Enterprise AI Digest #35 - PODCAST on Spotify and Apple
Topics:
ERP: Business Central - Components
AI: Securing AI Models in Microsoft Azure
Data: Microsoft Fabric Databases
Business Central - Components
Microsoft Dynamics 365 Business Central (BC) Online is a cloud-based ERP solution designed for small and medium-sized businesses. Its architecture ensures high availability, scalability, and security by leveraging Microsoft Azure’s infrastructure. Below are the key components that drive Business Central Online’s performance:
Compute
Business Central Online runs on Microsoft Azure Virtual Machines (VMs), dynamically allocated based on system load.
Uses Azure Kubernetes Service (AKS) for containerized deployment, ensuring auto-scaling and resilience.
Compute power is optimized with multi-tenant hosting, allowing seamless performance without dedicated hardware.
Database
Azure SQL Database is the backbone, providing high availability and disaster recovery.
Uses geo-replication to ensure data redundancy across multiple regions.
Elastic pools manage resource allocation efficiently, preventing performance bottlenecks for tenants.
Storage
Azure Blob Storage is used to store binary large objects (BLOBs) such as reports, files, and attachments.
Dataverse integration allows structured and unstructured data exchange between Business Central and Power Platform.
Data is automatically backed up using Azure Backup & Recovery, ensuring minimal data loss risk.
Performance & Security
Auto-scaling enables handling of seasonal spikes in demand.
Encryption at Rest and In Transit ensures data security.
Multi-region availability ensures 99.9% uptime as per Microsoft's SLA.
Securing AI Models in Microsoft Azure
Microsoft’s Azure AI Foundry and Azure OpenAI Service ensure data privacy, model security, and zero-trust architecture to safeguard organizations against potential threats.
Zero-Trust Architecture – AI models run in isolated Azure Virtual Machines (VMs), preventing unauthorized access to Microsoft’s infrastructure.
Strict Data Privacy – Customer data is not used to train shared models, and logs are not shared with model providers.
Comprehensive Model Security – Before deployment, AI models undergo rigorous security checks:
Malware analysis – Scans for embedded malicious code.
Vulnerability assessment – Identifies potential CVEs and zero-day threats.
Backdoor detection – Flags unauthorized code execution risks.
Model integrity checks – Detects tampering or corruption.
Microsoft Fabric Databases
Microsoft Fabric provides multiple data storage options tailored for different use cases. This comparison highlights key differences between Lakehouse, Warehouse, Eventhouse, SQL Database, and Power BI Datamart, focusing on data type, volume, developer persona, operations, and security.
Key Fabric Data Stores: Overview
Lakehouse – Best suited for big data processing, supports structured, semi-structured, and unstructured data. Developers use Spark SQL, PySpark, and R for data engineering.
Warehouse – Designed for enterprise-scale relational data, supports SQL-based analytics with ACID-compliant transactions.
Eventhouse – Ideal for real-time event processing, supporting KQL and T-SQL for analytics.
Fabric SQL Database – A structured data store for AI and app developers, providing T-SQL support with advanced security controls.
Power BI Datamart – A self-service analytics database that integrates with Power BI, designed for data analysts and business users.