Cloud hosting for e-commerce in detail

Running an online shop requires a very serious approach to hosting. Your site generates your revenue, and many businesses can fail due to slow loading, poor security or unprofessional server management. To choose the right hosting provider, you should always consider your traffic, budget, sales volume and level of experience with server administration.

We sat down with one of our largest e-commerce clients to learn more about the challenges of cloud hosting and the solutions they found. Here's their experience.

Company profile

Our client is a Swiss company based in Geneva, specializing in selling luxury men's products worldwide. The company chose to develop their websites and their own ERP in-house, to have a solution tailored to their specific needs. The tech stack relies primarily on PHP and MySQL at all levels.

This choice of in-house development is common among e-commerce companies with specific requirements for workflow, inventory management or customer experience customization. But it also means a strong dependency on hosting infrastructure, which must be able to keep up with business growth.

Hosting requirements

For this company, the priorities are clear: reliable hosting with full control over servers, backed by a high level of customer service. In e-commerce, every second of loading time counts. A page that takes more than 3 seconds to load can lose up to 40% of its visitors. Reliability and performance are not optional, they are prerequisites.

Beyond raw performance, this client needed a hosting environment that could support multiple deployment workflows. Their development team pushes updates several times per week, including during peak sales periods. Any hosting solution had to accommodate frequent deployments without downtime, support rolling updates across application servers and provide instant rollback capabilities in case of issues. Traditional VM-based hosting, with its rigid change management processes, was fundamentally incompatible with this pace of delivery.

The struggles with previous hosting

Before joining Hidora, the company had been running dedicated VMs for several years. When their shop traffic started to increase, problems multiplied: frequent slowdowns and downtime.

The previous provider's managed service offered no flexibility in terms of server management. Debugging was a slow and painful process. Changes had to be requested through tickets, and even the smallest tasks took several hours. For an e-commerce site where every minute of downtime represents a direct revenue loss, this situation had become untenable.

Discovering Hidora

The team met Hidora at an IT convention in Geneva. After a demonstration of the solution, they were impressed by what the platform could offer and decided to move all their operations.

Several factors motivated this choice:

• Geographic proximity. Working with a local provider gives easy access to support and the people behind it. The Hidora team immediately convinced them with their level of expertise and the proposed architecture.
• Data sovereignty. Security and data protection are a top priority for a company handling luxury product transactions. Hosting data in Switzerland was essential.
• Technical flexibility. Full control over nodes, combined with the platform's ease of use, provides an independence that the former host couldn't deliver.

Migration to containers

The migration from VM-based infrastructure to container-based infrastructure was relatively quick. The team encountered some difficulties configuring servers to meet all their needs, but Hidora's technical department provided all the necessary support. Within a few days, all applications were migrated.

The result was an immediate increase in speed, uptime and user satisfaction. This kind of hands-on support is an integral part of our managed services.

Impact on costs

Although price wasn't the deciding factor for this company, the pay-per-use pricing model allowed them to cut their hosting costs in half. This is a significant advantage, especially in a sector where margins are closely watched.

The pay-as-you-go model means you only pay for resources actually consumed. During quiet periods (nights, weekends outside of sale seasons), costs decrease automatically. During traffic peaks (Black Friday, end-of-year holidays), the infrastructure scales up without manual intervention.

The DevOps contribution

Hidora helped the company set up a GitLab node and implement all the necessary workflows to streamline the development-to-production process. This type of DevOps support allows development teams to focus on code and ensure it runs smoothly on the servers they have access to.

This transition saved considerable time and eliminated the frustrations associated with the previous hosting model.

Performance optimization for e-commerce

Beyond the infrastructure migration, several performance optimizations were implemented that had a direct impact on conversion rates and user experience.

Caching strategy. A multi-layered caching approach was put in place, combining Varnish for full-page caching, Redis for session and object caching, and browser-level caching with optimized headers. This reduced the average page load time from 2.8 seconds to under 1.2 seconds, a difference that translates directly into higher conversion rates.

Database optimization. The MySQL databases were migrated to a primary-replica topology with read replicas handling catalogue browsing queries. This offloaded the primary database, which now handles only writes and transactional queries (cart operations, order processing). During peak events like Black Friday, the read replicas absorb the surge in browsing traffic without any degradation to checkout performance.

Image delivery. Product images, which account for the majority of page weight on any e-commerce site, were moved to a dedicated storage node with automatic WebP conversion and responsive image serving. This alone reduced page payload by over 40% for mobile users.

These optimizations are not one-time efforts. The Hidora team monitors performance metrics continuously and recommends adjustments as the catalogue grows and traffic patterns evolve.

Scaling Strategies for Peak Traffic

E-commerce traffic is inherently unpredictable. A successful marketing campaign, a mention by an influencer, or a flash sale can multiply traffic by 5x or 10x within minutes. Traditional infrastructure planning, where you provision servers weeks in advance, simply cannot handle this reality.

For this client, we implemented a multi-layered scaling strategy that has proven itself across several Black Friday and holiday seasons.

Horizontal auto-scaling for application servers. The platform automatically adds application nodes when CPU or memory utilization exceeds defined thresholds. During Black Friday 2024, the client's infrastructure scaled from 4 application nodes to 12 within 15 minutes of the sale going live, then gradually scaled back down over the following 48 hours. No manual intervention was required, and no customer experienced degradation.

Database connection pooling. A common bottleneck during traffic spikes is not the application layer but the database. Each new application node opens dozens of database connections, and MySQL has a hard limit on concurrent connections. We deployed ProxySQL as a connection pooler between the application servers and the database cluster. This reduced the number of actual database connections by 80% while supporting 5x more concurrent users.

Pre-warming for planned events. While auto-scaling handles unexpected surges, planned events like Black Friday benefit from pre-warming. The day before a major sale, the infrastructure is scaled to a baseline that matches the expected initial load. This eliminates the 2-3 minute delay that auto-scaling requires to detect the surge and spin up new nodes. For a luxury e-commerce site where the first minutes of a sale drive a disproportionate share of revenue, those 2-3 minutes matter.

Geographic load distribution. Since this client sells worldwide, a CDN (Content Delivery Network) was configured for all static assets. Product images, CSS, JavaScript, and fonts are served from edge locations close to the customer, reducing latency for international buyers. The origin servers in Switzerland handle only dynamic content (pricing, cart operations, checkout), keeping the load manageable even during global traffic peaks.

Graceful degradation under extreme load. Even with auto-scaling, there are scenarios where traffic exceeds all projections. Rather than letting the site crash, we implemented circuit breakers that progressively disable non-essential features under extreme load. Product recommendations, recently viewed items, and live chat are the first to be throttled. The core purchasing flow (browse, add to cart, checkout) is always protected. This approach ensures that revenue-generating functionality remains available even in worst-case scenarios.

Technical architecture

This e-commerce client's typical environment includes:

• Load balancers to distribute traffic
• Application servers to run business logic
• Cache and storage nodes to optimize performance
• Database servers and SQL proxy servers
• Over 15 environments in total, including GitLab, backup and staging environments

Swiss e-commerce specific requirements

The Swiss e-commerce market has specific characteristics that directly influence hosting choices. First, compliance with the LPD (Swiss Federal Data Protection Act) imposes strict guarantees on where customer data is stored and processed, especially for payment transactions. A hosting provider that stores data outside Switzerland exposes the merchant to growing legal risks.

Second, Swiss consumers expect fast loading times and impeccable availability. Cart abandonment rates increase by 7% for every additional second of loading time. For a site generating CHF 500,000 in annual revenue, an extra second of latency can represent a loss of CHF 35,000 per year.

Third, seasonal peaks (Black Friday, Christmas, sales) require infrastructure that can scale rapidly. Traditional hosting with dedicated servers cannot absorb a 5x to 10x traffic spike without manual planning weeks in advance. Auto-scaling solves this problem transparently.

Three platform highlights

The client highlighted three key advantages:

1. Control over nodes. The independence this provides is invaluable for a technical team that knows what they want.
2. Ease of use. Setting up a new environment and being in production in no time is a game changer for time-to-market.
3. Active support. The platform is actively maintained and supported, and client feedback is taken into account in product evolution.

For e-commerce companies looking for performant, secure and sovereign cloud hosting, this client's experience perfectly illustrates what Hidora can deliver. Feel free to explore our cloud offerings or contact us to discuss your project.