Source-linked AI summary
Advanced Data Processing in the Business Network System
Daniel Ritter
TL;DR
Enterprises struggle to reconstruct and query Business Networks because relevant data is fragmented across heterogeneous systems and processes. The paper presents BNS, a provenance-aware platform that discovers, transforms, infers, queries, and updates network views; it reports a reference implementation applied to real-world enterprise landscapes, while identifying provenance redeployment and database performance as areas for further research.
Problem
Enterprises need visibility into Business Networks, but their relevant data is hidden across heterogeneous enterprise landscapes, making basic network questions difficult to answer.
Method
BNS combines data discovery, network inference, query and update processing, and provenance tracking in a continuously running architecture for reconstructing interconnected network perspectives.
Results
The reference implementation provides end-to-end network query, traversal, and update processing and was applied to several real-world enterprise landscapes.
Takeaways & Limitations
BNS constitutes a holistic network data management platform spanning information retrieval, network mining, inference, querying, traversal, and provenance-aware source-data loading and enhancement.
Takeaways & Limitations
Further research is required for redeploying user information to source systems, and database selection may limit performance improvements.
Abstract
from arXiv · showhide
The discovery, representation and reconstruction of Business Networks (BN) from Network Mining (NM) raw data is a difficult problem for enterprises. This is due to huge amounts of e.g. complex business processes within and across enterprise boundaries, heterogeneous technology stacks, and fragmented data. To remain competitive, visibility into the enterprise and partner networks on different, interrelated abstraction levels is desirable. We show the query and data processing capabilities of a novel data discovery, mining and network inference system, called Business Network System (BNS) that reconstructs the BN--integration and business process networks - from raw data, hidden in the enterprises' landscapes. The paper covers both the foundation and the key data processing characteristics features of BNS, including its underlying technologies, its overall system architecture, and data provenance approach.
I. INTRODUCTION
The paper addresses the difficulty of discovering and understanding Business Networks from heterogeneous, fragmented enterprise data. It presents BNS as a system for data processing, provenance, and continuous network querying across real-world enterprise landscapes.
- I. INTRODUCTION: Business Networks connect intra- and inter-enterprise participants through technical, social, and business relations, but their data is difficult to understand across heterogeneous systems.
- I. INTRODUCTION: BNS combines network discovery, inference, query processing, and provenance to reconstruct Business Networks from hidden enterprise data.The system implements continuous, scalable end-to-end network query, traversal, and update processing.
- I. INTRODUCTION: The paper contributes requirements for BNS data processing, a provenance approach, and a system applied to several real-world enterprise landscapes.
- I. INTRODUCTION: The paper proceeds from theoretical Network Mining and Business Network Management foundations to BNS requirements, architecture, processing, related work, and future work.
II. THE BUSINESS NETWORK SYSTEM
BNS derives operational requirements from Business Network and Network Mining foundations, then organizes discovery, inference, provenance, querying, updating, and visualization into a scalable system. Its architecture moves source data through validation and reconstruction while preserving links to origins.
- II. THE BUSINESS NETWORK SYSTEM: Network Mining discovers enterprise data that BNS transforms into interrelated business-process, integration, and social perspectives.
- II. THE BUSINESS NETWORK SYSTEM: BNS must support remote, scalable querying, traversal, and full-text search across interconnected Business Network perspectives.The requirements also include standard exchange and visualization formats, user enrichment and enhancement, and provenance across transformations.
- II. THE BUSINESS NETWORK SYSTEM: Source data is copied through protocol adapters, stored as raw data linked to its origin, and processed in the cloud for cross-domain network reconstruction.
- II. THE BUSINESS NETWORK SYSTEM: Inference produces network perspectives for visualization, simulation, and analytics, while client updates and provenance preserve links from queries back to source models.
A. Query Processing Overview
BNS separates inbound data handling, network inference, and client access so validated raw data can continuously produce queryable network views. Its APIs support search, traversal, visualization, enrichment, and application development.
- A. Query Processing Overview: Validated uploads enter a raw-data knowledge base, decoupled inference programs reconstruct network data, and automatically updated indices support client queries and traversal.Client requests operate on the current Business Network independently of inbound processing and inference.
- A. Query Processing Overview: Java/OSGi and HTTP/JSON interfaces let users build interfaces and applications for enriching, enhancing, analyzing, optimizing, and monitoring computed networks.
- A. Query Processing Overview: The architecture exposes integration-network visualizations and conversation details as views of the reconstructed network.
- A. Query Processing Overview: The client API supports keyword and field-specific searches, type restrictions, and traversal from participant systems to connected metadata such as hosts and locations.
- A. Query Processing Overview: Decoupled query and traversal components, together with model-centric index generation, process requests within short time even on larger networks.
B. Update Processing in BNS
BNS handles source loading, enrichment, and enhancement through provenance-aware transformations that preserve identity and support updates across raw and inferred network data. Aggregation remains a central design challenge, with trade-offs between preservation, simplicity, and user experience.
- B. Update Processing in BNS: Source records are merged by origin and local identifier, while composed keys prevent collisions across sources during raw-data ingestion.Records with the same origin and identifier update one another; records differing by origin or key remain distinct.
- B. Update Processing in BNS: Black-box transformations map one raw record to one Business Network record with direct lineage, while generated keys and mappings connect source and network identifiers.
- B. Update Processing in BNS: Aggregation can destroy source records into a surrogate, preserve equivalent records alongside a surrogate, or expose equivalence for applications to merge.These alternatives differ in how much source information and merge responsibility remain available to users and applications.
- B. Update Processing in BNS: Without preserving equivalent objects, later loads can leave Business Network objects less informative or unavailable, and re-evaluation can only be optimized rather than avoided.
- B. Update Processing in BNS: Enrichment adds labels or groups without modifying source data, whereas enhancement modifies, adds, or deletes computed instances and may redeploy changes to source systems.
- B. Update Processing in BNS: User enhancements require sufficient lineage and source-specific information for redeployment, while the inference system passes that information through without further domain-specific checking.
IV. RELATED WORK
Related work positions BNS alongside process mining, linked data, IT service management, virtualization, and software-ecosystem research. These areas provide complementary methods or models but do not fully cover enterprise-wide network inference and management.
- Process Mining supports business-process discovery from logs but does not use network models or inference, making it complementary to Business Network Management.
- IT service management and virtualization research could inform BNS when physical entities are added to its meta-model.
- Linked-data research has largely neglected intra-enterprise linked data and focused mainly on RDF-based approaches.
- Software-ecosystem research models interconnected enterprises within software development, so its modeling and governance approaches complement rather than overlap with BNS.
V. DISCUSSIONS AND FUTURE WORK
The paper presents a reference implementation that extends Business Network foundations with client access and data provenance for enterprise-ready network data management. It identifies unresolved provenance work and database-selection opportunities for performance improvement.
- BNS integrates information retrieval, network mining, inference, querying, traversal, continuous source loading, and user enhancements in one network data management platform.
- Data provenance supports tracing network data from original source artifacts through client queries to visualized computed-network instances.
- Further research is required for redeploying user information to source systems, while database selection may improve performance, including through SAP HANA.