The performance question most hosting decisions get wrong
When choosing hosting, most businesses compare advertised uptime percentages and basic response times. But these metrics hide the performance characteristics that determine whether your application handles real production load successfully.
We measured 8 hosting configurations under identical conditions to understand what actually happens when traffic patterns shift, when databases work harder, and when systems experience the kind of load that reveals infrastructure quality.
The gap between marketing claims and measured performance affects more than just page speed. Poor p95 and p99 response times translate directly into abandoned shopping carts, failed API calls, and user frustration that costs revenue. Understanding these numbers helps engineering teams make hosting decisions based on production reality rather than synthetic benchmarks.
Methodology: controlled testing across hosting types
We deployed identical WordPress applications with WooCommerce across 8 different hosting configurations, ranging from shared hosting to dedicated managed infrastructure. Each environment ran the same application stack to isolate infrastructure variables from application performance differences.
Hardware and software specifications
Every test environment used:
- WordPress 6.4.2 with identical plugins and themes
- WooCommerce 8.3.1 with 10,000 sample products
- MySQL 8.0 with consistent configuration
- PHP 8.2 with identical memory limits and extensions
- Same SSL certificates and CDN configuration disabled
The hosting configurations tested included:
- Shared hosting (major European provider)
- VPS with 4 CPU cores, 8GB RAM
- Cloud hosting (AWS t3.large equivalent)
- Managed WordPress hosting (premium tier)
- Dedicated server (8 cores, 32GB RAM)
- Container-based hosting
- Managed infrastructure with performance tuning
- High-availability setup with load balancing
Load testing approach
We used Apache JMeter to simulate realistic traffic patterns over 72-hour periods. The load profile included:
- Baseline: 50 concurrent users browsing products
- Peak traffic: 300 concurrent users during 2-hour windows
- Database-heavy operations: search queries, cart updates, checkout processes
- Mixed content types: static assets, dynamic pages, API calls
Each test measured response times, throughput, error rates, and resource utilization every 30 seconds. We focused on p50, p95, and p99 percentiles because these reveal how systems perform under stress, not just average conditions.
Results: performance gaps under real load
The measurements revealed significant differences between hosting types, particularly under sustained load. Here are the key metrics that matter for production applications:
Response time percentiles (milliseconds)
| Hosting Type | p50 Response | p95 Response | p99 Response | Error Rate |
|---|---|---|---|---|
| Shared Hosting | 2,400ms | 8,900ms | 15,200ms | 4.2% |
| Basic VPS | 1,100ms | 3,800ms | 7,100ms | 1.8% |
| Cloud Instance | 950ms | 2,900ms | 5,400ms | 1.2% |
| Managed WordPress | 800ms | 2,200ms | 4,100ms | 0.9% |
| Dedicated Server | 420ms | 1,100ms | 2,300ms | 0.3% |
| Container Platform | 380ms | 980ms | 1,900ms | 0.2% |
| Managed Infrastructure | 290ms | 650ms | 1,200ms | 0.1% |
| High-Availability Setup | 310ms | 580ms | 950ms | 0.0% |
The performance differences become more pronounced at higher percentiles. While shared hosting might handle average requests reasonably, the p99 response times show that 1% of users wait over 15 seconds for pages to load. This directly impacts conversion rates and user experience.
Throughput under sustained load
Peak throughput measurements revealed how each hosting type handles traffic spikes:
- Shared hosting: 45 requests/second maximum before degradation
- Basic VPS: 120 requests/second sustained
- Cloud instance: 180 requests/second with auto-scaling
- Managed WordPress: 250 requests/second with optimized caching
- Dedicated server: 420 requests/second with proper tuning
- Container platform: 580 requests/second with load balancing
- Managed infrastructure: 750 requests/second with performance optimization
- High-availability setup: 850 requests/second with failover capability
These numbers reflect real-world performance under the mixed load patterns that production applications experience. Pure synthetic benchmarks often show higher theoretical maximums that don't reflect actual user behavior.
Database performance patterns
Database query performance varied significantly across hosting types. We measured complex WooCommerce queries that involve multiple table joins and product filtering:
- Shared hosting: Average query time 340ms, with frequent timeouts during peak load
- Basic VPS: Average query time 120ms, consistent performance
- Managed infrastructure: Average query time 35ms, with query optimization and proper indexing
The database performance differences compound as traffic increases. Slow database queries create cascading effects that impact every aspect of application performance, from page generation to API response times.
Analysis: what these numbers mean in production
The performance measurements translate directly into business outcomes that engineering teams need to consider when making infrastructure decisions.
User experience impact
Response time percentiles reveal the experience different users have with your application. While p50 times show what typical users experience, p95 and p99 times show what happens to users during peak load or when systems are under stress.
Research consistently shows that response times above 3 seconds lead to significant user abandonment. In our tests, shared hosting exceeded this threshold for 5% of requests even under moderate load. During traffic spikes, this percentage increases dramatically.
The managed infrastructure configurations maintained sub-second response times for 95% of requests even under sustained load. This consistency matters more than average response times because user experience is determined by the slowest interactions, not the fastest ones.
For e-commerce applications, every second of additional load time reduces conversions. The difference between 1-second and 3-second page loads isn't just about user satisfaction – it directly affects revenue.
Scalability characteristics
The throughput measurements show how different hosting approaches handle growth. Shared hosting reaches capacity quickly and degrades poorly under increased load. Once the threshold is exceeded, response times increase exponentially rather than linearly.
Dedicated servers and managed infrastructure services show more predictable scaling characteristics. Performance remains consistent as load increases, up to well-defined capacity limits. When those limits are reached, the systems can be scaled horizontally or vertically without architectural changes.
Container-based and high-availability setups demonstrated the most graceful scaling behavior. Instead of hitting hard capacity walls, these configurations maintain performance levels by distributing load across multiple resources.
Reliability under stress
Error rates reveal how hosting infrastructure behaves when pushed beyond comfortable operating parameters. Shared hosting showed error rates above 4% during peak testing periods. These errors manifest as failed page loads, incomplete transactions, and timeout errors that users experience directly.
The managed infrastructure and high-availability configurations maintained error rates below 0.1% even under sustained heavy load. This reliability difference becomes crucial for applications that cannot afford to lose transactions or user sessions.
Database connection limits, memory constraints, and CPU throttling affect different hosting types differently. Properly configured database systems handle connection pooling and query optimization transparently, while basic hosting often exposes these limitations directly to applications.
Caveats and what we'd do differently
These measurements provide useful comparisons, but several factors limit their applicability to specific production environments.
Application-specific variables
We tested WordPress with WooCommerce because it represents a common, resource-intensive application pattern. However, different applications have different performance characteristics. Single-page applications with API backends will show different patterns than server-rendered applications.
Custom applications with optimized database schemas, efficient caching strategies, and minimal plugin overhead will perform better across all hosting types. The relative differences between hosting configurations might be smaller for highly optimized applications.
Static site generators and CDN-heavy architectures reduce the impact of hosting performance differences. Applications that serve mostly static content will see less variation between hosting providers than dynamic, database-driven applications.
Geographic and network factors
Our tests used servers located in the same European data center to control for network latency variables. Production applications serve users from different geographic locations, and network performance varies significantly between providers.
CDN configuration, edge caching, and regional server distribution affect user experience more than raw hosting performance for globally distributed applications. The hosting performance differences we measured matter most for users geographically close to the servers.
Network reliability, peering relationships, and bandwidth allocation differ between providers in ways that our controlled testing didn't capture. Real-world network conditions include packet loss, routing inefficiencies, and congestion that affect application performance.
Testing methodology limitations
Synthetic load testing doesn't perfectly replicate real user behavior patterns. Actual users browse unpredictably, abandon sessions partway through, and generate load spikes that don't follow consistent patterns.
We focused on HTTP response times and throughput metrics, but didn't measure factors like SSL handshake performance, DNS resolution times, and connection establishment overhead. These factors contribute to overall user experience but require different measurement approaches.
The 72-hour testing window captured performance during normal operating conditions but didn't test behavior during infrastructure maintenance, security updates, or hardware failures. Long-term reliability requires extended observation periods.
Configuration optimization differences
Each hosting environment used default or recommended configurations from the providers. In production, experienced teams optimize database settings, caching layers, and application configurations specifically for their workloads.
The performance gaps between hosting types would be smaller with expert optimization applied consistently across all environments. However, the managed infrastructure and high-availability configurations included professional optimization as part of the service, which reflects their real-world advantage.
Some hosting types have optimization potential that our testing didn't explore. Dedicated servers can be tuned extensively for specific applications, potentially achieving performance that exceeds our measurements.
Takeaways for infrastructure decisions
These measurements provide engineering teams with data points for making informed hosting decisions based on actual performance requirements rather than marketing claims.
Response time percentiles matter more than averages because user experience is determined by the slowest interactions. Applications that require consistent performance should evaluate p95 and p99 response times under realistic load conditions.
Throughput capacity determines how applications handle growth and traffic spikes. Understanding these limits helps teams plan for scaling and avoid performance surprises during important business periods.
Error rates under stress reveal infrastructure reliability characteristics that affect user experience and business operations. Systems that maintain low error rates during peak load provide better foundation for growing applications.
The performance differences between hosting types are significant and measurable. These differences translate directly into user experience and business outcomes that justify infrastructure investment decisions.
Managed infrastructure services provide consistent performance advantages across all measured metrics. The optimization, monitoring, and proactive management included with these services deliver measurable improvements that basic hosting cannot match.
Want these kinds of numbers for your own stack? Request a performance audit.
