Decomposing Enterprise Monoliths Without Rewrite Traps

Many engineering organizations attempt to escape monolithic friction by decomposing their entire system into dozens of microservices simultaneously. Without clear domain boundaries, this merely converts in-process function calls into fragile, high-latency HTTP or gRPC network requests with distributed transaction complexity.

Domain-Driven Boundary Identification

SazM approaches decomposition by profiling change frequency and operational contention:

1. High Friction / High Velocity: Services that deploy multiple times daily and suffer deployment queue bottlenecks (e.g. Lead Intake, User Onboarding).
2. High Contention / High Latency: Services that hold long-running database locks or interface with unstable third-party APIs.
3. Stable Core: Core accounting or ledger tables that change rarely and operate reliably within the monolith.

The Anti-Corruption Layer (ACL)

When modern services must read or write data shared with the legacy monolith, an Anti-Corruption Layer translates between domain models:

export function adaptLegacyUserData(legacyRow: Record<string, any>): CanonicalUser {
  return {
    id: legacyRow.vch_user_uuid || legacyRow.int_user_id.toString(),
    email: legacyRow.txt_email_address.trim().toLowerCase(),
    tier: legacyRow.int_status === 1 ? 'enterprise' : 'standard',
    createdAt: new Date(legacyRow.dt_created).toISOString(),
  };
}

By isolating the modern schema from legacy naming idiosyncrasies, the modern system remains clean and verifiable.

Event-Driven Decoupling via Transaction Outbox

Instead of cross-database transactions, publish domain events to an Outbox table:

BEGIN TRANSACTION;
UPDATE orders SET status = 'shipped' WHERE id = ?;
INSERT INTO domain_events (id, topic, payload, created_at)
VALUES (uuid(), 'order.shipped', json_object('order_id', ?), datetime('now'));
COMMIT;

A background worker polls the outbox and streams events to downstream services with guaranteed delivery.