Scaling MediaWiki
- Scaling MediaWiki: A High-Performance Server Configuration for Enterprise Knowledge Management
The deployment of MediaWiki for large-scale enterprise knowledge bases, internal documentation platforms, or high-traffic public wikis demands a meticulously engineered server configuration. This document details a high-availability, performance-optimized stack designed specifically to handle significant read/write loads, complex database queries, and large media file serving inherent to modern MediaWiki deployments (targeting MediaWiki 1.40+).
This architecture emphasizes low-latency storage, high core counts for concurrent PHP-FPM processing, and robust caching layers to ensure sub-second page load times even under peak load.
---
- 1. Hardware Specifications
The recommended configuration is built around a dual-socket server platform, leveraging modern CPU architectures for high Instructions Per Clock (IPC) and massive memory bandwidth, critical for database performance and PHP opcode caching.
- 1.1. Server Platform Baseline
The foundation is a 2U rackmount server chassis certified for high-density memory and PCIe lane availability.
| Component | Specification Detail | Rationale |
|---|---|---|
| Chassis Model | Dell PowerEdge R760 / HPE ProLiant DL380 Gen11 Equivalent | High expandability for NVMe and 100GbE networking. |
| Form Factor | 2U Rackmount | Balance between density and cooling effectiveness. |
| Power Supplies (PSU) | 2x 2000W Platinum Rated (Hot-Swappable) | N+1 redundancy required for high-load enterprise environments. Server Power Management |
| Baseboard Management | BMC/iDRAC/iLO 5.0+ | Essential for remote diagnostics and firmware updates. Remote Server Management Protocols |
- 1.2. Central Processing Units (CPUs)
MediaWiki processing (PHP execution, template rendering, parser operations) benefits significantly from high core counts and fast clock speeds. The configuration targets servers supporting 3rd Generation Xeon Scalable (Sapphire Rapids) or equivalent AMD EPYC processors.
| Parameter | Specification | Impact on MediaWiki |
|---|---|---|
| Model (Example) | 2x Intel Xeon Gold 6448Y (32 Cores / 64 Threads each) | Total 64 Cores / 128 Threads |
| Base Clock Speed | 2.5 GHz | Ensures consistent performance during sustained load. |
| Max Turbo Frequency | Up to 4.5 GHz | Crucial for burst performance during complex parser operations. CPU Scheduling for Web Servers |
| L3 Cache Size | 60 MB per CPU (Total 120 MB) | Reduces memory latency for frequently accessed PHP opcodes and database indexes. |
- 1.3. System Memory (RAM)
Memory capacity is paramount, as the entire structure of MediaWiki relies heavily on caching mechanisms: the PHP Opcode Cache (OPcache), MediaWiki's internal object caching (e.g., using Memcached or Redis), and most importantly, the database buffer pool (InnoDB/MariaDB). We specify DDR5 ECC RDIMMs for maximum bandwidth and reliability.
| Parameter | Specification | Notes |
|---|---|---|
| Total Capacity | 1.5 TB (Terabytes) | Allows for massive database buffer pools and extensive object caching. |
| Configuration | 12 x 128 GB DDR5-4800 ECC RDIMMs (Populating 12 channels across 2 sockets) | Optimizes memory interleaving and bandwidth utilization. DDR5 Memory Standards |
| Memory Type | DDR5 ECC RDIMM | Error Correction Code is mandatory for data integrity. |
- 1.4. Storage Subsystem (I/O Criticality)
The storage architecture must address two distinct I/O profiles: 1. **Database I/O:** High random read/write ($R/W$) latency sensitive (for transaction logs and index lookups). 2. **File/Attachment I/O:** High sequential throughput for serving large images and media files.
This configuration utilizes a tiered NVMe approach for the database and high-speed local SSDs for the MediaWiki application code and session data, offloading static assets to a dedicated Content Delivery Network (CDN) where possible, though local acceleration is configured.
| Tier | Device Type | Capacity | Interface/Protocol | Role |
|---|---|---|---|---|
| Tier 1 (Database Primary) | 4x 3.84 TB Enterprise NVMe SSD (e.g., U.2/M.2) | 15.36 TB Usable | PCIe Gen 4/5 x4 | MariaDB/MySQL Data Files and Binary Logs. Configured in RAID 10 via software or hardware controller for maximum IOPS and redundancy. Database High Availability |
| Tier 2 (OS/Application/Cache) | 2x 1.92 TB Enterprise SATA SSD | 3.84 TB Usable | SATA III 6Gbps | Operating System, PHP session files, OPcache persistence, and local object cache storage (if using APCu persistence). |
| Tier 3 (Local Media Cache) | 4x 7.68 TB SAS SSD | 30.72 TB Usable | SAS 12Gbps | Local storage for frequently accessed media files, serving as a fast fallback/L2 cache layer before CDN eviction. |
- 1.5. Networking Interface Cards (NICs)
High throughput is non-negotiable for serving large amounts of data and handling database replication traffic.
| Interface | Specification | Function |
|---|---|---|
| Primary Network | 2x 25GbE (SFP28) | Application traffic, Load Balancer communication. Configured in Active/Passive LACP bonding. Network Bonding Techniques |
| Backend/Replication Network | 1x 10GbE (SFP+) | Dedicated link for MariaDB replication streams and monitoring agents. |
| Management Network | 1x 1GbE (RJ45) | Dedicated for BMC/iDRAC/iLO access. |
---
- 2. Performance Characteristics
The performance profile of this MediaWiki stack is defined by its ability to manage concurrency (PHP-FPM processes) while minimizing the latency introduced by the database layer. Benchmarks below are illustrative of performance targets achievable with optimized software tuning (e.g., PHP 8.3, MariaDB 11.x, Varnish Cache 7.x).
- 2.1. Software Stack Tuning Highlights
- **PHP:** PHP-FPM configured with `pm = dynamic`, setting `pm.max_children` based on available RAM (e.g., 1000-1500 children, reserving 60% of RAM for the database buffer pool). OPcache configured with high memory limits (e.g., 1GB). PHP Performance Tuning
- **Database (MariaDB/Percona):** InnoDB Buffer Pool set to approximately 800GB (50% of total RAM). Query caching disabled in favor of application-layer caching. Database Buffer Pool Sizing
- **Caching Layer:** Varnish Cache (L7) deployed in front of Nginx/Apache, configured with a 100GB in-memory cache size, supplemented by Redis (L6) for session storage, locking, and complex object caching (e.g., parser output, session tokens). Web Caching Strategies
- 2.2. Benchmark Results (Targeted Performance)
The following table represents performance under a simulated load profile mimicking a medium-to-large enterprise wiki (e.g., 50,000 active pages, 100 concurrent editors, 500 concurrent readers).
| Metric | Target Value | Measurement Tool/Context |
|---|---|---|
| Average Read Latency (P95) | < 300 ms | Simulated 80% Read / 20% Write traffic via custom JMeter script. |
| Database Query Latency (P99) | < 15 ms | Measured directly from MariaDB slow query log analysis for complex article reads. |
| Transactions Per Second (TPS) | > 1,200 TPS | Sustained write operations (edits/uploads) over a 1-hour period. |
| Media Serving Throughput | > 5 GB/s | Measured via `iperf3` testing local SAS SSD tier serving compressed image assets. |
| PHP Execution Time (Average) | < 50 ms | Time to execute `index.php` without accessing the database (OPcache hit rate > 98%). PHP Opcode Caching |
- 2.3. Scalability Assessment
This configuration achieves vertical scaling limits primarily dictated by the memory bandwidth of the chosen CPU platform and the limits of the storage controller throughput.
- **Read Scaling:** Excellent. The large RAM pool ensures the working set of popular pages and database indexes remains resident in memory, minimizing disk access. The Varnish layer handles 80-90% of read requests without hitting the application server. Varnish Configuration for MediaWiki
- **Write Scaling:** Good, but limited by the single database instance. While synchronous replication can be added (see Section 4), the initial configuration relies on the speed of the NVMe RAID 10 array to absorb write spikes. Database Replication Topologies
---
- 3. Recommended Use Cases
This high-specification server configuration is optimized for environments where uptime, speed, and data integrity are paramount, and where the content management system (CMS) is a mission-critical asset.
- 3.1. Enterprise Internal Documentation Hubs
Organizations managing thousands of complex technical manuals, standard operating procedures (SOPs), and regulatory documents benefit immensely. The fast search indexing (Elasticsearch integration) and rapid rendering of complex templates (e.g., templates involving data tables or external data feeds) are supported by the high core count and massive RAM. MediaWiki Search Integration
- 3.2. High-Traffic Public Knowledge Bases
For public-facing wikis experiencing unpredictable traffic spikes (e.g., gaming wikis, large open-source project documentation), the combination of aggressive L7/L6 caching and fast storage prevents cascading failures during traffic surges. The 25GbE network interfaces ensure that the server can saturate the upstream network connection when serving cached content. CDN Integration Best Practices
- 3.3. Media-Intensive Deployments
Environments utilizing extensive media (high-resolution schematics, video transcripts, large image galleries) benefit from the dedicated local media cache tier (Tier 3 SAS SSDs). This configuration ensures that even if the primary CDN experiences a momentary lapse, local serving remains fast. MediaWiki File Management
- 3.4. Development and Staging Environments
While potentially over-provisioned for simple staging, this architecture is ideal for staging environments that must replicate production load profiles perfectly (Performance Testing). This ensures that performance regressions introduced during deployment are caught before hitting production. Staging Environment Synchronization
---
- 4. Comparison with Similar Configurations
To justify the investment in this high-end specification, it is crucial to compare it against more modest or horizontally scaled alternatives.
- 4.1. Comparison with Entry-Level Configuration
The Entry-Level configuration typically utilizes a single-socket modern CPU, 256GB RAM, and standard SATA SSDs, often relying solely on the database server for caching.
| Feature | High-Performance Scaling (This Document) | Entry-Level (Single Socket) |
|---|---|---|
| CPU Cores | 128 Threads (Dual Socket) | 16-24 Threads (Single Socket) |
| RAM Capacity | 1.5 TB DDR5 | 256 GB DDR4 |
| Storage Latency (DB) | Sub-millisecond NVMe RAID 10 | Millisecond-level SATA SSD RAID 10 |
| Max Concurrent Editors (Sustained) | > 100 | ~ 15-20 |
| Cache Layer Complexity | Varnish (L7) + Redis (L6) + OPcache | Basic Nginx caching only. |
| Cost Index (Relative) | 4.0x | 1.0x |
- 4.2. Comparison with Clustered/Horizontal Scaling Approach
A common alternative is horizontal scaling, using multiple smaller, cheaper nodes (e.g., 4x commodity servers) behind a load balancer. This trades complexity for raw power per node.
| Feature | Vertical Scaling (This Document) | Horizontal Scaling (4 Small Nodes) |
|---|---|---|
| Complexity Overhead | Low (Single OS, Single DB Management) | High (Distributed Cache Management, Load Balancer Tuning, Distributed Locks) Load Balancing Strategies |
| Database Bottleneck | Single Point of Failure (SPOF) for Writes (unless clustered) | Write latency is distributed across multiple small database instances (requires sharding or complex clustering). |
| Infrastructure Footprint (Rack Space) | 2U | 4U - 8U (4 Servers) |
| Network Dependency | Lower (Internal bus communication is fast) | Higher (Relies heavily on low-latency 25GbE switching between nodes) |
| Maximum Single-Node Performance | Very High (Leverages massive local memory bus) | Moderate (Limited by individual node resources) |
The vertical scaling approach described here is superior when the primary bottleneck is the single, massive database instance required by MediaWiki's relational structure, as scaling the database layer horizontally introduces significant application-level complexity (e.g., managing inter-wiki links across shards). Database Sharding Challenges
- 4.3. Impact of CPU Architecture on MediaWiki Performance
The choice of CPU directly impacts the execution rate of the MediaWiki parser, which is single-threaded for many operations but benefits greatly from parallelization in template expansion.
- **IPC (Instructions Per Clock):** Modern CPUs (like the Sapphire Rapids example) offer substantial IPC gains over older generations (e.g., Skylake), meaning higher clock speeds are less critical than efficient core design. This directly translates to faster page rendering. CPU Architecture Comparison
- **Memory Channels:** Dual-socket systems provide 12+ memory channels, ensuring the CPUs are never starved waiting for data from the 1.5TB RAM pool. This is crucial for avoiding stalls during large object retrieval from the database buffer pool. Memory Channel Utilization
---
- 5. Maintenance Considerations
While the performance is high, the complexity and density of this setup require specific maintenance protocols to ensure long-term stability and uptime.
- 5.1. Power and Cooling Requirements
High-density 2U servers equipped with dual high-TDP CPUs and numerous NVMe drives generate significant thermal output.
- **Power Density:** The combined peak draw for this system can exceed 1500W under sustained load. Data center racks must be Power Density rated (minimum 8kW per rack). Data Center Power Infrastructure
- **Cooling:** Requires high CFM (Cubic Feet per Minute) airflow supplied by the rack containment system. Standard office cooling is insufficient. Ensure the server room maintains a consistent temperature (Target: 18°C - 24°C). Server Thermal Management
- 5.2. Firmware Management and Lifecycle
The reliance on high-speed components (DDR5, PCIe Gen 5 NVMe) necessitates strict firmware management.
- **BIOS/UEFI:** Must be kept current to ensure the latest microcode fixes (especially Spectre/Meltdown mitigations) and optimal memory timing profiles are applied. Server Firmware Best Practices
- **Storage Controller Firmware:** Crucial for NVMe RAID performance and longevity. Firmware updates often require planned downtime. NVMe Drive Lifespan Management
- 5.3. Backup and Disaster Recovery (DR) Strategy
Given the consolidation of critical data (application, configuration, database) onto this single platform, a robust DR strategy is mandatory.
- **Database Backup:** Continuous Archiving (WAL shipping) to an offsite location is required for minimal Recovery Point Objective (RPO). Daily full snapshots of the Tier 1 NVMe array are also recommended. Database Backup Strategies
- **Application State:** The configuration files (`LocalSettings.php`, extensions, skins) must be version controlled (e.g., Git) and backed up nightly to the Tier 2 storage, then replicated offsite. Configuration Management for Wikis
- 5.4. Monitoring and Alerting
Effective monitoring is key to preempting failures before they impact service availability.
- **Hardware Health:** Monitor PSU voltage, fan speeds, and internal temperatures via the BMC interface at 1-minute intervals. Set alerts for any deviation outside 90% of nominal range. Hardware Monitoring Tools
- **I/O Saturation:** Monitor disk queue depth and IOPS separation between the database tier (Tier 1) and the media cache tier (Tier 3). Spikes in Tier 1 queue depth indicate a database bottleneck requiring immediate investigation (e.g., slow queries, missing indexes). Storage I/O Monitoring
- **Software Caching:** Monitor Varnish hit rates and Redis memory utilization. A sudden drop in Varnish hit rate signals a potential issue with cache invalidation logic or an upstream configuration change. Application Performance Monitoring (APM)
- 5.5. Security Hardening
The large memory footprint and high network connectivity require stringent security measures.
- **Kernel Security:** Utilize security modules like SELinux or AppArmor to enforce mandatory access controls on the PHP-FPM and MariaDB processes. Linux Security Modules
- **Network Segmentation:** The 25GbE application interfaces should be segmented from the 1GbE management network via physical separation or VLAN tagging, even if accessing the same physical switch infrastructure. Network Security Segmentation
- **Memory Protection:** Ensure kernel configurations enforce Address Space Layout Randomization (ASLR) aggressively to mitigate common exploitation vectors against PHP processes. Operating System Hardening
This detailed specification provides the blueprint for a MediaWiki deployment capable of sustaining enterprise-level demands for performance, reliability, and scale.
Intel-Based Server Configurations
| Configuration | Specifications | Benchmark |
|---|---|---|
| Core i7-6700K/7700 Server | 64 GB DDR4, NVMe SSD 2 x 512 GB | CPU Benchmark: 8046 |
| Core i7-8700 Server | 64 GB DDR4, NVMe SSD 2x1 TB | CPU Benchmark: 13124 |
| Core i9-9900K Server | 128 GB DDR4, NVMe SSD 2 x 1 TB | CPU Benchmark: 49969 |
| Core i9-13900 Server (64GB) | 64 GB RAM, 2x2 TB NVMe SSD | |
| Core i9-13900 Server (128GB) | 128 GB RAM, 2x2 TB NVMe SSD | |
| Core i5-13500 Server (64GB) | 64 GB RAM, 2x500 GB NVMe SSD | |
| Core i5-13500 Server (128GB) | 128 GB RAM, 2x500 GB NVMe SSD | |
| Core i5-13500 Workstation | 64 GB DDR5 RAM, 2 NVMe SSD, NVIDIA RTX 4000 |
AMD-Based Server Configurations
| Configuration | Specifications | Benchmark |
|---|---|---|
| Ryzen 5 3600 Server | 64 GB RAM, 2x480 GB NVMe | CPU Benchmark: 17849 |
| Ryzen 7 7700 Server | 64 GB DDR5 RAM, 2x1 TB NVMe | CPU Benchmark: 35224 |
| Ryzen 9 5950X Server | 128 GB RAM, 2x4 TB NVMe | CPU Benchmark: 46045 |
| Ryzen 9 7950X Server | 128 GB DDR5 ECC, 2x2 TB NVMe | CPU Benchmark: 63561 |
| EPYC 7502P Server (128GB/1TB) | 128 GB RAM, 1 TB NVMe | CPU Benchmark: 48021 |
| EPYC 7502P Server (128GB/2TB) | 128 GB RAM, 2 TB NVMe | CPU Benchmark: 48021 |
| EPYC 7502P Server (128GB/4TB) | 128 GB RAM, 2x2 TB NVMe | CPU Benchmark: 48021 |
| EPYC 7502P Server (256GB/1TB) | 256 GB RAM, 1 TB NVMe | CPU Benchmark: 48021 |
| EPYC 7502P Server (256GB/4TB) | 256 GB RAM, 2x2 TB NVMe | CPU Benchmark: 48021 |
| EPYC 9454P Server | 256 GB RAM, 2x2 TB NVMe |
Order Your Dedicated Server
Configure and order your ideal server configuration
Need Assistance?
- Telegram: @powervps Servers at a discounted price
⚠️ *Note: All benchmark scores are approximate and may vary based on configuration. Server availability subject to stock.* ⚠️