Optimizing the Enterprise Search

NoSQL solutions are used for search in large data volumes of unstructured data, mostly for the search of the public data. The problem is how to use the technology for effective search of enterprise data where users access rights apply. The search mechanism has to be organized in a way to work fast on a large volume of data to return only the results a user has access to. An objective of our research is to create such a search mechanism. We evaluate several types and brands of NoSQL databases and select the most appropriate for the solution of the problem on hand. We use a polyglot persistence model that advocates to use several technologies inside a solution to profit from the strength of each. The solution we created uses an SQL database for the input and maintenance of the current data and a NoSQL Elasticsearch database for the search and retrieval of data. We created a methodology as well to evaluate the performance of the solution (and a search mechanism in particular). The evaluation results on a 1 million document database show that the data requests of a flow generated by 2000 user base are processed in 0.14 seconds in average. The evaluation tests are performed on a one node Elasticsearch database on a commodity server hence the proposed solution should be regarded as an effective way to handle search of enterprise data.