Database Migration for DSN Software

Challenge

DSN Software, Inc. provides cloud-based dental practice management for oral & maxillofacial surgery, periodontal, endodontic, and general dentistry practices, hosted on a significant AWS footprint of several dozen EC2 instances each running Microsoft SQL Server with FileStream. As the customer base grew past 95 dental practices to several hundred, the architecture's economics and operational profile became a strategic problem: SQL Server licensing across that many EC2 instances was expensive and getting worse; HIPAA-grade data isolation depended on careful, manual configuration per-customer; redundant infrastructure and operational overhead were tying up engineering time; and new-customer onboarding from competing platforms was slow because every new tenant required hand-rolled ETL. The technical pattern that worked at 10 customers wasn't going to work at 100+, and DSN's leadership needed to know whether AWS-native services could carry the business forward before committing to a full migration.

Solution

BUILDSTR led a two-phase engagement leveraging the AWS Migration Acceleration Program (MAP) in order to get DSN to a modernized database foundation ready to scale with the next chapter of the business.

Phase 1

BUILDSTR stood up Amazon RDS Custom for SQL Server with equivalent specs to the existing EC2 fleet, used AWS DMS to replicate data from EC2 source to RDS target, ran functional, performance, failover, and SLA-driven testing against critical jobs and reports, and delivered a recommendation to DSN leadership confirming RDS was a viable path forward. One of the biggest identified solution areas of need was the use of FileStream in SQL Server for storing unstructured data directly in the database. This design was originally the easiest path for the customer, but as they scaled, it became a drag on both operations and development.

Phase 2

BUILDSTR delivered the migration through four workstreams in parallel.

Workstream 1

WS1 was the "Mobilize" Phase of the AWS MAP approach, which landed the AWS landing zone (account structure + tagging, VPC/network, IAM, CloudTrail/CloudWatch, backup), deployed the migration factory, and ran the pilot migration with FileStream-to-S3 conversion and an automated ETL pipeline.

Workstream 2

WS2 was building the New Customer Onboarding Pipeline: a production-ready, event-driven ETL pipeline using Agentic AI for schema detection, validation, and anomaly triage, with **human-in-the-loop approval gates** at key checkpoints — designed first for the top two competing vendors and architected for future competitor sources.

Workstream 3

WS3 executed three migration waves of customers in 5-cluster groupings of 5–10 DBs each, with dual-write FileStream→S3 validation, per-customer UAT, and documented rollback procedures. Application-side changes to support the new storage model were deployed through DSN's existing containerized application, enabling zero-downtime cutover per wave.

Workstream 4

WS4 was BUILDSTR HyperCare, where instead of handing over the keys and wrapping the project once the new solution was in Production, BUILDSTR stays engaged and keenly focused on triaging any potential issues and making any necessary adjustments that can only arise once a solution is truly in Production . The target architecture moved DSN onto Amazon RDS for SQL Server Custom with S3 storage, with each customer's data kept isolated via per-customer EBS volume configuration — preserving the multi-tenant isolation HIPAA requires while modernizing the underlying platform.

Results

- Over 100 Dental Practices migrated across 3 waves from both the data center-hosted and EC2-hosted SQL Server with FileStream onto RDS SQL Server Custom + S3, with per-customer data isolation preserved.

- New Customer Onboarding Pipeline in production: event-driven, AI-driven schema detection and validation with human-in-the-loop approval gates, reducing onboarding time from WinOMS/OMSVision sources by over 40%.

- 60% reduction in monthly license, operations, and infrastructure spend after waves complete