Welcome
/
Business Cases
/
Migrating a data platform to the cloud: an analytical warehouse on GCP to steer compute infrastructure
Manufacturing
Architecture & Engineering
Data Platform Migration

Migrating a data platform to the cloud: an analytical warehouse on GCP to steer compute infrastructure

For an international high-tech industrial group, we migrated the platform that tracks the usage of its virtual machines to an analytical warehouse on Google Cloud. In seven months, platform costs fell by 40% and processing speed rose by 30% in production.

In platform costs

-40%

In processing speed in production

+30%

From audit to go-live

7 months

Usage dimensions structured: CPU, GPU and users

3

Summary

Title

At a glance

Sector

Manufacturing

Sub-sector

High tech

Offer

Expertise

Engine

Architecture & Engineering

Product

Data Platform Migration

Duration

7 months

Discuss a similar project

Challenge

Infrastructure usage data too poorly structured to forecast, ensure reliability and govern.

The group runs a large fleet of virtual machines. Their usage, in CPU, GPU and number of users, was collected on a legacy platform where data remained poorly structured.

This situation weighed on three areas.

  • Forecasting. Anticipating infrastructure needs required usage data that remains usable over time.
  • Data quality. Without a common structure, usage figures were hard to make reliable.
  • Governance. The data remained hard to document and control.

The need: a modern, structured data platform with higher performance and controlled costs.

Solution

Audit the existing platform, design the target, then migrate.

We led the migration from audit to production. Three principles guided the work.

  • Start from what exists. An audit of the existing Splunk platform set the scope, the uses to keep and the limits to remove.
  • An architecture built for analytics. Usage data moves into an analytical warehouse on BigQuery, structured for forecasting and control.
  • Industrialized processing. Dagster orchestration on GKE and Spark processing make the pipeline reliable and reproducible.

Set-up

Audit, target architecture, migration, go-live.

  • Audit. Analysis of the existing Splunk platform, its flows and its uses.
  • Target architecture. Design of the platform on Google Cloud Platform: BigQuery analytical warehouse, orchestration and processing.
  • Migration. Transfer of CPU, GPU and user usage data into the analytical warehouse.
  • Industrialization. Pipeline orchestration with Dagster on GKE and Spark processing.
  • Go-live. Cutover to the new platform and measurement of costs and performance.

A seven-month engagement, from audit to go-live. Tech stack: Splunk, Google Cloud Platform, BigQuery, Dagster, GKE, Spark.

Results

40% lower platform costs and 30% faster processing, in seven months.

Infrastructure usage data is now structured in an analytical warehouse on Google Cloud. In production, the new platform costs 40% less than the old one and processes data 30% faster. All in seven months, from audit to go-live.

Two effects extend the migration.

  • Forecasting on a reliable base. Teams anticipate infrastructure needs from structured usage data.
  • A foundation for quality and governance. Gathered in a single warehouse, the data is simpler to control and document.

Expertises mobilisées

Savoir-faire

  • Cloud migration
  • Data architecture
  • Data warehouse
  • Pipeline orchestration
  • Cost optimization

Technologies

  • Google Cloud Platform
  • BigQuery
  • Dagster
  • GKE
  • Spark
  • Splunk

Signature

Four mechanisms that make the difference.

  • Audit before architecture. We design the target from the actual uses of the existing platform.
  • A warehouse built for analytics. BigQuery structures the data for forecasting, not only for storage.
  • Orchestrated pipelines. Dagster on GKE and Spark make processing reliable and reproducible.
  • Gains measured in production. We measure costs and performance on the live platform.

For you

A modern data platform, faster and less costly.

This approach addresses four situations.

  • Your legacy data platform is expensive and slows down your processing.
  • Your data is collected but too poorly structured for analysis and forecasting.
  • You are preparing a cloud migration and want to control its scope.
  • You need to strengthen the quality and governance of your data.

We audit your existing platform, design the target architecture and lead the migration through to production, with measured gains.

Vue aérienne d'un marais avec de petits cours d'eau sinueux traversant des zones de végétation brune et des berges sableuses.

contact

A similar challenge? Let's talk.

Describe your context. A Diametral expert will get back to you shortly with an initial scoping.

Thank you, your request has been received. A Diametral expert will get back to you shortly.

Oops, something went wrong. Please try again or email us at contact@diametral.com.

Projects in the same category

Aerial view of crossing paths between fields

Energy

Strategy

Data & AI Roadmap

Data & AI strategy scoping: target team, use case portfolio and 12-month roadmap

For the digital department of a major European energy group, we built the strategy of a new cross-functional Data & AI team in 45 days. Of 38 use cases identified across six business units, 12 were prioritized. Six months later, four were launched, two of them in production.

Aerial view of a marsh in autumn

Hospitality

Applied AI & Data Science

AI Assistant

AI assistant for hotel teams: operational support that drives productivity and guest satisfaction

For an international hotel group, we helped design an AI assistant that answers the operational questions of hotel teams using the support knowledge base and each hotel's own information. Agentic platform, optimization and integration into the group's RAG platform: the assistant went into production in 8 months and is rolling out to more than 1,500 hotels.

Luxury

Acculturation & Change

Acculturation Program

AI acculturation for 700 employees: the program of a luxury beauty House.

We support an international luxury beauty House in building the AI literacy of its 700 employees, across every function. The program runs seven sessions over four months. Its executive committee sponsors it, and we measure its impact at every step.

Close-up of a cracked salt flat

Services

Product Delivery

Recruiting Automation

Recruitment automation: sourcing, scoring and outreach built with recruiters

Before promising adoption to our clients, we put it to the test in-house. We automate our own recruitment end to end, starting from our recruiters' actual work: sourcing, scoring and outreach, in compliance with GDPR. Target: 20 qualified interviews per week per recruiter, compared with 6 to 10 today.

Close-up of mushroom gills

Retail

Applied AI & Data Science

Price Recommendation

Price reduction campaign optimization: product recommendations validated on historical performance

For a major food retailer, we built a recommendation system that shows, product by product, where to invest the price budget. Each strategy was validated on past campaigns before being applied. The recommendations now apply automatically to local campaigns, covering nearly a quarter of revenue.

Aerial view of tall grass flattened by the wind

Retail

Product Delivery

Verbatim analysis

Customer verbatim analysis with generative AI: the voice of the customer leveraged at scale for product and quality teams

For a home appliance manufacturer in Brazil, we turned thousands of previously untapped customer verbatims into a decision source for product and quality teams. Two experts supported by AI agents delivered the POC and the MVP in 1.5 months, compared with six months for a traditional organization.

Granite rocks stacked in balance

Manufacturing

Architecture & Engineering

Data Platform Migration

Migrating a data platform to the cloud: an analytical warehouse on GCP to steer compute infrastructure

For an international high-tech industrial group, we migrated the platform that tracks the usage of its virtual machines to an analytical warehouse on Google Cloud. In seven months, platform costs fell by 40% and processing speed rose by 30% in production.

Close-up of pink oyster mushrooms

Luxury

Applied AI & Data Science

Talk-to-your-data

AI assistant for finance teams: querying an on-premises OLAP cube in natural language

For a luxury House, we designed an AI assistant that lets finance teams query their OLAP cube in natural language, without writing a single line of MDX. In production since January 2026, the CFO uses it every day.

Aerial view of a sandbank crossed by a channel

Services

Product Delivery

Supply Chain Operations

Supply chain of a professional sports club: reliable stock flows to prepare forecasting and replenishment

At a professional sports club, stock movements, bills of materials, end-of-season returns and alerts ran on spreadsheets and paper, alongside the ERP. We mapped the actual flows, set adoption indicators before building and put in place a weekly usage report to the sponsor.

Snow-capped rocky summit lit by the setting sun

Banking

Acculturation & Change

Learning Expedition

Learning expedition for a board of directors: testing the payments strategy against the most advanced players

For the board of directors of the payments subsidiary of a major French banking group, we designed a three-day learning expedition. Top-level meetings with payments frontrunners, in another ecosystem, informed the adjustment of the strategy.