Network Checklist
Network checklist for enterprise deployments.
Network Architecture
Planning
- Topology Verify the architecture (star/ring/mesh) meets high availability requirements.
- Redundancy Ensure at least two independent physical links between nodes.
- Bandwidth Estimate traffic as: read/write bandwidth × node count × replica count, plus 20% headroom.
IP Allocation
- Separate management network from data network.
- Allocate a contiguous IP address range for storage nodes. We recommend a /24 subnet.
- Reserve at least 15% of IP addresses for expansion.
Hardware Requirements
Switches
| Item | Requirement |
|---|---|
| Backplane bandwidth | ≥ 1.2 × aggregate line rate of all ports |
| Port Type | 10G/25G/100G SFP+/QSFP+ fiber ports |
| Flow Table Capacity | ≥ Node count × 5 |
| Spanning Tree | Enable RSTP or MSTP for fast convergence |
Cabling
- Fiber attenuation test (single mode ≤0.35dB/km).
- Verify port mapping (Node A eth0 ↔ Node B eth0).
- Cable labeling (including source/destination IP + port number).
OS Configuration
Kernel Tuning
# Check the following parameter settings
net.core.rmem_max = 16777216
net.core.wmem_max = 16777216
net.ipv4.tcp_keepalive_time = 600
net.ipv4.tcp_slow_start_after_idle = 0NIC Configuration
- Enable jumbo frames (MTU=9000, requires full path support).
- Verify bonding mode (recommend LACP mode4).
- Disable IPv6 (if not needed).
Security
Firewall
# Necessary open ports
- TCP 9000 (S3 API; also used for internal node-to-node communication)
- TCP 9001 (RustFS Console)
- TCP 443 (only if you terminate HTTPS on a reverse proxy / load balancer)Access Control
- Enable port security (MAC address limits) on switches.
- Encrypt traffic between storage nodes with IPsec.
- Enable TLS 1.3 for management interfaces.
Performance Testing
Benchmarks
- Inter-node latency:
ping -c 20 <target IP>(runiperf3 -son the target first for throughput tests) - Cross-rack bandwidth:
iperf3 -c <target IP> -P 8 -t 30 - Failover: Randomly disconnect core links to observe recovery time.
Criteria
| Metric | Requirements |
|---|---|
| Node latency | ≤ 1 ms (same data center) / ≤ 5 ms (cross-AZ) |
| Bandwidth utilization | Peak ≤ 70% of design capacity |
| Failover | < 500ms BPDU convergence |
Documentation Requirements
- Network topology diagram (including physical connections and logical IPs)
- Switch configuration backup files (with timestamps)
- Baseline test reports (including raw data)
- Change log (including maintenance windows)
We recommend running a 72-hour stress test at 110% of the expected peak traffic before deploying to production.
This checklist covers the key network requirements for deploying distributed object storage in an enterprise environment. For technical support, contact RustFS.