Learning & development · Resilient infrastructure

Build on a
solid foundation.

Nyxever helps engineers and organizations build resilient infrastructure: open technology, clear architecture guidance, and the practical knowledge to run it well.

The technology is open. Getting the solution right takes sound architecture and people who understand what they operate. That is where Nyxever focuses.

open technology inspectable standards and source
clear architecture decisions documented
shared practice lessons published
building status

Foundation first.

Resilient systems are designed, not patched together. They start from the outcome and the constraints, rest on a foundation that is understood layer by layer, and are run by people who know why each decision was made. Tools are chosen on fit, cost, risk and operability; open technology is preferred where it is a real equivalent.

Build the foundation first

Compute, network, storage and identity done properly make everything above them simpler. Shortcuts at the bottom of the stack are paid for at every layer on top.

Design for resilience

Explicit failure domains, tested recovery, and systems that degrade predictably. Resilience is an architecture property, decided before the first workload lands.

Share what works

Best practices, decision records and lessons from the lab are published, so others can learn from the reasoning, not just copy the configuration.

Understand it. Operate it. Decide with confidence.

The best infrastructure decisions are made by people who understand what is running beneath them. Nyxever’s learning material is concept-first: the problem, the architecture, the tradeoffs, then the configuration. Public cloud, managed services and self-hosted infrastructure each solve real problems; knowing which tradeoffs apply is the skill.

Concepts before configuration

Documentation for every layer of the foundation explains why before how, so the knowledge transfers to whatever stack you run.

Read the docs →

Best practices, explained

Architecture patterns and operating practices for resilient infrastructure, with the reasoning and the failure modes they protect against.

Architecture principles →

Decisions you can learn from

Each significant design choice is recorded with the context, the options considered and why one was chosen, including what was learned along the way.

Decision records →

A foundation to build on and learn from.

The Nyxever reference foundation is built on open technology and organized around the infrastructure functions that matter at runtime. Each layer states the problem it solves, what is decided, and which open standards are being evaluated. It is where the guidance is tested before it is taught.

The problem: run workloads reliably, survive a node failure, and move work between hosts without downtime. Decided: two core technologies. Proxmox VE for virtualization and HA clustering, and Kubernetes for container orchestration, running on Proxmox VE virtual machines.

  • Proxmox VE: KVM/QEMU virtual machines, LXC containers, live migration
  • Cluster quorum and fencing, so a single node failure needs no manual intervention
  • Kubernetes: scheduling, self-healing and declarative deployment of application workloads, described through the Kubernetes API so changes are reviewable and reproducible
  • Proxmox VE handles hardware, VM placement and host-level HA; Kubernetes handles scheduling, restarting and scaling containers
  • Standards in scope: OCI images and runtimes, CRI, CNI, CSI
  • Under evaluation: the Kubernetes distribution and installation method

The problem: keep workloads separated, encrypt traffic between services, and make every external connection intentional. DNS, routing and firewall rules live in version control. Under evaluation: the open standards below.

  • Segmentation: 802.1Q VLANs, VXLAN / BGP EVPN, Kubernetes NetworkPolicy
  • Service-to-service: TLS 1.3 with mutual authentication (mTLS)
  • Site and cloud connectivity: IPsec/IKEv2 and the WireGuard protocol as candidates
  • Naming: split-horizon DNS, DNSSEC, dynamic updates (RFC 2136)

The problem: keep data durable when a disk or node fails, serve it through standard interfaces, and back it up somewhere else. Decided: Ceph for distributed storage and ZFS for local storage.

  • Ceph: block (RBD), file (CephFS) and object storage in one cluster
  • ZFS on individual nodes for local workloads and boot volumes
  • Standard interfaces: S3 API, NFS, CSI for Kubernetes volumes
  • Encryption at rest; replication or erasure coding across failure domains
  • Backups following 3-2-1, with off-site or cloud object storage where it makes sense

The problem: every service needs to prove who it is, secrets must never sit in plaintext, and people should sign in once. Under evaluation: the software for each part; the standards are chosen.

  • PKI: X.509 certificates issued automatically via ACME (RFC 8555)
  • Short-lived certificates to limit exposure without manual rotation
  • Secrets management: encryption, audit logging, leasing and rotation
  • Authentication: OIDC / OAuth 2.0 and SAML 2.0, with SCIM for provisioning
  • Workload identity: SPIFFE / SVID

AI sits between identity and observability deliberately, with human intent at the center. Humans define what matters, set the boundaries and interpret the output. AI speeds up pattern recognition in logs, anomaly detection, documentation and assisted configuration; the judgment and accountability stay with the operator.

  • Local or hosted inference, decided per workload by data sensitivity, latency and cost
  • Open-weight models and open formats (GGUF, safetensors, ONNX) under evaluation
  • Model Context Protocol (MCP) for scoped, logged tool access
  • GPU access through the compute layer (PCIe passthrough or Kubernetes workloads)
  • Human-in-the-loop by design: no autonomous action without operator approval

The problem: see what every layer is doing, find failures quickly, and alert on what affects users. Telemetry is collected inside the platform boundary and can be forwarded to a managed platform where that is the right fit. Under evaluation: the storage and dashboard tooling.

  • Instrumentation: OpenTelemetry and OTLP, vendor-neutral by design
  • Metrics: OpenMetrics exposition format
  • Logs: structured logging, syslog (RFC 5424), OTLP logs
  • Alerting on service-level objectives and error budgets, with runbooks linked
Speed isn’t the enemy. Cutting corners is.

Clear guidance for resilient solutions.

Open technology is the starting point. A resilient solution also needs clear architecture, sound decisions and a team that can run it. Nyxever engagements start from your workloads and constraints and end with something you can operate and explain: decisions written down, tradeoffs stated, and a system designed to outlast the project that created it.

Vision & Strategy

Direction & Planning

Clarity before commitment. An honest assessment of where you are, agreement on what “done” means, and a roadmap that holds up under real constraints and can be explained at every level of the organization.

You get Current-state assessment, target outcomes, and a phased roadmap.

Systems Architecture

Design & Engineering

Architecture designed to be understood, operated and handed off. Explicit failure domains, documented tradeoffs, and reliability designed in from the start. If you can’t explain why a decision was made, the decision isn’t finished.

You get Architecture review or design, with decision records and an operating model.

Cybersecurity

Security Architecture

Security grounded in your actual threat surface. Threat modeling, control design and validation built into architecture decisions from the beginning, with assessments that report plainly on what works and what doesn’t.

You get Threat model, prioritized control gaps, and a remediation plan.

AI Fluency: Foundation & Frameworks

Applied AI

The knowledge to evaluate models, frameworks and infrastructure with clarity: local versus hosted inference, open-weight models, retrieval-augmented generation, and responsible deployment. AI integrated in ways that are transparent, auditable, and aligned with what your organization needs.

You get Workshops and an evaluation framework for your AI use cases.

Training

Hands-On Instruction

Instruction built around doing. Linux administration, private cloud infrastructure, security fundamentals, and the operational layer that managed services usually hide. The lab is the infrastructure: real systems, real failures, real understanding.

You get Hands-on labs on real infrastructure, tailored to your team.

Knowledge grows when it’s shared.

Practitioners learn fastest from each other. Bring questions, share what you’ve built, and challenge the decisions. If you’ve solved a similar problem differently, with a managed cloud, other tooling or something in between, that perspective makes everyone’s architecture better.

Built to teach. Designed to last.

Nyxever grew out of more than twenty years of systems architecture across cloud, enterprise networking, identity and security, and distributed systems, in environments where getting it wrong has consequences. The lesson that kept repeating: the teams that make the best decisions are the ones who understand what is running beneath them.

Nyxever turns that experience into three things: a reference foundation built on open technology, learning material that explains the reasoning behind it, and direct guidance for organizations designing their own resilient solutions.

The foundation sits at the layer most managed services abstract away: compute, storage, networking, identity, observability and AI on infrastructure you control. Self-hosting isn’t always the right answer, but understanding that layer gives you a real choice about what to run yourself and what to delegate.

Nyxever is founded and led by Charles Côté, systems architect.

Let’s talk about what you’re building.

Whether you’re working through a strategy question, evaluating an architecture, planning a resilient platform, or developing your team’s skills, reach out. Every engagement starts with a direct conversation about the problem.

Change is constant and accelerating. The only foundation worth building is one designed to adapt.