Rosatom moves consolidated reporting to the KHD 2.0 platform

On 22 July 2026, Arenadata reported the completion of Rosatom’s move to the KHD 2.0 platform for consolidated financial reporting. The system brings together data from more than 400 entities, and almost 500 users took part in testing designed to resemble a real reporting close.

The project is useful as an example of a critical enterprise migration where replacing one database is not enough. Data models, analytical forms, integrations, quality controls and operational procedures all need to move together.

What KHD 2.0 contains

According to Arenadata, the analytical layer uses the Forsight platform. Arenadata DB and QuickMarts provide storage and processing, while the Atom.Most integration bus connects source systems. The project also includes data-quality controls and integration monitoring.

Development started in January 2024, and production operation began in February 2026. The team completed a monthly close in March and a quarterly close in April before handing the system to regular support in June.

The publication reports a 30-fold improvement in form loading, a six-fold improvement in loading source data and an 11-fold improvement in producing a trial balance. These figures are reported by the project participants, not by an independent benchmark, and should be read in the context of the previous architecture and this particular data set.

What matters in this type of migration

Financial reporting cannot be judged by query speed alone. A production result requires:

• reconciliation of balances and movements between the old and new systems;
• traceability from a reported figure to source records and transformation rules;
• governed reference data and named data-quality owners;
• repeatable ingestion without duplication after failures;
• access separation and action logging;
• backup, recovery and consistency checks;
• reporting-close tests under peak load;
• clear ownership of integrations after launch.

The staged rollout is particularly relevant: establish technical readiness, run monthly and quarterly closes, and only then transfer the platform to regular support. It exposes the system to different volumes and calendar conditions before the project team steps back.

Questions for another enterprise

Before selecting a stack, define control data sets and acceptable reconciliation tolerances. Performance should be measured across representative queries, batch ingestion and concurrent users. Each integration needs an owner, a format agreement, retry rules and an error-notification path.

Recovery deserves a separate design. A database backup is insufficient if analytical models, reference data, integration-bus configuration and secrets cannot be restored consistently. Our guide to backup and resilience covers these tests, while API integration architecture explains reliable data exchange.

Practical conclusion

KHD 2.0 demonstrates a large reporting migration approached as a complete system change rather than a product swap. The number of source organisations, rehearsal of real closing cycles and transfer into normal support are the most meaningful public signals.

Another organisation should still validate its own models, workload and operating process. The published acceleration factors are useful context, but an investment decision needs reproducible tests and verified reporting accuracy.

Source: the official Arenadata announcement dated 22 July 2026.