Network Performance Testing Results & Analysis

Overview

This document chronicles comprehensive network performance testing conducted to establish baseline performance before implementing a 10GbE lab network upgrade. Testing revealed significant network topology insights and performance bottlenecks that will guide the upgrade strategy.

Testing Environment

Hardware Under Test

  • NAS: Synology DS918+ (192.168.10.5)
  • Test Client 1: Laptop via WiFi (192.168.6.3 → 192.168.2.12)
  • Test Client 2: Jumpbox via Cat6 (192.168.100.10)
  • Target: Synology DS918+ running iperf3 in Docker container

Network Topology Discovered 

Internet → Router → Multiple VLANs/Subnets:
├── 192.168.2.x   (Laptop WiFi - with routing overhead)
├── 192.168.6.x   (Laptop WiFi with VPN)
├── 192.168.10.x  (NAS subnet)
└── 192.168.100.x (Jumpbox - direct switched connection)

Test Results Summary

Test Source Connection Type Speed (Mbps) Performance Level Network Path
Laptop + VPN WiFi 21 ❌ Severely Limited VPN bottleneck
Laptop WiFi 155 ⚠️ Bottlenecked Cross-subnet routing
Laptop Multi-stream WiFi 124 ⚠️ Bottlenecked Cross-subnet routing
Jumpbox Cat6 Wired 875 ✅ Excellent Direct switched

Detailed Test Results

Test 1: Initial Baseline (With VPN Active)

Configuration: Laptop → WiFi → VPN → NAS Command: iperf3 -c carbonite.markalston.net -p 5201

Client: 192.168.6.3 → Server: 192.168.10.5
Average Speed: 21.1 Mbps (20.8 Mbps received)
Performance: SEVERELY LIMITED

Key Issues Identified:

  • VPN encryption/routing overhead
  • Multiple network hops
  • Potential VPN server bottleneck

Test 2: VPN Disabled

Configuration: Laptop → WiFi → NAS (no VPN) Command: iperf3 -c carbonite.markalston.net -p 5201

Client: 192.168.2.12 → Server: 192.168.10.5
Average Speed: 155 Mbits/sec (152 Mbits/sec received)
Performance: MODERATE - Well below 1GbE potential

Analysis:

  • ✅ 7x improvement over VPN test
  • ❌ Still only 16% of 1GbE theoretical maximum
  • Cross-subnet routing overhead identified
  • Different subnets requiring router/VLAN traversal

Test 3: Multi-Stream Performance

Configuration: Laptop → WiFi → NAS (4 parallel streams) Command: iperf3 -c carbonite.markalston.net -p 5201 -P 4 -t 30

Client: 192.168.2.12 → Server: 192.168.10.5
Total Bandwidth: 125 Mbits/sec (124 Mbits/sec received)
Individual Streams: 24-35 Mbits/sec each
Performance: BOTTLENECKED - Lower than single stream

Key Findings:

  • Multi-stream performance actually decreased
  • Indicates network infrastructure limitation, not single-stream TCP limitation
  • Confirms routing/switching bottleneck rather than protocol issue

Test 4: Bidirectional Testing

Configuration: Laptop → WiFi → NAS (reverse direction) Command: iperf3 -c carbonite.markalston.net -p 5201 -R -t 10

Download (NAS → Laptop): 192 Mbits/sec (183 Mbits/sec received)
Upload (Laptop → NAS): 155 Mbits/sec
Asymmetry: Download ~25% faster than upload
Retransmissions: 349 (indicating some congestion)

Analysis:

  • Slight asymmetry in performance
  • Download marginally better than upload
  • Retransmissions suggest network congestion under load

Test 5: Direct Wired Connection (Jumpbox)

Configuration: Jumpbox → Cat6 → Switch → NAS Command: iperf3 -c carbonite.markalston.net -p 5201

Client: 192.168.100.10 → Server: 192.168.10.5
Average Speed: 875 Mbits/sec (873 Mbits/sec received)
Performance: EXCELLENT - 93% of 1GbE theoretical maximum
Retransmissions: 693 (normal under high throughput)

Outstanding Results:

  • ✅ Proper 1GbE performance achieved
  • ✅ 93% efficiency - excellent for real-world conditions
  • ✅ Expected file transfer: ~109 MB/s
  • ✅ Proves network infrastructure capability

Performance Analysis & Network Topology Insights

Identified Network Bottlenecks

  1. Cross-Subnet Routing Overhead
    • Laptop (192.168.2.x) requires routing to reach NAS (192.168.10.x)
    • ~83% performance penalty due to VLAN/routing traversal
    • Multiple network hops introduce latency and processing overhead
  2. VPN Impact
    • VPN reduces performance by ~93% (21 Mbps vs 155 Mbps)
    • Encryption overhead and additional network hops
    • Critical to disable VPN for accurate baseline testing
  3. Direct Switched Connection Benefits
    • Jumpbox (192.168.100.x) achieves full 1GbE performance
    • Same network segment or optimized routing to NAS
    • Demonstrates infrastructure’s true capability

Current Network Performance Summary

Connection Type Speed File Transfer Rate Efficiency
WiFi + VPN 21 Mbps ~2.6 MB/s 2% of 1GbE
WiFi (Cross-subnet) 155 Mbps ~19 MB/s 16% of 1GbE
Wired (Direct) 875 Mbps ~109 MB/s 93% of 1GbE

10GbE Upgrade Impact Projections

Current Planned Upgrade

  • UCG-Fiber Gateway: Advanced routing with 2.5GbE + 10GbE SFP+ ports
  • Synology E10G18-T1: 10GbE adapter for DS918+ NAS
  • Phase 2: USW Pro XG 8 PoE switch for full lab 10GbE

Expected Performance Improvements

From Laptop (Worst Case Scenario) 

Current:    155 Mbps  →  Expected: 8,500+ Mbps  (55x improvement)
Transfer:   19 MB/s   →  Expected: 1,000+ MB/s  (53x improvement)

From Jumpbox (Best Case Scenario) 

Current:    875 Mbps  →  Expected: 8,500+ Mbps  (10x improvement)
Transfer:   109 MB/s  →  Expected: 1,000+ MB/s  (9x improvement)

Why the Massive Laptop Improvement?

The UCG-Fiber with direct 10GbE connection will likely bypass the current cross-subnet routing bottleneck that’s limiting the laptop to 155 Mbps, providing direct high-speed access.

Testing Methodology & Tools

iperf3 Configuration

  • Server Setup: Docker container on Synology DS918+
  • Docker Command: networkstatic/iperf3 -s
  • Port: 5201 (default)
  • Test Duration: 10-30 seconds per test
  • Multiple Streams: 4 parallel connections tested

Test Commands Used 

# Basic throughput test
iperf3 -c carbonite.markalston.net -p 5201

# Multi-stream test
iperf3 -c carbonite.markalston.net -p 5201 -P 4 -t 30

# Reverse direction test
iperf3 -c carbonite.markalston.net -p 5201 -R -t 10

Environment Variables

  • VPN Impact: Tested both with/without VPN
  • Network Segmentation: Multiple subnets identified during testing
  • Connection Types: WiFi vs wired performance comparison

Key Findings & Recommendations

Critical Discoveries

  1. Network topology has significant performance impact
    • Cross-subnet routing reduces performance by 80%+
    • Direct switched connections achieve full 1GbE potential
  2. Current infrastructure is capable
    • 875 Mbps proves network backbone can handle high throughput
    • NAS can sustain near-1GbE speeds consistently
  3. 10GbE upgrade will be transformational
    • Expected 10-55x performance improvement depending on connection path
    • Direct 10GbE connection should bypass current routing limitations

Recommendations

  1. Proceed with UCG-Fiber + E10G18-T1 Phase 1
    • Immediate ~10x improvement for direct connections
    • Should resolve cross-subnet routing bottleneck
  2. Post-upgrade testing plan
    • Retest from laptop to verify routing improvement
    • Retest from jumpbox to measure pure 10GbE performance
    • Target: 8,000+ Mbps (8+ Gbps) throughput
  3. Future network optimization
    • Consider network topology consolidation
    • Evaluate VLAN/subnet strategy for optimal performance

Baseline Established

This testing provides excellent baseline measurements for comparison after the 10GbE upgrade implementation. The dramatic difference between cross-subnet (155 Mbps) and direct-switched (875 Mbps) performance clearly demonstrates both the current limitations and the massive potential for improvement with the planned infrastructure upgrade.

Next Steps:

  1. Implement UCG-Fiber + E10G18-T1 upgrade
  2. Repeat identical tests from both laptop and jumpbox
  3. Document performance improvements
  4. Plan Phase 2 switch upgrade based on results

Testing completed: Current date Document prepared for 10GbE lab network upgrade planning

This project is for educational and home lab purposes.