Databases & In-Memory

Yes. The directive covers primary processing, the databases themselves, backups, IAM, encryption keys and audit logs. A replica or snapshot in a foreign region is still customer data outside Nigeria. Our designs keep primary, replicas, WAL archives and backups in two Nigerian data centres. This is general guidance, not legal advice; your compliance team should confirm the interpretation with the regulator.

PostgreSQL runs ledgers, payment hubs and risk engines in banks across Europe, and it has done so for over a decade. It offers ACID transactions, point-in-time recovery, row-level security and logical replication. The gap is usually operational, not functional, and that gap is exactly what our support contract closes.

It means a contract with response times, a named escalation path and engineers who fix the problem rather than point at a mailing list. We track upstream CVEs, test patches against your versions and take responsibility for the outcome. The software stays free of licence fees and the bank retains ownership of the platform.

Valkey is the Linux Foundation fork created after Redis changed its licence in 2024. It is API-compatible with Redis 7.2, keeps the BSD licence and is backed by AWS, Google and Oracle among others. For a bank that wants no licence risk and no dependency on a single vendor, Valkey is the safer choice, and migration from Redis is a restart with the same data files.

Apache Geode is the open source core; Tanzu GemFire adds commercial features such as WAN replication tooling, management console and vendor support. If you need multi-site replication between two Nigerian data centres and a vendor-backed SLA, GemFire is the practical choice. For a single-site authorisation cache, Geode with our support is sufficient.

Critical CVEs are patched or mitigated within 72 hours, high within 14 days. Minor upgrades are applied as rolling restarts through Patroni with no application downtime. Major upgrades are planned exercises using logical replication, rehearsed in staging and executed with a rollback path.

Yes, and many of our deployments do, using operators such as CloudNativePG or Zalando for PostgreSQL and the Valkey and GemFire operators. Kubernetes adds self-healing and consistent provisioning. For the largest ledgers we still recommend dedicated nodes with local NVMe storage, which the operators also support.

For a typical estate of ten to twenty databases, assessment and build take six to seven weeks and the migration waves another four to eight, depending on data volume and application testing capacity. Each cutover itself is a matter of minutes thanks to logical replication.