CVE commitment
Enterprise-grade MPP analytics, on-prem in Nigeria, with no cost per query.
Analytics & Data Warehouse
Apache Cloudberry, Greenplum, WarehousePG and SynxDB for SQL at scale; Apache Spark, Iceberg and Hadoop for the lake beneath them. We build the warehouse the regulator can inspect and the analysts actually use, on hardware you own, with a support contract instead of a monthly bill that grows with every query.
- Apache Cloudberry
- Greenplum
- Apache Spark
- Apache Iceberg
- Apache Hadoop
The technologies we cover
Products in this family
Apache Cloudberry
Apache-governed MPP data warehouse derived from Greenplum, on a modern PostgreSQL core with vectorised execution.
Bare metalVirtual machinesKubernetesPrivate cloudAir-gappedOfficial support

Greenplum
The reference MPP warehouse for banks that already run it, supported through its end of life and beyond.
Bare metalVirtual machinesPrivate cloudAir-gappedOfficial support

WarehousePG
Community-maintained open source fork of Greenplum 6 and 7 for estates that need continuity without a licence.
Bare metalVirtual machinesPrivate cloudAir-gappedOfficial support
- SynxDB
SynxDB
Commercially supported distribution based on Greenplum with a migration path from Greenplum 6.
Bare metalVirtual machinesPrivate cloudOfficial support
Apache Spark
Distributed batch and ML processing over Iceberg tables, run on Kubernetes or YARN.
Virtual machinesKubernetesPrivate cloudAir-gappedOfficial support
Apache Iceberg
Open table format with ACID transactions, schema evolution and time travel over S3-compatible storage.
KubernetesPrivate cloudAir-gappedOfficial support
Apache Hadoop
HDFS and YARN for deep historical storage and legacy batch workloads already written for the ecosystem.
Bare metalVirtual machinesPrivate cloudAir-gappedOfficial support
Why it matters for data localisation
Where this family meets the CBN directive
A data warehouse holds the most complete copy of a bank's customer and transaction history, and it is often the copy that has quietly moved abroad. Snowflake, BigQuery and Redshift accounts created for analytics teams process payment data in foreign regions, keep their own backups and time-travel snapshots there, and manage access through the cloud provider's identity service. Each of those items (primary processing, databases, backups, IAM, encryption keys and audit logs) must be inside Nigeria by 1 January 2027 with no dependency on a foreign cloud provider. Cloudberry, Greenplum, WarehousePG and SynxDB are MPP warehouses that run on servers in a Nigerian data centre and answer the same SQL. Iceberg on S3-compatible object storage gives the lake open formats, and Spark and Hadoop cover batch processing and deep history. This family delivers the evidence the inspector expects: the location of every table and backup, keys held in a local KMS, access governed by the bank's directory, and query audit logs retained in-country. It also removes the per-query bill that made analysts ration their questions.
The journey with NuxFamily
- 01Assess
- 02Design
- 03Build
- 04Migrate
- 05Operate
- 06Evolve
Every family is delivered through the same six-stage journey, with official 24×7 support and knowledge transfer built in.
Our expertise
Credentials, not adjectives
We built and operated Greenplum warehouses for European banks and insurers when Greenplum was the standard for regulatory and risk analytics, and we have followed the technology through Greenplum 6 and 7, the WarehousePG fork and the Apache Cloudberry project. Our Spark and Hadoop experience comes from retail and industrial data lakes measured in petabytes. We know what a month-end close looks like when the warehouse is under pressure, and we design so that it finishes on time.
Sectors
- Banking
- Insurance
- Retail
- Industry
15+
Years with these technologies
35+
Production deployments
1.5 PB Greenplum warehouse across 96 segment hosts
Largest scale delivered
Official vendor support
Support tiers for this family
Essential
- Coverage
- 8×5, Nigeria business hours
- P1 response
- 4 h
- Corrective support on Cloudberry, Greenplum, WarehousePG, SynxDB, Spark, Iceberg and Hadoop
- Security patches and CVE advisories for your installed versions
- Knowledge base, tuning guides and ticket portal
Business
- Coverage
- 24×7
- P1 response
- 1 h
- Everything in Essential
- Quarterly health checks: segment balance, skew, bloat and catalogue statistics
- Version and patch lifecycle management for warehouse and Spark
- Backup, restore and mirror failover reviews
Mission Critical
Most chosen- Coverage
- 24×7 with named engineer
- P1 response
- 15 min
- Everything in Business
- Designated engineer who knows your schemas, workloads and SLAs
- Architecture review, capacity planning and query tuning for month-end
- Support for major upgrades and Greenplum to Cloudberry migration
- CBN audit accompaniment with evidence of data residency
Version policy
Response times and tier names are indicative and confirmed contractually.
Use cases
How it is used in a regulated bank
Use case 01
Regulatory reporting to the CBN
Prudential, credit-risk and payment returns to the CBN are assembled from the core banking system, the card platform and treasury, often through spreadsheets and manual extracts. Late or inconsistent figures carry a regulatory cost. A governed pipeline produces every report from the same tables, with lineage from each field back to its source.
Technologies
- Apache Cloudberry
- Apache Iceberg
- Apache Spark
- OpenMetadata
- Airflow
Expected outcome
Every return is reproducible from a named snapshot, and every figure can be explained to an inspector by following its lineage. The finance team stops rebuilding the same numbers in spreadsheets.
Metric: Reporting close reduced from days to under 4 hours after end of day; 100 % of report fields with documented lineage
- 1Extract from core and cards. Change data capture and end-of-day extracts land in Iceberg tables with the source timestamp preserved, so every figure can be traced to a point in time.
- 2Land in bronze tables. Raw records are stored unmodified in Iceberg on S3-compatible storage in Nigeria, versioned by snapshot for reproducibility.
- 3Conform and validate. Spark jobs apply the bank's data quality rules (completeness, referential integrity, balance checks) and quarantine failures for review.
- 4Load the warehouse. Conformed data is loaded into Cloudberry through external tables, partitioned by reporting period and distributed by account key.
- 5Compute regulatory measures. SQL models compute the return line items with the CBN taxonomy encoded as versioned reference tables, so a rule change is a data change with history.
- 6Publish with lineage. The report is generated, signed off in the workflow tool and registered in OpenMetadata with field-level lineage to source tables.
- 7Archive evidence. The submitted file, the input snapshots and the query log are archived immutably in object storage for the retention period.
Use case 02
Credit-risk analytics over years of history
Use case 03
Migration from a cloud data warehouse to Greenplum or Cloudberry on-prem
Reference architecture
What a compliant deployment looks like
Sources
Lake and table format
Processing and warehouse
Consumption
Migration path
From where you are to a compliant platform
01
3-4 weeksAssess
Activities
- Inventory schemas, tables, queries, scheduled jobs and BI connections in the current warehouse
- Classify datasets by CBN scope and by consumers, and record twelve months of cloud spend
- Profile query patterns to choose distribution keys and partitioning
- Size the target cluster and object storage, including growth for three years
02
4-6 weeksDesign and build
Activities
- Deploy Cloudberry or Greenplum with mirrored segments, and object storage with erasure coding
- Install Spark on Kubernetes, the Iceberg catalogue, Airflow and OpenMetadata
- Configure encryption, LDAP or Keycloak integration, resource groups and query audit logging
- Benchmark the cluster against representative queries before migration starts
03
6-10 weeks, per data domainMigrate
Activities
- Export domains to Iceberg, load the warehouse and set up incremental synchronisation
- Port views, procedures and pipelines; run automated dialect tests
- Dual-run key reports for two cycles and reconcile results
- Repoint BI tools domain by domain and decommission cloud objects with documented deletion
04
OngoingOperate
Activities
- 24×7 support with CVE patching by severity
- Quarterly reviews of skew, bloat, statistics and resource-group configuration
- Month-end readiness checks and capacity forecasting
05
From month 6Evolve
Activities
- Move Greenplum estates to Cloudberry or WarehousePG on a planned schedule
- Extend the lakehouse to feature stores and on-prem AI use cases
- Hand over operations with a paired-operation quarter and documented runbooks
FAQ
Questions architects ask us
If it holds customer or payment data, in our reading it does. The warehouse processes that data, keeps backups and time-travel copies, and controls access through the cloud provider's identity service, all outside Nigeria. The directive names each of those elements. This is general guidance, not legal advice; confirm the interpretation with your compliance team.
They share the same architecture and SQL. Greenplum remains the choice if you already run it and need continuity. Apache Cloudberry is the open source successor under Apache governance with a newer PostgreSQL core, and it is our default for new deployments. WarehousePG is the community fork for teams that want a drop-in open source continuation of Greenplum, and SynxDB is the commercially supported route. We support all four and help you choose on the basis of your workload and upgrade timeline.
For the steady, predictable workloads of a bank (daily loads, month-end closes, regulatory returns), a correctly sized MPP cluster on local NVMe storage matches or beats a cloud warehouse, and the cost does not rise with usage. Where public cloud wins is elastic burst for rare, very large jobs; we design headroom for month-end rather than for hypothetical peaks.
Iceberg gives the bank an open, engine-independent copy of every dataset with ACID guarantees, schema evolution and time travel. Spark, the warehouse and data-science tools all read the same tables, so a change of engine later is not a second migration. It also provides the immutable snapshots that make regulatory reports reproducible.
Corrective support with contractual response times on the warehouse engine, Spark, Iceberg and Hadoop, CVE patching by severity, version lifecycle management and, at the higher tiers, a named engineer, quarterly health checks, query tuning and upgrade support, including the move from Greenplum to Cloudberry.
For a warehouse of ten to thirty terabytes with a few hundred pipelines, assessment and build take seven to ten weeks and the domain-by-domain migration another six to ten. Dual running for two reporting cycles per domain is included, because the comparison of results is what gives the finance team confidence to switch.
Yes. None of the components requires internet access at runtime. Images, packages and patches are delivered through an internal signed registry, and we run the same tooling for our own air-gapped deployments.
Only if you already have it or need very cheap deep storage for history that is rarely queried. New deployments usually use S3-compatible object storage with Iceberg instead. We support HDFS estates and help move them to object storage when it makes sense.
Over two decades
Built by the team behind the platforms of Santander, ING, Bankinter, Mapfre and Inditex
More than twenty years designing, building and operating private clouds for institutions that cannot afford to fail, and a delivery model where we stay with you from assessment to operation.
See our track record20+
Years building private clouds
40+
Private clouds delivered
Talk to an architect about this family
Tell us where you are today and we will come back with a first view of the target architecture and the migration path.