Services Oriented Architecture (SOA) is becoming a popular style for building complex systems-of-systems that allow businesses to work together across organizational boundaries. However concerns have been raised about the comprehensibility and maintainability of SOA composite applications. Integrating and deploying SOA applications requires artifacts in a variety of web-based languages (WSDL, XSD, BPEL, etc.) often produced by code-generation tools. It becomes difficult for a human to discover and understand the dependencies between these artifacts in an existing system. In this paper, we describe ongoing research on using search techniques to facilitate SOA maintenance by allowing users to query collections of artifacts making up a SOA composite application. The main focus in this paper is a case study using our prototype search tool SOAMiner to identify a set of abstractions that extract useful and critical information for maintainers, thereby bridging the heterogeneity of SOA artifacts while opportunistically exploiting their structure. Results of the study indicate that the highest priority abstractions for SOA are datatype summaries, service invocation (calling) relationships, and data usage relationships.