Community has funded Knowledge Systems Inc. to explore a highly decentralized approach to digital libraries. There may be a hierarchical organization, with clusters of servers.

Decentralized: Proxy Servers

The WWW has an arbitrary number of clients and servers. There is no good way to scale servers to handle sudden growth in usage. Caching in the network, i.e., having proxy servers, reduces network and server loads. Caching can be done in the client, in a proxy server, in a super-proxy server, etc.

Decentralized: Hyper-G

The Hyper-G system deals with scalability by restricting clients to generally connect to a local server. Local servers cache data, and access other Hyper-G servers as needed to get desired information. Local servers index and search full-text documents, and maintain their part of the distributed link database.

Centralized/Decentralized: NCSTRL

The Networked Computer Science Technical Report Library (NCSTRL) accomodates its Lite sites with a central server (and a backup server). Standard sites provide full service, supporting local search. Thus, NCSTRL supports a combination of centralized and distributed searching.

Related Works