Quest for Requirements: Scrutinizing Advanced Search Queries for Cloud Services with Fuzzy Galois Lattices

Abstract

In software and requirements engineering, requirements elicitation is considered an essential step towards building successful systems. Despite extensive existing research in the field of distributed requirements engineering, the topic of requirements elicitation for cloud systems remains still uncovered. Cloud challenges (e.g., heterogeneous and globally distributed users, volatile requirements, frequent change requests) cannot always be satisfied by existing methods. We present a new approach for eliciting requirements for cloud services by analyzing advanced search queries. Our approach builds fuzzy Galois lattices for the terms that compose advanced search queries, thus enabling a thorough analysis of stored search data. This can support cloud providers in observing requirements clusters and new classes of cloud services, identifying the threshold for achieving satisfied consumers with a minimal set of requirements implemented, and thus designing novel solutions, based on market trends. Moreover, the Galois lattices approach enables large-scale consumers' involvement and ensures the elicitation of real requirements unobtrusively. Abstract-In software and requirements engineering, requirements elicitation is considered an essential step towards building successful systems. Despite extensive existing research in the field of distributed requirements engineering, the topic of requirements elicitation for cloud systems remains still uncovered. Cloud challenges (e.g., heterogeneous and globally distributed users, volatile requirements, frequent change requests) cannot always be satisfied by existing methods. We present a new approach for eliciting requirements for cloud services by analyzing advanced search queries. Our approach builds fuzzy Galois lattices for the terms that compose advanced search queries, thus enabling a thorough analysis of stored search data. This can support cloud providers in observing requirements clusters and new classes of cloud services, identifying the threshold for achieving satisfied consumers with a minimal set of requirements implemented, and thus designing novel solutions, based on market trends. Moreover, the Galois lattices approach enables large-scale consumers' involvement and ensures the elicitation of real requirements unobtrusively.

A Fuzzy Galois Lattices Approach to Requirements Elicitation for Cloud Services

“…However, we used an available small dataset containing real-world data [18] coming from one of the companies interviewed [17] as a starting point. Moreover, we generated the queries such that they define data storage cloud services with ten features, as described in Table 1.…”

“…This work is an extension of an existing paper published in the 2014 IEEE 10th World Congress on Services [18]. There, we introduced the preliminary idea of building fuzzy Galois lattices to support cloud providers in the requirements elicitation activity.…”

“…In this extension, we have three new contributions: 1. We enhance the preliminary algorithm introduced in [18], by improving its performance through a pre-processing technique that calculates frequencies of search queries. This is included as a functionality in the StakeCloud Tool we developed.…”

“…All the queries whose distances to the supremum are smaller than the breakingP oint value for a given sim are considered similar for the sim value specified. Each element of the set ends up in a cluster, as long as the distance between it and the supremum is smaller than the breaking point (lines [14][15][16][17][18]. Then, the dataset is updated (line 19) to include only those elements that did not become members of clusters or were left out as single unclustered elements, and the steps above are repeated until the dataset is empty.…”

“…The analysis can continue until the provider has obtained enough information to make a decision what service(s) are worth launching, e.g., by unbundling query (2) to analyze the queries it consists of, generating the graphs that visualize to what extent query (7) can be satisfied by the given infima, generating the graphs that show also the extent to which the features of the selected nodes are partially satisfied, etc. For another analysis example, please refer to [18]. This analysis is performed exclusively remotely, based on the input dataset.…”