Open source software sits at the foundation of almost every modern technology stack, yet the companies that depend on it most often underinvest in understanding the landscape that sustains it. The tools that emerged and matured through 2021 offer a useful lens on where enterprise infrastructure is heading — and where the real leverage points are for organizations building on top of it.
ITSulu builds on and contributes to open source projects internally. We use a wide range of open source tools in our own infrastructure and in client engagements. This perspective shapes how we think about open source not as a cost reduction strategy but as a quality and speed advantage that compounds over time.
Machine Learning Infrastructure: TensorFlow and the Ecosystem Around It
TensorFlow, released under the Apache 2.0 license and maintained primarily by Google, remains one of the most widely deployed machine learning frameworks across industries. The 2.0 release in September 2019 addressed many of the usability complaints that had accumulated around the original API, and enterprise adoption accelerated significantly as a result.
What makes TensorFlow particularly relevant for enterprise AI deployments is not just the framework itself but the ecosystem it anchors. Deployment tooling, model serving infrastructure, hardware optimization layers, and monitoring frameworks all build on TensorFlow's abstractions. Organizations that have invested in TensorFlow expertise can deploy models across CPU, GPU, and TPU targets without fundamentally rewriting their pipelines, which matters enormously at the scale where AI infrastructure costs become material.
Healthcare, search, retail, education, and financial services all have significant TensorFlow deployments in production. The common pattern is not a single flagship model but dozens of smaller models running inference continuously against operational data — fraud detection, demand forecasting, personalization, anomaly detection. Each of these use cases benefits from the maturity of TensorFlow's production serving infrastructure, which has been hardened by years of large scale deployment at Google and across the open source community.
Kubernetes Ecosystem: minikube, kubecost, and the Cost Management Frontier
Kubernetes has become the de facto standard for container orchestration at scale, but the ecosystem tools built around it arguably matter as much as the core platform for day to day operations. minikube provides a local Kubernetes environment for development and testing, enabling developers to work against a realistic cluster configuration without requiring cloud resources for every iteration cycle.
kubecost represents a category of tooling that emerged as Kubernetes adoption scaled: infrastructure cost attribution. When dozens of teams share a Kubernetes cluster and costs are allocated at the cloud account level rather than the namespace or deployment level, the result is a cloud bill that nobody can explain and nobody is accountable for. kubecost maps Kubernetes resource consumption to costs at the namespace, deployment, and label level, giving platform teams the data to have productive conversations with application teams about efficiency.
The pattern that kubecost enables — transparent cost attribution at the team and application level — consistently produces better infrastructure economics than either top down cost mandates or invisible billing. Teams that can see the cost impact of their resource requests make different decisions than teams that cannot. Open source tooling that makes this visibility accessible without enterprise license costs has driven adoption far faster than commercial alternatives would have.
Observability and Monitoring: Grafana and the Open Observability Stack
Grafana has become the standard visualization layer for the open source observability stack, pairing with Prometheus for metrics, Loki for logs, and Tempo for distributed traces. The combination of these four tools gives operations teams a full observability platform that competes directly with expensive commercial alternatives.
What distinguishes the Grafana ecosystem from commercial observability tools is not just cost but control. Organizations running their own observability stack retain ownership of their data, avoid vendor lock in on ingestion costs, and can customize dashboards and alerting logic to match their specific operational needs. For organizations running Kubernetes workloads, the native Kubernetes metrics exposed by kube-state-metrics and the DCGM exporter for GPU metrics integrate directly with Prometheus and Grafana without additional configuration.
The operational leverage from mature observability is difficult to overstate. Mean time to resolution drops when engineers have the data to understand what is happening rather than speculating. Alert fatigue decreases when alerting rules are built on signals that actually predict incidents rather than proxy metrics that fire constantly. The time investment in building a good observability foundation pays back in every significant incident that gets resolved faster because the data was there.
Data Infrastructure: Delta Sharing and the Lakehouse Architecture
Delta Sharing, released as an open source protocol by Databricks, addresses one of the fundamental challenges of multi cloud and multi organization data infrastructure: sharing data without copying it. Traditional data sharing approaches require either giving a partner direct access to your storage (a security risk) or copying data to a neutral location (an operational overhead and a freshness problem).
Delta Sharing allows an organization to share live, read only access to specific datasets with external parties through a REST API, without moving the data. The recipient queries the data through their own tools — Pandas, Spark, Tableau, PowerBI — against the original source. The data owner retains access control and can revoke sharing at any time.
For enterprises running analytics across organizational boundaries — whether that means sharing data with subsidiaries, partners, or customers — Delta Sharing provides a clean governance model that commercial data exchange platforms have historically charged significant premiums to enable.
Container Platform: OKD and the OpenShift Foundation
OKD is the community distribution of OpenShift, Red Hat's enterprise Kubernetes platform. For organizations that want OpenShift's operational tooling and security posture without the Red Hat subscription cost, OKD provides a comparable foundation. The tradeoff is that OKD requires more operational expertise to run and maintain, while the subscription version includes Red Hat's support infrastructure.
For organizations building internal platform engineering capabilities, OKD is a compelling way to develop expertise with the OpenShift ecosystem before committing to an enterprise subscription. The security defaults, the integrated developer tooling, and the operator framework that OpenShift has built on top of vanilla Kubernetes represent genuine operational value that the open source version preserves.
Developer Productivity: Nx and Monorepo Architecture
Nrwl's Nx (now simply Nx) has emerged as the leading tool for managing JavaScript and TypeScript monorepos at enterprise scale. As organizations have moved toward micro frontend architectures and large shared component libraries, the build and test overhead of traditional CI approaches has become a significant bottleneck. Nx's computation cache and affected command system dramatically reduce build times by only rebuilding and retesting the parts of the codebase that actually changed.
The productivity gains from mature monorepo tooling are consistently underestimated before adoption. Organizations that have migrated from multi repo to monorepo architectures with Nx regularly report 40 to 70% reductions in CI pipeline duration, with corresponding improvements in developer iteration speed and release frequency.
Open Source Licensing: What Enterprise Teams Need to Understand
Using open source software in a commercial context requires understanding the licensing terms that govern it. Most enterprise teams have a working understanding of the distinction between permissive licenses and copyleft licenses, but the practical implications are worth spelling out because licensing errors create legal exposure that most technology leaders would rather avoid.
Permissive licenses — Apache 2.0, MIT, BSD — allow modification and redistribution with few conditions beyond attribution. TensorFlow, Grafana, and most of the Kubernetes ecosystem use permissive licenses. Organizations can use these tools in commercial products, modify them, and distribute modified versions without licensing their own code under the same terms. The primary obligation is attribution and, for Apache 2.0, patent grant preservation.
Copyleft licenses — GPL, AGPL, LGPL — carry the obligation to release modifications under the same license when distributing. AGPL specifically extends this requirement to software provided as a service, which is the variant most relevant for organizations building cloud hosted applications. MongoDB and CockroachDB switched to SSPL, a custom license with similar SaaS distribution implications, for exactly this reason.
For organizations building internal tools, the distinction matters less — you are not distributing the software, so copyleft obligations generally do not trigger. For organizations building products that incorporate open source components, a license audit is a standard step in any acquisition due diligence and should be a standard step in any new open source adoption as well. The cost of the audit is small. The cost of discovering a licensing problem post-acquisition is not.
How ITSulu Can Help
ITSulu uses open source tools extensively across our AI, Kubernetes, cloud, and networking practices. We help clients evaluate, implement, and get production value from the open source ecosystem — from Kubernetes observability stacks built on Grafana and Prometheus, to AI infrastructure on TensorFlow and GPU optimized Kubernetes, to Odoo ERP on open source foundations.
If your organization wants to take better advantage of the open source tools available to it, or needs help building the internal expertise to operate them at production scale, we can help structure an approach that fits your team's capabilities and your organization's risk tolerance.
Contact ITSulu today to discuss open source infrastructure strategy for your environment.