DEV Community: Shannon Dias

How to Install Drupal 11 on an Ubuntu Dedicated Server

Shannon Dias — Thu, 17 Sep 2026 07:18:16 +0000

Drupal 11 is a flexible, open-source CMS, but getting it running securely in production takes more than a Composer command.

This tutorial walks through a full deployment on an Ubuntu 24.04 dedicated server, including:

Installing and securing Nginx, PHP 8.3, and MariaDB
Creating the Drupal project with Composer's recommended project structure (/web as the public root)
A hardened Nginx server block that blocks PHP execution inside uploaded files, denies hidden files, and routes Drupal Image Styles correctly
Enabling HTTPS with Let's Encrypt and testing certificate renewal
Setting temporary write permissions for installation, then locking them down afterward
Configuring Trusted Host Patterns, UFW firewall rules, and scheduled Drush cron under the www-data user
Backups, server monitoring, and a full 26-point pre-launch checklist

Read the Full Tutorial

The complete step-by-step walkthrough, including every command and Nginx config block, is available here:

👉 Read more on the Fit Servers blog

How to Install RabbitMQ on Ubuntu 24.04 (Dedicated Server Guide)

Shannon Dias — Thu, 17 Sep 2026 06:17:05 +0000

Why RabbitMQ?

RabbitMQ lets applications communicate asynchronously through queues instead of waiting on each other directly — which makes it a go-to choice for background jobs, microservices, event-driven systems, and notification pipelines.

Running it on a dedicated server gives you full control over CPU, memory, storage, networking, and firewall rules — but it also means you are responsible for configuring it securely.

What the Full Guide Covers

We put together a complete, step-by-step tutorial for Ubuntu 24.04 LTS that walks through:

Adding the Team RabbitMQ APT repositories and signing key
Installing the required Erlang/OTP packages
Installing and verifying the RabbitMQ server
Enabling the web-based Management Plugin
Creating a dedicated administrator account and removing the default guest user
Setting up virtual hosts and application-specific users with scoped permissions
Understanding RabbitMQ's core components (producers, exchanges, queues, consumers)
Configuring network listeners and UFW firewall rules per-service
Enabling TLS for AMQP (port 5671) and HTTPS for the Management UI (port 15671)
Monitoring queue depth, memory usage, and troubleshooting common startup/connectivity issues
A full production security checklist

Read the Full Tutorial

👉 Install RabbitMQ on Ubuntu 24.04 — Full Guide

If you're deploying RabbitMQ for real production workloads, the hardware underneath it matters — isolated CPU, memory, and network throughput make a real difference once message rates climb. That's where a dedicated server comes in.

CPU Core Count vs Clock Speed: Which Matters More?

Shannon Dias — Thu, 17 Sep 2026 05:29:35 +0000

CPU Core Count vs Clock Speed: Which Matters More?

When choosing a dedicated server, it's easy to default to comparing two numbers: core count and clock speed (GHz). But neither one tells the full story on its own — the right choice depends on how your application actually uses the CPU.

What's covered:

The difference between physical cores and logical threads
Why a 4.0 GHz CPU isn't automatically faster than a 3.5 GHz CPU
Single-threaded vs multi-threaded workload behavior
Per-core performance factors: architecture, cache, branch prediction
Amdahl's Law and diminishing returns from adding cores
CPU priorities for gaming servers, virtualization, video encoding, and databases
Why real-world benchmarking beats comparing spec sheets

Instead of asking "more cores or more GHz?", the better question is: how does my application use the CPU?

📖 Read the full guide: https://www.fitservers.com/blogs/server-cpu-cores-vs-clock-speed/

Build Your Own S3 Storage: MinIO Setup on Dedicated Servers

Shannon Dias — Fri, 11 Sep 2026 08:55:05 +0000

Build Your Own S3 Storage: MinIO Setup on Dedicated Servers

Cloud object storage is convenient - until the egress bill shows up. If you're storing backups, media, logs, or large unstructured datasets, self-hosting an S3-compatible storage layer on your own dedicated server can cut costs significantly while giving you full control over your data.

MinIO is a free, open-source object storage server that's fully API-compatible with Amazon S3. In the full tutorial, we cover:

Standalone vs. distributed (erasure-coded) deployment options
Hardware planning for high-storage servers (NVMe/SSD, drive topology, networking)
Step-by-step installation and systemd service configuration
Performance tuning for high-throughput workloads (file descriptors, jumbo frames, kernel tuning)
Enabling SSL/TLS with Let's Encrypt
Locking down access with firewalls and credential management
Creating and managing buckets with the MinIO Client (mc)
Monitoring, safe backup strategies, and common troubleshooting fixes

👉 Read the full step-by-step tutorial: https://www.fitservers.com/tutorials/howto/setup-minio-dedicated-server/

PCIe Gen5 NVMe for GPU Servers: Fixing Storage Bottlenecks

Shannon Dias — Fri, 11 Sep 2026 08:26:50 +0000

Why storage can quietly become your GPU's biggest limitation

A powerful GPU can only work as fast as the data reaching it. In AI training, model checkpointing, and rendering workloads, if storage can't keep pace, the GPU spends part of its time idle instead of computing — even though the GPU itself hasn't changed at all.

This post breaks down:

PCIe, NVMe, and Gen5 — three separate technologies that get bundled together, explained individually
Throughput comparison — SATA SSD vs. Gen3 vs. Gen4 vs. Gen5 NVMe (Gen5 reaching roughly 10–14+ GB/s sequential in the right platform)
GPUDirect Storage (GDS) — how it and DMA let an NVMe SSD write directly into GPU VRAM, bypassing the CPU "bounce buffer" that traditional data paths rely on
Multi-GPU PCIe lane planning — a single flagship GPU needs 16 lanes; an 8-GPU server can consume all 128 lanes on a dual-socket platform, which is why PCIe switches are needed for storage and networking
How to diagnose the bottleneck yourself — GPU utilization, I/O wait, and disk throughput metrics worth checking before upgrading anything

If you're architecting GPU infrastructure for ML/AI, this is a practical way to figure out whether your next dollar should go to compute or storage.

👉 Full tutorial: https://www.fitservers.com/blogs/pcie-gen5-nvme-gpu-servers-storage-bottleneck/

AI Inference at the Edge: Dedicated Servers vs Serverless APIs

Shannon Dias — Thu, 10 Sep 2026 07:19:08 +0000

Where should your AI inference actually run?

Generative AI, real-time computer vision, and dynamic recommendation engines are pushing inference workloads out of centralized cloud regions and toward the edge, where low latency is non-negotiable. That raises a real architectural question for CTOs, DevOps engineers, and founders: dedicated infrastructure, or a serverless AI API?

In the full article, we cover:

The difference between AI training and inference
What "edge AI inference" actually means (it's broader than just IoT devices)
Why latency matters for real-time applications like voice agents and industrial automation
A full side-by-side comparison: hardware control, pricing model, scaling, data privacy, and custom model support
The economics of serverless vs dedicated at different traffic volumes
A hybrid architecture pattern that uses both

If you're deciding how to host your next AI deployment, this breakdown gives you a practical framework rather than a one-size-fits-all answer.

👉 Read the full guide: https://www.fitservers.com/blogs/ai-inference-at-the-edge-dedicated-servers-vs-serverless-apis/

The Ultimate Guide to the Best Dedicated Server Specifications for Docker & Kubernetes in 2026

Shannon Dias — Thu, 03 Sep 2026 07:35:27 +0000

When deploying containerized applications at scale, the underlying hardware is just as critical as the orchestration software itself. As we move deeper into 2026, Kubernetes (K8s) and Docker power heavy data analytics pipelines, large language model (LLM) inferences, complex edge computing frameworks, and massive stateful databases.

For system administrators, DevOps engineers, and IT architects, choosing the right dedicated server specifications is a balancing act between performance, scalability, and cost.

Why Choose Bare-Metal Dedicated Servers in 2026?

Zero Virtualization Overhead: Cloud virtual machines (VMs) rely on a hypervisor, which consumes CPU cycles and introduces latency. Running containers directly on bare metal yields a 10% to 15% performance improvement.
Hardware Passthrough for AI/ML: Bare metal allows direct hardware passthrough to NVIDIA GPUs and accelerators without public cloud abstraction.
Predictable Latency for Microservices: Dedicated servers let you implement 100Gbps physical networks with eBPF-based routing (like Cilium), drastically reducing microservice latency.

Kubernetes Core Hardware Requirements

1. Processors (CPU)

Container orchestration relies heavily on concurrency. High core counts generally outweigh raw clock speed. Opt for AMD EPYC (Zen 5), Intel Xeon Scalable, or ARM AmpereOne processors.

2. Memory (RAM)

DDR5 is Mandatory: Provides double the bandwidth of DDR4.
Ratio: A standard rule of thumb is 4GB to 8GB of RAM per CPU core to prevent OOMKilled errors.

3. Storage (NVMe)

The Control Plane’s etcd database is extraordinarily sensitive to disk write latency. You must use PCIe Gen 4 or Gen 5 NVMe drives to prevent cluster failure.

4. Network Specifications

10 Gbps is the bare minimum, with 25 Gbps to 100 Gbps network interface cards (NICs) becoming standard for handling heavy internal pod traffic.

👉 Read the complete guide, configuration tables, and FAQ on the official Fit Servers Blog!

NVMe Gen 3 vs 4 vs 5: Stop Overpaying for Server Storage

Shannon Dias — Thu, 03 Sep 2026 07:01:33 +0000

When shopping for dedicated servers, it’s easy to get caught up in marketing specs boasting the fastest drive speeds on the market. But ask yourself: Does your application actually need Gen 5 NVMe, or are you paying for an edge case?

According to our infrastructure analysis at Fit Servers, 87% of dedicated server buyers overpay for storage speed they never use.

Quick Generation Overview

Specification	Gen 3	Gen 4	Gen 5
Max Bandwidth	3.9 GB/s	7.8 GB/s	15.6 GB/s
Typical IOPS	~500K	~1M+	~2M+
Price Premium	Baseline	+50-80%	+100-150%

Workload Sweet Spots

WordPress & Content Sites: Gen 3 is enough. Storage accounts for less than 5% of total page load time; caching and PHP rendering dominate.
Databases & High-Concurrency APIs: Gen 4 is the sweet spot. Delivering 2-3x the practical IOPS of Gen 3, it handles 50+ concurrent connections efficiently without breaking the bank.
AI/ML & Real-Time Systems: Gen 4 suffices for training (since GPU compute is the bottleneck). Gen 5 has a case only for real-time inference APIs with aggressive sub-50ms latency SLAs.

Optimize Before You Upgrade

Before spending extra budget on storage generation upgrades, make sure you've checked:

Database indexes & query tuning (30–50% improvement)
Caching layers like Redis (60–80% reduction in disk load)
Connection pooling (PgBouncer/ProxySQL)

👉 Read the full benchmark report and decision matrix here: Fit Servers Blog

Your "EU Region" Cloud Server Might Not Be GDPR-Safe

Shannon Dias — Thu, 27 Aug 2026 08:56:51 +0000

If your SaaS handles personal data belonging to EU citizens, the physical location of your servers — and the legal jurisdiction of the company that owns them — is a core compliance question, not a minor infra detail.

Data residency vs. data sovereignty

These two terms get conflated constantly:

Data residency: your data is physically stored within a given country's borders.
Data sovereignty: which legal jurisdiction actually governs that data, and who can compel its disclosure.

AWS, Google Cloud, and Azure can give you EU data residency by letting you pick a Frankfurt availability zone. What they can't give you is EU data sovereignty - because they're US-incorporated companies, and therefore still subject to the US CLOUD Act, which can compel them to hand over data to US authorities regardless of where it's physically stored. That directly conflicts with how GDPR treats data transfers outside the EU.

Why Frankfurt specifically

Germany's own privacy law (BDSG) goes beyond GDPR's baseline.
Frankfurt is the EU's financial capital, so its data centers are built to extremely high physical security and redundancy standards.
It's home to DE-CIX, one of the largest internet exchange points in the world, with peak throughput over 19.6 Tbps.

That last point matters for performance, not just compliance - DE-CIX means direct peering with thousands of networks, translating to under 15ms latency to most major European tech hubs.

The fix

Independent, EU-owned bare metal dedicated servers hosted in Frankfurt give you both data residency and data sovereignty, since the entity controlling the hardware isn't subject to the CLOUD Act in the first place.

Full write-up, including a latency benchmark table and hardware checklist (NVMe, unmetered bandwidth, hardware DDoS protection, isolated VLANs), is here:

👉 Read the full guide

Singapore's Internet Infrastructure: The Southeast Asian Hub for APAC Tech Companies

Shannon Dias — Thu, 27 Aug 2026 07:49:55 +0000

Why Singapore, and why now

If you're deploying infrastructure anywhere near Southeast Asia, Singapore has probably already come up in your planning - and the numbers explain why.

Singapore is the landing point for 27 international submarine cables, roughly 29% of the 97 cables operating across ASEAN. New systems including Bifrost, Apricot, Candle, and Asia United Gateway East will keep extending that lead through 2029.

Peering keeps regional traffic local

Cable density is only half the story. Singapore also hosts SGIX and a dense private peering ecosystem operated by providers like Equinix and Digital Realty. That means traffic between, say, Indonesia and Thailand doesn't need to cross an international cable at all - it peers locally, which is why round trips complete in single-digit milliseconds instead of tens of milliseconds.

The latency numbers

Jakarta: 30–40ms
Bangkok: 35–45ms
Manila: 40–50ms
India: 40–70ms
Sydney: 50–70ms
Tokyo: 65–80ms

Overall, Singapore delivers sub-50ms latency to 600M+ people across Southeast Asia and reaches 2B+ users across broader APAC within 100ms.

Hyperscalers are voting with capital

AWS, Google, and Microsoft have committed a combined $20B+ to the region, most of it anchored around Singapore's core infrastructure - a signal that's hard to ignore if you're deciding where to build.

When alternatives make sense

Singapore isn't always the answer. Thailand and Malaysia's Johor region offer meaningfully lower costs, and Johor in particular connects via terrestrial fiber directly into Singapore's cable infrastructure. For cost-constrained early-stage deployments, that's a reasonable starting point - with a migration path to Singapore as traffic and revenue scale.

Full breakdown, latency tables, and regulatory comparisons (Singapore vs. Hong Kong):
👉 https://www.fitservers.com/blogs/singapore-internet-infrastructure/

Ashburn vs Dallas vs Chicago Best US Data Center Locations for Crypto Infrastructure

Shannon Dias — Fri, 14 Aug 2026 05:28:35 +0000

Unlike a typical website, blockchain infrastructure is latency-sensitive in specific ways: validator nodes need fast global peer sync, exchange-facing servers need the shortest possible round trip to liquidity venues, and RPC endpoints need to stay close to the users calling them.

Three US cities consistently come up when choosing where to host crypto infrastructure: Ashburn, VA, Dallas, TX, and Chicago, IL. Each has a distinct strength.

Ashburn, Virginia The Internet's Backbone

Ashburn sits in "Data Center Alley," the densest concentration of data centers in the world, with close to 50 million square feet of data center space and 4,900+ megawatts of commissioned power as of 2025. For crypto infrastructure, this means rich peering with other blockchain providers and low-hop routing across the global network.

Dallas, Texas The Geographic Center

Dallas wins through geography: its central US position gives near-equal latency to both coasts. The Infomart carrier hotel connects 140+ competing network providers, and Dallas's inland location keeps it clear of Gulf Coast hurricane risk a real plus for redundancy planning.

Chicago, Illinois The Financial Markets Hub

Chicago's edge is narrower but sharp: CME Group's Aurora, IL facility runs the matching engine behind regulated Bitcoin and Ether futures. If your infrastructure touches trading or derivatives, Chicago offers a measurable latency advantage the other two can't match.

Quick Take

Need global peer sync for a validator or full node? Ashburn.
Need balanced nationwide coverage and redundancy? Dallas.
Building trading infrastructure tied to CME? Chicago.

The full article includes a detailed comparison table and a checklist of infrastructure essentials (NVMe storage, unmetered bandwidth, DDoS protection, redundant power) for any crypto-facing server.

👉 Read more on the Fit Servers blog: https://www.fitservers.com/blogs/best-usa-dedicated-server-locations-crypto-infrastructure/

E-Commerce Hosting in Brazil: Compliance + Speed for Local Markets

Shannon Dias — Fri, 14 Aug 2026 04:32:06 +0000

Brazil is the largest e-commerce market in Latin America, growing faster than most mature markets in North America and Europe. For international retailers, that's the opportunity — but succeeding there depends on two things most launch checklists underrate: real-world latency for Brazilian shoppers, and whether your data handling holds up under LGPD.

What the guide covers

Brazil's network hubs — São Paulo, Rio de Janeiro, Fortaleza, and João Pessoa, and why physical distance to them directly affects checkout conversion
Dedicated vs. shared/cloud hosting for e-commerce specifically — isolated performance, NVMe storage for database-heavy catalogs, and hardware-level DDoS protection
LGPD in practice — lawful basis for processing, data subject rights, international transfer rules, and the Encarregado (data protection contact) requirement
The Marco Civil da Internet myth — clearing up the common but inaccurate claim that Brazilian law mandates in-country data storage
Pix and checkout performance — with instant payments now around 44% of online purchase value in Brazil, slow confirmation directly undermines conversion
A practical Brazil-ready hosting checklist covering location, connectivity, storage, DDoS protection, uptime SLAs, compliance support, and scalability

Full article here: https://www.fitservers.com/blogs/ecommerce-hosting-brazil/