Main Page

Discover cloud and AI on infocepo.com

infocepo.com – Cloud, AI & Labs

Welcome to the infocepo.com portal.

This wiki is intended for system administrators, cloud engineers, developers, students, and enthusiasts who want to:

• Understand modern architectures (Kubernetes, OpenStack, bare-metal, HPC…)
• Deploy private AI assistants and productivity tools
• Build hands-on labs to learn by doing
• Prepare large-scale audits, migrations, and automations

The goal: turn theory into reusable scripts, diagrams, and architectures.

---

Getting started quickly

Recommended paths

1. Build a private AI assistant

• Deploy a typical stack: Open WebUI + Ollama + GPU (H100 or consumer-grade GPU)
• Add a chat model and a summarization model
• Integrate internal data (RAG, embeddings)

2. Launch a Cloud lab

• Create a small cluster (Kubernetes, OpenStack, or bare-metal)
• Set up a deployment pipeline (Helm, Ansible, Terraform…)
• Add an AI service (transcription, summarization, chatbot…)

3. Prepare an audit / migration

• Inventory servers with ServerDiff.sh
• Design the target architecture (cloud diagrams)
• Automate the migration with reproducible scripts

Content overview

• AI guides & tools : assistants, models, evaluations, GPUs
• Cloud & infrastructure : HA, HPC, web-scale, DevSecOps
• Labs & scripts : audit, migration, automation
• Comparison tables : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.

---

future

The world after automation

AI Assistants & Cloud Tools

AI Assistants

ChatGPT

• https://chatgpt.com ChatGPT – Public conversational assistant, suited for exploration, writing, and rapid experimentation.

Self-hosted AI assistants

• https://github.com/open-webui/open-webui Open WebUI + https://www.scaleway.com/en/h100-pcie-try-it-now/ H100 GPU + https://ollama.com Ollama

Typical stack for private assistants, self-hosted LLMs, and OpenAI-compatible APIs.

• https://github.com/ynotopec/summarize Private summary – Local, fast, offline summarizer for your own data.

Development, models & tracking

Discovering and tracking models

• https://ollama.com/library LLM Trending – Model library (chat, code, RAG…) for local deployment.
• https://huggingface.co/models Models Trending – Model marketplace, filterable by task, size, and license.
• https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending – Vision-language models (image → text).
• https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation – Image generation model comparisons.

Evaluation & benchmarks

• https://lmarena.ai/leaderboard ChatBot Evaluation – Chatbot rankings (open-source and proprietary models).
• https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard – Benchmark of embedding models for RAG and semantic search.
• https://ann-benchmarks.com Vectors DB Ranking – Vector database comparison (latency, memory, features).
• https://top500.org/lists/green500/ HPC Efficiency – Ranking of the most energy-efficient supercomputers.

Development & fine-tuning tools

• https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending – Major recent open-source projects, sorted by popularity and activity.
• https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning – Advanced framework for LLM fine-tuning (instruction tuning, LoRA, etc.).
• https://www.perplexity.ai Perplexity AI – Advanced research and synthesis oriented as a “research copilot”.

AI Hardware & GPUs

GPUs & accelerators

• https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100 – Datacenter GPU for Kubernetes clusters and intensive AI workloads.
• NVIDIA 5080 – Consumer GPU for lower-cost private LLM deployments.
• https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator – Hardware accelerator dedicated to LLM inference.

---

Open models & internal endpoints

(Last update: 2026-02-13)

The models below correspond to logical endpoints (for example via a proxy or gateway), selected for specific use cases.

Endpoint	Description / Primary use case
ai-chat	Based on gpt-oss-20b – General-purpose chat, good cost / quality balance.
ai-translate	gpt-oss-20b, temperature = 0 – Deterministic, reproducible translation (FR, EN, other languages).
ai-summary	qwen3 – Model optimized for summarizing long texts (reports, documents, transcriptions).
ai-code	gpt-oss-20b – Code reasoning, explanation, and refactoring.
ai-code-completion	gpt-oss-20b – Fast code completion, designed for IDE auto-completion.
ai-parse	qwen3 – Structured extraction, log / JSON / table parsing.
ai-RAG-FR	qwen3 – RAG usage in French (business knowledge, internal FAQs).
gpt-oss-20b	Agentic tasks.

Usage idea: each endpoint is associated with one or more labs (chat, summary, parsing, RAG, etc.) in the Cloud Lab section.

---

News & Trends

• https://www.youtube.com/@lev-selector/videos Top AI News – Curated AI news videos.
• https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper – Example of real-time transcription with speaker detection.
• https://github.com/openai-translator/openai-translator OpenAI Translator – Modern extension / client for LLM-assisted translation.
• https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM – Conversational search based on LLMs and OpenSearch.

---

Training & Learning

• https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained – Introduction to Transformers, the core architecture of LLMs.
• Hands-on labs, scripts, and real-world feedback in the CLOUD LAB project below.

---

Cloud Lab & Audit Projects

The Cloud Lab provides reproducible scenarios: infrastructure audits, cloud migration, automation, high availability.

Audit project – Cloud Audit

ServerDiff.sh

Bash audit script to:

• detect configuration drift,
• compare multiple environments,
• prepare a migration or remediation plan.

Example of Cloud migration

Example: migration of virtual environments to a modernized cloud, including audit, architecture design, and automation.

Task	Description	Duration (days)
Infrastructure audit	82 services, automated audit via ServerDiff.sh	1.5
Cloud architecture diagram	Visual design and documentation	1.5
Compliance checks	2 clouds, 6 hypervisors, 6 TB of RAM	1.5
Cloud platform installation	Deployment of main target environments	1.0
Stability verification	Early functional tests	0.5
Automation study	Identification and automation of repetitive tasks	1.5
Template development	6 templates, 8 environments, 2 clouds / OS	1.5
Migration diagram	Illustration of the migration process	1.0
Migration code writing	138 lines (see MigrationApp.sh)	1.5
Process stabilization	Validation that migration is reproducible	1.5
Cloud benchmarking	Performance comparison vs legacy infrastructure	1.5
Downtime tuning	Calculation of outage time per migration	0.5
VM loading	82 VMs: OS, code, 2 IPs per VM	0.1
Total		15 person-days

Stability checks (minimal HA)

Action	Expected result
Shutdown of one node	All services must automatically restart on remaining nodes.
Simultaneous shutdown / restart of all nodes	All services must recover correctly after reboot.

---

Web Architecture & Best Practices

Principles for designing scalable and portable web architectures:

• Favor simple, modular, and flexible infrastructure.
• Follow client location (GDNS or equivalent) to bring content closer.
• Use network load balancers (LVS, IPVS) for scalability.
• Systematically compare costs and beware of vendor lock-in.
• TLS:
  • HAProxy for fast frontends,
  • Envoy for compatibility and advanced use cases (mTLS, HTTP/2/3).
• Caching:
  • Varnish, Apache Traffic Server for large content volumes.
• Favor open-source stacks and database caches (e.g., Memcached).
• Use message queues, buffers, and quotas to smooth traffic spikes.
• For complete architectures:
  • https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture
  • https://github.com/systemdesign42/system-design System Design GitHub

---

Comparison of major Cloud platforms

Feature	Kubernetes	OpenStack	AWS	Bare-metal	HPC	CRM	oVirt
Deployment tools	Helm, YAML, ArgoCD, Juju	Ansible, Terraform, Juju	CloudFormation, Terraform, Juju	Ansible, Shell	xCAT, Clush	Ansible, Shell	Ansible, Python
Bootstrap method	API	API, PXE	API	PXE, IPMI	PXE, IPMI	PXE, IPMI	PXE, API
Router control	Kube-router	Router/Subnet API	Route Table / Subnet API	Linux, OVS	xCAT	Linux	API
Firewall control	Istio, NetworkPolicy	Security Groups API	Security Group API	Linux firewall	Linux firewall	Linux firewall	API
Network virtualization	VLAN, VxLAN, others	VPC	VPC	OVS, Linux	xCAT	Linux	API
DNS	CoreDNS	DNS-Nameserver	Route 53	GDNS	xCAT	Linux	API
Load Balancer	Kube-proxy, LVS	LVS	Network Load Balancer	LVS	SLURM	Ldirectord	N/A
Storage options	Local, Cloud, PVC	Swift, Cinder, Nova	S3, EFS, EBS, FSx	Swift, XFS, EXT4, RAID10	GPFS	SAN	NFS, SAN

This table serves as a starting point for choosing the right stack based on:

• Desired level of control (API vs bare-metal),
• Context (on-prem, public cloud, HPC, CRM…),
• Existing automation tooling.

---

Useful Cloud & IT links

• https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared – AWS / Azure / GCP service mapping.
• https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map – Global Internet mapping.
• https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape – Overview of cloud-native projects (CNCF).
• https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki – Wikimedia infrastructure, real large-scale example.
• https://openapm.io OpenAPM – SRE Tools – APM / observability tooling.
• https://access.redhat.com/downloads/content/package-browser RedHat Package Browser – Package and version search at Red Hat.
• https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Barometer of IT freelance daily rates.
• https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor) – Salary indicators.

---

Advanced: High Availability, HPC & DevSecOps

High Availability with Corosync & Pacemaker

Basic principles:

• Multi-node or multi-site clusters for redundancy.
• Use of IPMI for fencing, provisioning via PXE/NTP/DNS/TFTP.
• For a 2-node cluster:

 – carefully sequence fencing to avoid split-brain,
 – 3 or more nodes remain recommended for production.

Common resource patterns

• Multipath storage, LUNs, LVM, NFS.
• User resources and application processes.
• Virtual IPs, DNS records, network listeners.

HPC

• Job orchestration (SLURM or equivalent).
• High-performance shared storage (GPFS, Lustre…).
• Possible integration with AI workloads (large-scale training, GPU inference).

DevSecOps

• CI/CD pipelines with built-in security checks (linting, SAST, DAST, SBOM).
• Observability (logs, metrics, traces) integrated from design time.
• Automated vulnerability scanning, secret management, policy-as-code.

---

About & Contributions

For more examples, scripts, diagrams, and feedback, see:

• https://infocepo.com infocepo.com

Suggestions for corrections, diagram improvements, or new labs are welcome. This wiki aims to remain a living laboratory for AI, cloud, and automation.