Cloud migration reduces critical processes by 75% at Catupiry

Catupiry is one of the most traditional and iconic companies in the Brazilian food sector. With 113 years of history, it has established itself as a benchmark in the market and operates a robust and widely distributed industrial structure. Its operation includes four strategically located manufacturing plants for the collection and processing of raw materials, as well as a centralized Distribution Center (DC) responsible for the distribution of production.

The company has also diversified its operations through different business units, including its own stores that combine a gourmet shop and a restaurant. Currently, it maintains a portfolio of over 130 products, serving both the retail and food service segments throughout Brazil, in addition to having structured channels for export.

Despite the strength of its brand and its century-long history, the company's executive leadership, represented by Tatiana Campos, Head of Technology, identified that sustaining growth in the coming years would depend on a profound modernization of the technological architecture. The goal was to create a foundation capable of supporting the physical and digital expansion of the business, offering greater scalability, efficiency, and support for the continuous evolution of the portfolio.

The real problem: operational limitations and impacts

Catupiry's information technology infrastructure operated under a centralized on-premise, physically concentrated at its headquarters in São Paulo. Although robust by legacy standards, the server and communication link architecture became a bottleneck for business scalability.

Bottlenecks and processing latency

The legacy systems ecosystem faced severe batch processing constraints and access concurrency. Extracting complex analytical reports required 2-3 hour continuous processing. The most critical impact was concentrated in the closing of milk management —the main raw material in the production chain—whose data consolidation required complex operational routines that extended for up to 12 consecutive hours.

Operational risks and single point of failure (SPOF)

The physical centralization at headquarters generated a high dependence on connectivity for all remote units. The four industrial plants and the distribution center relied on communication tunnels directed to the local infrastructure in São Paulo. In the event of any power outage, hardware failure, or link failure at headquarters, the entire supply chain, billing, and shipping of the factories risked total paralysis, representing a critical operational risk with a high financial impact per hour of downtime.

Solution architecture

To address the issue of decentralization and low performance, Catupiry implemented Skyone Autosky , an integrated platform that enables the migration of monolithic Client-Server applications and ERPs to public cloud environments (Hyperscalers such as AWS, Oracle Cloud Infrastructure – OCI, Microsoft Azure, or Google Cloud Platform – GCP) using the concept of Zero - Code Change Migration .

The manufacturing plants, distribution center, and stores access the environment via a web browser, using an HTTPS connection on port 443 with low bandwidth requirements, around 100 Kbps. All traffic passes through a cybersecurity layer comprised of Zero Trust Network Access (ZTNA), QR Code-based multifactor authentication (MFA), and a Web Application Firewall (WAF).

After this layer of protection, access is directed to the Skyone Autosky platform, which houses the portal and the administrative core of the operation, as well as the auto-scaling engines and server templates.

From there, the operation is routed to a client-segregated environment (VPC), where application instances, running on dynamic ephemeral VMs, communicate directly with corporate databases. This environment supports various technologies, including Oracle, SQL Server, PostgreSQL, and HANA.

Complementing the architecture, Skyone Studio adds data- and artificial intelligence-focused features, bringing together Lakehouse functionalities, iPaaS pipelines, and generative AI and machine learning projects.

Components and technology stack deployed

• Skyone Autosky portal & core: responsible for the orchestration, provisioning, and centralized management of Catupiry's segregated environment.
• Application Abstraction Layer (Web Access / Plugin): automatic conversion of the native client-server ERP interface for direct execution via web browsers (HTML5) on any operating system (Windows, macOS, Linux), eliminating the need for heavy local emulation or traditional VPNs.
• Intelligent auto-scaling engines: Skyone's proprietary algorithms that monitor infrastructure variables (CPU utilization, RAM, and active user volume) minute by minute, dynamically provisioning or terminating ephemeral instances to absorb demand peaks (such as accounting and production closing) without human intervention.
• Security based on Zero Trust architecture: complete isolation of the environment in virtual private networks (VPCs). Network connections operate strictly through encrypted secure ports (HTTPS on port 443 or via TCP port 491). Authorization of connections to the dynamic IPs of the servers occurs only in real time after double authentication (MFA and identity privilege check), being revoked upon each disconnection.
• Data persistence and interoperability: native compatibility with the enterprise database engines used in the company's ERPs (supporting ecosystems such as Oracle, SQL Server, PostgreSQL, or SAP HANA).

Technical challenges and risk mitigation

The transition from an analog and centralized manufacturing model to a digitally native operation has imposed complex engineering and management challenges:

• Critical risk of production line interruption: In the perishable food sector, unplanned system shutdowns result in immediate losses of raw materials ( fresh). To mitigate this risk, Skyone's Professional Services team designed a homologation environment identical to the production environment using Server Templates . This allowed for the execution of load stress tests and validation of analytical reports before go-live.
• Abstraction of legacy code: the ERP architecture lacked native APIs or modern web connectors. Rewriting the source code would require substantial investments and years of development. The trade-off was resolved by Autosky's application virtualization technology, which encapsulates the original binary logic and delivers only the rendered presentation layer to the end-user's browser.
• Hybrid workforce management and factory connectivity: the network infrastructure of the plants needed to operate with tolerable latencies even on limited internet connections. Because Autosky processes all logic on cloud servers and transmits only interface and screen coordinates, the bandwidth requirement per user was reduced to a stable 100 Kbps, making the operation resilient to fluctuations from regional telecommunications operators.

Implementation strategy

The migration process was planned and executed through a rigorous phased schedule, structured to take advantage of seasonal windows of low industrial activity

        January Carnival Easter ▼ ▼ ▼ ┌─────────────────────┐┌────────────────────┐┌─────────────────────┐ │ Technical Alignment ││ Environment ││ Go-Live Window │ │ and Design of ││ Homologation and ││ Final Migration, │ │ Cloud Architecture ││ Load Testing ││ Validation and Rollout │ └─────────────────────┘└────────────────────┘└─────────────────────┘

1. Phase 1 – Alignment and diagnosis (January): beginning of technical discussions, mapping of infrastructure dependencies, local network topology, and inventory of historical databases.
2. Phase 2 – Homologation and simulation (Carnival period): Automated construction of the mirror environment (Staging) using templates . Execution of data validation routines, connectivity tests, and perimeter security assessments.
3. Phase 3 – Final migration and rollout (Easter period): migration of databases in a production environment, disconnection of local links, and redirection of factory traffic directly to the public cloud. The transition was completed with zero downtimeperceivedby end users, who maintained the same operational experience but with immediate performance gains.

Measurable results and business impact

The gains resulting from the adoption of the Skyone Autosky platform were directly quantified in Catupiry's operational performance and computational efficiency KPIs:

• 75% reduction in critical shutdown time: milk management processing, which previously took 12 continuous hours, has been optimized to just 3 hours.
• 83% optimization in report generation: complex Business Intelligence (BI) extractions that previously paralyzed workstations for up to 3 hours now run in 20 to 30 minutes.
• 50% increase in overall system performance: upgrading the hardware processing layer to optimized cloud instances increased the fluidity of application input/output by up to 50%.
• Operational risk mitigation: complete decoupling of the four factories and the distribution center from physical dependence on the headquarters in São Paulo. Each unit now connects independently and redundantly to cloud servers, eliminating single points of failure.

Data activation and AI readiness

The migration structured the necessary technological foundation for the company's data governance. Catupiry began the curation, purging, and cleaning of a historical database encompassing 22 years of operational and commercial records.

With the support of the Skyone Studio ecosystem, this data is being consolidated into Lakehouse frameworks and integrated data pipelines via iPaaS, preparing the private data environment to feed advanced Machine Learning (ML) and Generative Artificial Intelligence (GenAI) models, aiming at demand prediction and intelligent automation of industrial processes .

Lessons learned

• The cloud as a foundation for AI: the successful adoption of generative or predictive artificial intelligence technologies intrinsically depends on the quality, accessibility, and centralization of the data infrastructure. Traditional organizations first need to consolidate their cloud architectures before reaping the strategic value of their historical data.
• Demystifying refactoring: this case study proves that modernizing mature enterprise systems doesn't necessarily require a complete rewrite or traumatic replacement of core systems (Rip and Replace). The no-code infrastructure virtualization approach provides accelerated ROI with low implementation risk.
• Two-way knowledge transfer: enterprise digital transformation projects generate extended value when they integrate teams. While Skyone's technical architecture team absorbed the complex business rules of the dairy manufacturing flow, Catupiry's technical team was trained in advanced concepts of cloud engineering, Zero Trust security, and data governance.

FAQ

How can we mitigate the risk of downtime during the migration of industrial ERP systems to the cloud?

Risk mitigation in industrial ERP system migrations is achieved through the use of cloud management platforms that supportzero-code changemigrations. Creating isolated testing environments using Server Templates ensures faithful simulation of production workloads. Scheduling the go-live for seasonal periods of low industrial activity (such as long weekends) allows for the final synchronization of databases to occur safely, eliminating operational impacts and ensuring business continuity.

What is zero-code change migration to the cloud?

Zero-code migration is a technological approach that allows transferring applications from traditional monolithic, client-server, or desktop architectures directly to public cloud environments (Hyperscalers) without the need to rewrite, refactor, or modify the original software source code. Through the use of advanced application virtualization layers (such as that provided by Skyone Autosky), the original system runs on optimized cloud instances, while the interface is translated and delivered in a lightweight and secure way directly to the end-user's web browser.

How does Skyone Autosky optimize costs with cloud infrastructure?

Skyone Autosky performs financial control and infrastructure cost optimization through proprietary auto-scaling algorithms that operate 24/7. These engines continuously monitor the actual usage of resources such as CPU and RAM by applications. When user demand drops, the platform automatically terminates ephemeral server instances, preventing the waste of idle resources. Additionally, Skyone offers predictable pricing in local currency per user, eliminating hidden variable fees and currency volatility associated with direct billing from public clouds.