The MEDLINE database, an online archival of medical research publications, is extensively used by physicians, clinicians and other medical researchers for literature search and information on prior case studies. The size of MEDLINE database is over 40GB. The PubMed is an implementation of MEDLINE and provides basic search capabilities freely via the Internet. The initial search often tends to return excessive irrelevant pointers; so researchers often need to conduct additional searches to refine the search data. In this paper, we will investigate a parallel architecture for searching the MEDLINE database, which is unique compared to the existing MEDLINE implementations. The proposed architecture will implement MEDLINE on a cluster of Personal Computers (PCs). This architecture (i) allows for refinement of searches, (ii) reduces the search time for the users, and (iii) increases the availability of the system. We present design guidelines to select the cluster configuration and an analytical model to predict performance improvements.