How to Migrate On-Premise Data to the Cloud in 2026
- Feb 24
- 9 min read
Quick Answer
On-premise to cloud migration in 2026 does not require custom code, months of developer time, or routing sensitive data through third-party vendor servers.
Compliance-focused enterprise IT teams use no-code cloud migration platforms with self-hosted architecture to move mission-critical data — IBM DB2 AS400, transactional ERPs, local data warehouses — to Snowflake, Redshift, and Azure SQL without sacrificing security, governance, or throughput. This guide covers the six architecture decisions that determine whether your migration succeeds safely at enterprise scale.
Why compliance-sensitive organizations need a different migration approach
Traditional on-premise to cloud migration introduces two problems that standard integration tools cannot solve simultaneously.
The first is security. Most cloud migration tools act as a man-in-the-middle — copying and storing your data on vendor servers during transit. For organizations in finance, healthcare, and government, this model violates data sovereignty mandates and exposes sensitive data to third-party infrastructure your security team has not approved.
The second is maintenance. Custom-coded ETL and ELT pipelines are slow to build, fragile by design, and require continuous developer attention. Every schema change in the source system breaks something downstream. Every API update creates a remediation project. Over time, the maintenance burden consumes the engineering capacity the organization needed for higher-value work.
No-code cloud migration with a self-hosted, customer-controlled architecture solves both problems. The migration engine runs inside your own security perimeter. Your data never touches vendor servers. And cloud migration automation handles schema discovery, performance tuning, and resumability without custom code.
Architecture decision 1: True data custody with self-hosted deployment
For compliance-focused teams, the primary requirement of any hybrid cloud migration is total data sovereignty. The migration engine must run entirely within your corporate security perimeter — on-premise or inside your virtual private cloud — with zero vendor data exposure at any stage.
The correct architecture establishes a secure hybrid data bridge with three properties.
Zero vendor data exposure. The migration application deploys as a standard Java Web Archive inside Apache Tomcat on a 64-bit Java runtime. Because it runs on your own infrastructure, your sensitive enterprise data never touches or passes through vendor servers at any point during the migration.
Outbound-only firewall footprint. The application server initiates only outbound HTTPS connections — port 443 — to source APIs and local JDBC writes. No inbound firewall ports need to be opened, which eliminates a major vector for external security threats and simplifies your security and compliance posture significantly.
Transit encryption. All communications between source systems, target cloud databases, and the licensing server use enterprise-grade TLS 1.2 encryption throughout the transfer.
This self-hosted deployment model gives your organization complete control over where data lives, how it moves, and who can access it — satisfying data governance requirements for GDPR, HIPAA, SOX, and national data sovereignty laws by architecture rather than by vendor assurance.
Architecture decision 2: Automated schema discovery to eliminate manual mapping
Manually translating legacy database schemas into cloud-compatible structures is one of the most time-consuming phases of any migration project. As source fields, tables, and custom objects change, manual mapping scripts break — causing schema drift, broken analytics, and unplanned remediation work.
Cloud migration automation makes manual schema mapping unnecessary. The migration platform dynamically queries your source systems' APIs and database catalogs to automate schema construction end-to-end.
On-the-fly schema generation automatically discovers all standard and custom objects, fields, and tables. It then builds the matching tables, columns, indexes, and primary keys directly in the cloud destination — Snowflake, AWS Redshift, Azure SQL, or another supported target — without any manual data modeling.
Reserved word collision resolution handles the naming conflicts that arise when source systems use different reserved naming conventions than the target database. The platform compares field and table names against the target database's reserved word list and automatically appends a suffix to conflicting names — preventing compilation errors that would otherwise require manual intervention.
Automatic column width expansion detects field-width expansions in source systems and widens the corresponding destination VARCHAR and NVARCHAR columns in real time. This prevents records from being rejected due to schema updates that occurred after the initial migration configuration — a common failure point in long-running migrations.
Architecture decision 3: High-performance tuning for enterprise data volumes
Migrating hundreds of millions of records to cloud platforms requires performance tuning that goes beyond default configuration. Without it, pipelines bottleneck at the destination, exhaust source system resources, or fail under memory pressure during heavy transfer runs.
Three tuning decisions determine whether a high-volume migration completes successfully.
Writer-to-finder thread ratio. The migration engine uses a thread pool that separates extraction threads — which query the source system — from writer threads — which load data into the cloud destination. Maintaining a three-to-one ratio of database writer threads to extraction threads is required to prevent the target cloud database from becoming the bottleneck during heavy write operations. This ratio is configurable through the platform's property files without writing code.
Dynamic time-slicing for large historical datasets. Large-scale queries covering wide date ranges on active transactional databases frequently trigger connection timeouts or API governor limits. The platform uses patented time-slicing technology to dynamically adjust query windows during extraction.
When a query returns more records than a configured threshold — and the interval is greater than 2 seconds — the platform automatically shortens the time interval and re-queries the source, breaking dense historical datasets into manageable chunks that complete reliably without timing out.
In-memory throttling. When network latency or destination write speed creates a backlog, the in-memory record pool can grow until it exhausts Java heap space and crashes the migration process. The platform's memory pool parameter caps the in-memory record buffer. When the limit is reached, extraction pauses automatically — allowing writer threads to catch up before additional records are loaded into memory. This prevents the JVM memory crashes that derail large migration runs.
Architecture decision 4: Checkpointed resumability for multi-day migrations
A primary concern during multi-day historical data migrations is pipeline resilience. Network disruptions, server reboots, and source system maintenance windows are operational realities. A migration that must restart from the beginning after any interruption wastes API budget, processing time, and engineering attention.
The correct approach uses checkpointed resumability — tracking migration job step metadata in a dedicated database table so any interrupted migration can resume from the exact point of the last successful transaction.
The platform maintains three runtime variables for every migration job: the starting date-time boundary of the data load, the timestamp when the current run initialized, and a unique audit run identifier. These variables update in the destination database only after each job step successfully completes.
When a migration resumes after an interruption, the platform reads the last logged transaction timestamp from the history table and continues from that exact checkpoint — with zero data gaps and zero record duplication. This resumability is automatic — no manual intervention, no re-run configuration, no risk of loading records twice.
Architecture decision 5: Handling non-incremental objects and complex tables
Incremental synchronization relies on source system update timestamps. But many core ERP and CRM administrative tables do not include modification timestamps, making incremental tracking impossible for those objects.
The platform automatically identifies non-incremental objects and applies a different sync strategy — truncating the target table and executing a full reload during each sync run. This ensures data consistency without requiring manual configuration for each non-incremental object type.
For NetSuite environments, this applies to 69 standard objects — including Subsidiary, Department, and Budget — that do not support timestamp-based incremental querying. The platform detects and handles these automatically as part of normal sync operation.
Saved searches as queryable tables extends this capability further. Enterprise teams can migrate pre-filtered source system saved searches into standard, queryable database tables. Special characters in search names are automatically replaced with underscores, and the destination table includes a record count tracking column that stores the exact count returned by the search on each run. This creates a queryable audit trail of historical dataset growth that supports both analytics and compliance reporting requirements.
Architecture decision 6: Governance and compliance controls for regulated migrations
A compliant hybrid cloud migration requires audit controls and historical change-tracking built into the migration architecture — not added as an afterthought after the data has already moved.
Versioned change auditing supports regulatory compliance by automatically maintaining a versioned history table alongside each primary replicated table. Before an update commits to the main table, the platform copies the existing row to the history table — creating a complete, ongoing audit trail of every record version. Because the history tables live on the customer's own database infrastructure, retention is limited only by the storage the customer provisions — with no platform-imposed ceiling.
Destination architectural governance determines where history tracking is available. For row-oriented target databases — Oracle, SQL Server, MySQL, MariaDB, PostgreSQL, DB2 — full history tracking is supported. For columnar cloud data warehouses — Snowflake, Amazon Redshift, Vertica — standard flat replication is fully supported, but native history tracking is not available due to the transactional overhead of columnar bulk loaders.
Organizations using Snowflake or Redshift as the primary destination should configure a separate row-oriented database as a metadata source to handle configuration tracking and history logging.
Technical specifications at a glance
System runtime: 64-bit Java 1.8 JRE — update 212 or later — Apache Tomcat 8.5 or 9.0, memory pool tuned to 2048/4096.
Connection security: Outbound HTTPS port 443 only, zero inbound open ports, TLS 1.2 encryption throughout.
Thread optimization: Three-to-one writer-to-finder thread ratio required to maximize cloud load and prevent destination deadlocks.
In-memory throttling: Record pool cap prevents JVM memory exhaustion during heavy transfer runs.
Schema drift control: Automatic column width expansion accommodates source schema changes in real time.
Resumability: Checkpoint tracking inside the migration history table — zero data gaps, zero duplication on restart.
Why Sesame Software is built for compliance-first cloud migration
Sesame Software has spent 23+ years helping compliance-focused enterprise IT teams move data from on-premise systems to cloud platforms — without routing that data through vendor infrastructure, without custom code, and without the security compromises that standard cloud migration tools introduce.
The self-hosted deployment model processes every migration inside the customer's own environment. Your data moves directly from on-premise sources to your cloud destination through pipelines running on your infrastructure. Sesame Software's servers are never in the data path — not during extraction, not during transformation, not during loading.
With 15 proprietary patents powering the replication engine, 20+ actively maintained connectors covering the legacy enterprise source systems that matter most — IBM DB2 AS400, Oracle, SQL Server, SAP, NetSuite, Salesforce — and cloud destinations including Snowflake, Redshift, Azure SQL, and Google Cloud, Sesame Software delivers the high-performance, governance-ready migration infrastructure that compliance-sensitive organizations require.
Predictable annual pricing based on connectors — no per-row charges, no consumption-based billing surprises as data volumes grow. Whether the migration is a one-time historical load or an ongoing hybrid cloud sync, the cost stays fixed.
Frequently asked questions
What is no-code cloud migration?
No-code cloud migration moves on-premise databases to cloud platforms using visual, configuration-driven tools — without custom ETL scripts, developer resources, or vendor-hosted processing. The migration platform handles schema discovery, extraction, transformation, and destination loading automatically. Enterprise IT teams configure the migration through a visual interface and the platform executes it — with no code written anywhere in the process.
How does self-hosted cloud migration satisfy compliance requirements?
Self-hosted cloud migration runs the migration engine inside the customer's own infrastructure — on-premise or in the customer's own cloud accounts. Because the platform processes data inside your environment, your sensitive enterprise data never passes through vendor-managed servers. This architecture satisfies GDPR data processor documentation requirements, HIPAA security perimeter obligations, and national data sovereignty laws by design rather than by vendor assurance.
What happens when a cloud migration is interrupted mid-run?
Checkpointed resumability allows any interrupted migration to restart from the exact point of the last successful transaction — not from the beginning. The platform tracks job step metadata including start timestamp, run ID, and completion status in a dedicated history table. On restart, it reads the last successful checkpoint and continues from there — with zero data gaps and zero record duplication.
How does the platform handle schema changes in source systems during migration?
Automated schema discovery detects field-width expansions, new fields, new tables, and modified data types in source systems and propagates those changes to the destination automatically. Column widths expand in real time when source field sizes increase. New fields create new destination columns. These changes apply without manual intervention or migration downtime — keeping the destination schema aligned with the source throughout the migration lifecycle.
What cloud destinations does Sesame Software support?
Sesame Software connects to all major cloud data warehouse and database destinations — Snowflake, AWS Redshift, Azure SQL, Google BigQuery, Oracle, PostgreSQL, SQL Server, MySQL, and others. The platform creates destination schemas automatically based on source structure, with no manual table creation or schema mapping required.
Does no-code cloud migration support regulatory audit requirements?
Yes. The platform maintains versioned history tables that create a complete audit trail of every record change throughout the migration — supporting the multi-year retention requirements of HIPAA, SOX, and GDPR. All pipeline operations are logged with timestamps, run IDs, and completion status in the migration history table — producing the audit documentation that compliance teams and regulators require.
