GIFRÖST / Compare

Choose the connection model that fits the environment.

Compare can connect to source and target databases directly or through local agents. The right choice depends on table size, network topology, encoding differences, and how much execution work you want to keep close to each database.

Two connection modelsUse direct access for simpler setups or agents for heavier, more distributed workloads.

Heterogeneous-readySource and target can be different database engines inside the same comparison workflow.

Performance trade-offDirect is easier to start with, while agents reduce network pressure and sorting issues.

Connection models

This step defines which database instance acts as the trusted source and which becomes the target for validation. In practice, the key decision is whether Compare should read data centrally or delegate more work to agents placed closer to each database.

Lower setup effort

Direct Database Connection

Compare connects straight to the database endpoints and executes the comparison flow from the main environment, without deploying local agents near source and target.

Best when

You need the quickest onboarding path and the scope is moderate in size and complexity.

Strengths

Simple rollout, fewer moving parts, and no agent lifecycle to manage.

Watch out for

Higher network transfer volume and possible sorting mismatches across engines.

Fast setupNo agent installSmaller scopes

Higher execution efficiency

Agent-Based Connection

Local agents are placed near the source and target databases to handle sorting, hashing, and part of the comparison workload closer to where the data lives.

Best when

You compare large datasets, operate across environments, or expect encoding and sorting differences.

Strengths

Lower network overhead, better scalability, and more control over where heavy processing happens.

Watch out for

Extra deployment work, agent maintenance, and local infrastructure requirements.

Large volumesLower trafficDistributed setups

Detailed guides

If the execution model is already clear, jump straight to the dedicated setup guide:

• Direct Database Connections
• Agent-Based Connections

Quick selection guide

01

Need the fastest path to a working comparison?

Start with Direct Database Connection when setup speed matters more than deep execution tuning.

02

Working with large tables or remote environments?

Choose Agent-Based Connection when you want hashing and sorting closer to the data source.

03

Databases differ in encoding or ordering behavior?

Prefer Agent-Based Connection to reduce the risk of mismatches caused by engine-specific sorting behavior.

Operational recommendation

For small and well-controlled environments, direct access is usually enough. For production comparisons across larger datasets or heterogeneous platforms, agent-based mode is the safer long-term default.

Connection architecture

The upper model keeps processing closer to the databases by using agents. The lower model keeps the setup simpler by connecting directly from the main Compare environment.

Supported databases

Compare supports a set of enterprise-grade database engines commonly used in migration, replication, and validation projects.

Oracle

Supported versions

21c19c18c12.2

PostgreSQL

Supported versions

8.4+

Informix

Supported versions

12.1014.10

IBM Db2

Supported versions

11.111.5x12.1

MS SQL Server

Supported versions

2014+

Testing connections

Once the source and target are selected, Compare validates connectivity automatically. The goal is to confirm that both sides are reachable and properly authenticated before the comparison moves to execution.

01

Select source and target

Choose the reference database and the environment that should be validated against it.

02

Check the status indicator

Green means the connection succeeded. Red indicates a network, credential, or configuration issue.

03

Retest manually when needed

Use Test Connection after adjusting credentials, host settings, or firewall rules.

04

Proceed only with two valid endpoints

Both connections should be healthy before moving to the next configuration step.