Every time I design a multi-tier cloud VPC, debug a hung WireGuard tunnel, or compress sprawling firewall rules, I end up needing quick subnet or MTU math.
Most calculators on page one of search results haven't changed since 2008: they assume classic RFC -2 host math, ignore cloud provider reservations, or upload pasted JWT bearer tokens and configs to a backend server.
Here are 3 networking math traps that regularly bite engineers in production—plus NetDevKit, a fast, 100% client-side browser workbench I built to solve them offline.
1. The AWS & Azure /28 Subnet Surprise (11 IPs, Not 14)
Under classic RFC 950 subnetting, a /28 block (16 total IPv4 addresses) gives you 14 usable hosts. Cloud VPCs reserve extra control-plane addresses inside every subnet:
| Environment | Reserved IPs | Usable Hosts in /28 | What's Reserved? |
|---|---|---|---|
| Standard RFC | 2 | 14 | .0 Network, .15 Broadcast |
| Google Cloud (GCP) | 4 | 12 | Network, .1 Gateway, Penultimate IP, Broadcast |
| AWS VPC / Azure VNet | 5 | 11 | .0 Net, .1 Router, .2 DNS, .3 Future, Broadcast |
During rolling Kubernetes node upgrades or multi-AZ RDS failovers, those missing IPs routinely trigger InsufficientFreeAddressesInSubnet errors.
Try it live (pre-filled
/28in AWS VPC mode + Terraform snippet):
IPv4 Subnet & Cloud VPC Calculator →
2. WireGuard Over PPPoE Hanging on TLS ClientHello
ping works across the VPN tunnel, SSH connects, and then curl https://... or git clone freezes indefinitely.
That is a classic Path MTU Discovery (PMTUD) blackhole:
- Standard Ethernet (
1500) minus IPv6 WireGuard overhead (80bytes) =1420MTU - Behind residential PPPoE (
8bytes overhead,1492WAN MTU), your safe ceiling drops to1412bytes (TCP MSS = 1372for IPv4 /1352for IPv6)
When upstream firewalls drop ICMP Fragmentation Needed packets, large TLS handshakes stall silently unless you clamp TCP MSS on forward chains:
nft add rule inet filter forward tcp flags syn tcp option maxseg size set rt mtu Calculate exact tunnel overhead & split-tunnel configs:
3. Firewall Exact Merging vs. BGP Supernets
Summarizing routes for BGP/OSPF is very different from compressing CIDR lists to fit inside AWS Security Group quotas (60 rules per group by default):
- Firewall / Security Group compression requires Exact Merge (
0unallocated gap IPs exposed to external traffic). - BGP route aggregation uses a Single Umbrella Supernet (which spans intermediate gaps and requires a
Null0discard route to prevent routing loops).
Audit & merge CIDR lists locally in your browser:
What Else Is Inside the Workbench?
Everything on NetDevKit runs 100% locally in vanilla JS & WebCrypto—zero server uploads, dark mode, shareable links, and one-click CLI/Terraform copy:
- IP & Routing: IPv4/Cloud VPC, VLSM Splitter (CSV export), IPv6 Compression, CIDR Merger, VPC Overlap Checker
- VPN & Security: WireGuard Config Builder, Tunnel MTU/MSS Calculator, SPF & DMARC DNS Generator
- DevOps & Crypto: Offline
JWTExpiry Inspector, JSON/Base64 Formatter, WebCryptoSHA-256/SHA-512& URL Encoder, Cron Simulator, Linuxchmod/Umask, Log Regex Matcher
Next up on my list: LLM GPU VRAM & KV-Cache Sizing and AWS EKS ENI Pod IP Capacity.
What infrastructure calculator or config generator do you wish existed without popups or server uploads? Let me know in the comments!

