BigQuery Data Warehouse

Modernizing a Global IT Services Provider's Db2 Estate for Cloud-Native Analytics and AI

3 min read

Mar 5, 2026 1:44:53 PM

Legacy Db2 z/OS and LUW to a cloud-native analytics and AI platform

A global IT services organization across 30+ countries relied on IBM Db2 for z/OS and Db2 LUW for service delivery, billing, and customer management. With end-of-support deadlines approaching, MIPS costs climbing, and Db2 DBAs retiring, the organization engaged Pythian to modernize its entire Db2 estate.

45%

Reduction in mainframe MIPS costs

$3.1M

Annual infrastructure savings

99.99%

Uptime maintained during migration

Account profile

Industry: Information Technology & Services

Organization scale: Global enterprise, $2B+ revenue, 30+ countries

Tech stack:

IBM Db2 for z/OS
Db2 11.5 LUW with pureScale clustering
IBM DataStage and Cognos Analytics
JCL batch processing and COBOL applications
Google BigQuery, Looker, Airflow, dbt

Three decades of Db2 and only one year to modernize

End-of-support timelines, an aging DBA workforce, and customer demands for real-time reporting created pressures that could no longer be deferred.

EOS meets a shrinking DBA team

Db2 11.5 LUW end-of-support (April 2027) set a hard deadline. Two of three senior z/OS DBAs announced retirement within 12 months. The team lacked bench depth to maintain the estate without external support.

Dual-variant, undocumented estate

he z/OS mainframe powered billing and SLA enforcement, while Db2 11.5 LUW handled data warehousing. Over 400 SQL PL stored procedures, DataStage pipelines, and JCL batch jobs connected the two—none documented to modern standards. pureScale clustering added HA complexity with no cloud fallback.

38% of budget, zero real-time analytics

MIPS costs consumed 38% of the IT budget and climbed annually. Cognos batch reports delivered SLA data 24 hours late. Competitors with real-time dashboards were winning renewals. The organization had no viable path to predictive analytics or AI.

Variant-aware modernization in three phases

Db2 for z/OS and Db2 LUW are different platforms that share a name. Each required its own migration strategy and tooling. Pythian stabilized first, then migrated in phases that delivered value at each step.

Discovery phase

Pythian cataloged 18TB across both variants: 412 SQL PL stored procedures (37 requiring manual refactoring), 89 DataStage jobs, RCAC policies across 14 regulated datasets, and a pureScale cluster with undocumented failover configurations. The team mapped every downstream system, report, and application dependency.

Strategic architecture

The solution centered on an analytics-first migration with phased mainframe offload:

Analytics to BigQuery first

Pythian migrated analytics workloads to Google BigQuery, targeting the highest-value, lowest-risk workloads while core billing remained on z/OS during the transition.

HA replacement

Pythian replaced pureScale HA with BigQuery's multi-region resilience and Cloud SQL for PostgreSQL where ACID guarantees were required.

The governance layer

Unity Catalog replaced both Hive Metastore and Apache Atlas, while Kerberos/Ranger policies were remapped to cloud-native IAM with the fine-grained, row-level controls regulators required.

Implementation roadmap

Phase 1: Stabilize and reduce MIPS (months 1–3)

Pythian augmented the DBA team for 24/7 coverage while retiring DBAs transitioned knowledge. RUNSTATS optimization and locking contention fixes delivered an immediate 12% MIPS reduction before migration began.

Phase 2: Migrate analytics and rebuild pipelines (months 4–9)

Pythian refactored SQL PL using SQLines and BigQuery Migration Service, with 60% converting automatically and 40% requiring manual rewriting for Db2-specific extensions and MQT dependencies. The team replaced DataStage with Airflow and dbt, and rebuilt all 67 Cognos reports as interactive Looker dashboards. Pythian mapped RCAC security policies to BigQuery column-level controls and validated EBCDIC-to-UTF-8 conversion across all character sets. A CDC-based dual-run process ensured data integrity throughout the transition.

Phase 3: Cut over and deploy AI (months 10–14)

Pythian cut over billing and SLA data to BigQuery, using hash-based reconciliation to validate 100% of records. The team decommissioned the mainframe for analytics workloads, while transaction processing followed a separate timeline. Pythian deployed Vertex AI for predictive SLA breach detection and automated capacity planning, and continued 24/7 managed services across the new environment.

From cost center to competitive advantage

The modernization didn't just cut costs. With Pythian's support, it changed what the organization could offer its customers.

45% reduction in mainframe MIPS costs

Offloading analytics from z/OS to BigQuery and optimizing RUNSTATS eliminated nearly half of all mainframe compute charges.

$3.1M in annual savings with 11-month payback

The organization eliminated MIPS charges, Db2 Enterprise licensing, and DataStage/Cognos fees. Full cost recovery arrived three months ahead of projection.

Real-time SLA dashboards replace 24-hour batch reports

Query times dropped from minutes to seconds. All 67 customer-facing dashboards now update live, and Pythian refactored all 412 stored procedures with zero post-migration defects.

Predictive AI replaces reactive troubleshooting

Vertex AI now predicts SLA breaches 48 hours in advance. Pythian's managed services resolved the DBA skills gap, replacing retiring mainframe talent while the internal team upskilled on BigQuery.

IBM Db2 consulting services

Ready to solve your data challenges?

Speak with a Pythian IBM Db2 expert ->

On this page

Ready to unlock value from your data?

With Pythian, you can accomplish your data transformation goals and more.

Speak with Pythian consultants now →

Services