← Back to homepage

Examples of Hardware Configuration

Traffic CPU RAM OS Cache Storage Network Max Connections
1 Gbps 8 cores 32 GB SSD 128 GB 1x SATA SSD 1TB 2x 1GbE ~50,000
10 Gbps 16 cores 64 GB SSD 256 GB 2x NVMe Gen3 2TB 2x 10GbE ~200,000
25 Gbps 32 cores 128 GB 2x256 GB RAID 4x NVMe Gen3 4TB 2x 25GbE ~500,000
50 Gbps 64 cores 256 GB 2x512 GB RAID 8x NVMe Gen4 4TB 4x 25GbE bond ~1,000,000
100 Gbps 128 cores 512 GB 2x1 TB RAID 16x NVMe Gen5 8TB 2x 100GbE ~2,000,000
Note: These configurations are provided as examples only. Actual requirements depend on your cache hit ratio, object sizes, and traffic patterns. Higher hit ratios reduce disk I/O requirements significantly.

Disk I/O Calculations

Formula (cache MISS scenario):
Disk Write I/O = Network Throughput × (1 - Hit Ratio)
Disk Read I/O = Network Throughput × Hit Ratio
Traffic Network Throughput 50% Hit Rate 80% Hit Rate 95% Hit Rate
1 Gbps 125 MB/s 62 MB/s R + 62 MB/s W 100 MB/s R + 25 MB/s W 119 MB/s R + 6 MB/s W
10 Gbps 1.25 GB/s 625 MB/s R + 625 MB/s W 1 GB/s R + 250 MB/s W 1.19 GB/s R + 62 MB/s W
25 Gbps 3.1 GB/s 1.55 GB/s R + 1.55 GB/s W 2.5 GB/s R + 625 MB/s W 2.95 GB/s R + 155 MB/s W
50 Gbps 6.25 GB/s 3.1 GB/s R + 3.1 GB/s W 5 GB/s R + 1.25 GB/s W 5.9 GB/s R + 312 MB/s W
100 Gbps 12.5 GB/s 6.25 GB/s R + 6.25 GB/s W 10 GB/s R + 2.5 GB/s W 11.9 GB/s R + 625 MB/s W

Disk Performance

Type Seq. Read Seq. Write IOPS Best For Status
HDD 7200 RPM 150 MB/s 130 MB/s ~150 Cold storage only Not recommended
SATA SSD 550 MB/s 500 MB/s ~90,000 Up to 1 Gbps Entry level
NVMe Gen3 3.5 GB/s 3 GB/s ~500,000 Up to 25 Gbps Recommended
NVMe Gen4 7 GB/s 6.5 GB/s ~1,000,000 Up to 50 Gbps Recommended
NVMe Gen5 14 GB/s 12 GB/s ~1,500,000 100 Gbps+ High performance

Cache Hit Ratio Impact

Higher cache hit ratios dramatically reduce bandwidth to origin servers and improve response times.

High Hit Ratio (90%+)

Origin bandwidth savings: 90%+
10x

Ideal for static content, CDN files, software updates

Medium Hit Ratio (60-80%)

Origin bandwidth savings: 60-80%
3-5x

Mixed content with some dynamic pages

Low Hit Ratio (30-50%)

Origin bandwidth savings: 30-50%
1.5-2x

Many unique URLs or short TTLs

Streaming/Gaming (95%+)

PlayStation, Xbox, Steam, Windows Update
20x+

Large files with slice-based caching

Real-world example: With 80% cache hit ratio, a 10 Gbps link only requires 2 Gbps to origin servers. The remaining 8 Gbps is served directly from cache at sub-millisecond latency.

Multi-Level Cache Architecture

Level Scope Function Size Latency
L1 - Memory Cache Hot objects Small frequently accessed objects (<128KB) 100 MB - 1 GB < 0.1 ms
L2 - NVMe Cache Warm objects Large objects, sliced files 1 TB - 100 TB ~0.5 ms
L3 - Sibling Cache Peer nodes Objects from other WebProxyCache nodes Distributed ~1-5 ms
L4 - Origin Internet Fetch from origin server Unlimited 10-200 ms

Scaling Tips

Vertical Scaling

  • Add RAM for memory cache
  • Upgrade to faster NVMe drives
  • Upgrade NIC (10G → 25G → 100G)
  • More CPU cores for concurrent connections
  • Enable compression (Brotli/Gzip)

Horizontal Scaling

  • Sibling cache peering between nodes
  • ECMP/PCC load balancing on router
  • Consistent hashing for cache distribution
  • Upstream proxy chaining
  • Geo-distributed deployments

Connection & Resource Limits

Resource Default Tuned Notes
File Descriptors (ulimit -n) 1,024 1,000,000 Auto-increased at startup
TCP Connections (net.core.somaxconn) 4,096 65,535+ Auto-tuned based on RAM
Conntrack Max (nf_conntrack_max) 262,144 2,000,000+ For TPROXY mode
Socket Buffers 128 KB 4 MB Optimized for high throughput
MySQL Semaphore - 50 concurrent Prevents goroutine accumulation
Auto-tuning: WebProxyCache automatically adjusts kernel parameters at startup based on available RAM and CPU cores. No manual sysctl configuration required.