Treating Multi-Cloud Strategy as a Product for Unified Engineering Experience
This article from QCon London 2026 highlights that multi-cloud strategies are often approached incorrectly, leading to technical chaos and increased costs. JP Morgan Chase's Luis Albinati and Surabhi Mahajan advocate for treating multi-cloud as a product, focusing on unified capabilities rather than siloed cloud providers, to achieve a more sustainable and governed architecture. They present practical steps like demand governance and standardizing observability across environments.
Read original on InfoQ ArchitectureThe Multi-Cloud Dilemma: A Product Problem, Not Just Technical
Many enterprises find themselves in a multi-cloud environment due to organic growth, acquisitions, or specific workload needs (like AI's GPU demand). This often leads to an unmanaged 'chaos tree' of disparate systems, increasing costs and complexity. The core argument from JP Morgan Chase at QCon London 2026 is that this is fundamentally an organizational and product management problem, not solely a technical one.
- Initial promises vs. reality: Cloud migration was sold on cost optimization and agility, but often results in higher spending and ungoverned workloads.
- Siloed operations: Teams often operate in silos, owning specific cloud providers (AWS, GCP, Azure), each with independent roadmaps, leading to a fragmented view.
- Remediation programs: The complexity forces companies to layer on numerous remediation programs like FinOps, governance, and SaaS compliance, each with its own tools and headcount.
Architecting for Unified Capabilities Across Clouds
The proposed solution shifts the architectural focus from individual cloud providers to horizontal capabilities. Instead of asking 'What does our AWS roadmap look like?', the question becomes 'Are we delivering unified observability everywhere?' This involves organizing around cross-cutting concerns like observability, cost management, identity, networking, and delivery. This approach helps break down silos and ensures a consistent experience and governance across the entire multi-cloud estate.
Frankenstein Logging to Unified Signals
A concrete example was the 'Frankenstein logging' platform at JPMC, funneling all logs (GCP, AWS, on-prem) into AWS S3, leading to massive egress costs and noisy data. The fix involved treating logging as a product: understanding user needs (SREs, cyber ops), deploying OpenTelemetry collectors at the edge in each cloud to filter low-value events, and standardizing on distributed trace IDs as a 'golden thread'. This cut log volume by 60-80% and reduced MTTR from hours to minutes.
Key Principles for Multi-Cloud Product Thinking
- Capabilities over siloed solutions: Focus on providing core functionalities that span all cloud environments.
- Shared discovery over isolated priorities: Foster collaboration and shared understanding of requirements across teams and clouds.
- Golden paths over bespoke one-offs: Establish standardized, well-documented patterns and solutions for common needs.
- Governance by design: Embed governance mechanisms from the outset, rather than layering them on as afterthoughts.
- Resilience over mere portability: Prioritize robust, fault-tolerant designs, recognizing that true portability isn't always the primary goal; sometimes better contracts and differentiated SLAs are more effective than uniform Kubernetes deployments.