Why fog/edge before the cloud
The promise of the cloud was to simplify. Twenty years later, it has produced: vendor lock-in, opaque costs, unpredictable latencies, regulatory drift (Schrems II, NIS2, AI Act), dependence on foreign hyperscalers for sensitive Italian data.
Fog/edge is the architectural answer: the infrastructure lives close to the customer, not in someone else’s datacenter behind another nation’s firewall. Latency under 5ms. Total sovereignty. Post-incident recovery without depending on a service provider.
Anthilla infrastructure stack — real numbers
| Component | Current status | Notes |
|---|---|---|
| Uptime | 122+ days | Since January 9, 2026, active snapshots |
| Go daemons (anth-*) | 11 active | Reactor, realtime, threatd, guardian, hashdb, whitelist, diskwatch, logkeeper, logarchive, watcher, blacksync |
| systemd timers | 16 active | backup, sync, hardening, monitoring, log rotation, integrity check |
| Threat IP block | ~29,500 | permanent kernel-level DROP, multi-layer anti-attack |
| fail2ban jail | 13 active | SSH, web scanning, login hardening, etc |
| Storage pool | 2 (dual) | production pool + replica pool, 0 errors, compression active, frequent/hourly/daily/weekly snapshots |
| Whitelist | 211 entries | CIDR+prefix, auto-sync FLUX user mobile |
5-layer threat management
- Detection: real-time log monitoring, filesystem watchdog, pattern matching
- Blocking: kernel DROP, silent close or 403, global blacklist
- Synchronization: firewall rule export (1min), 3-way sync verification (1min), blacklist sync (inotify zero-polling)
- Intelligence: anth-historical-analyzer (5min log .gz retro), threat-update (hourly), threat-intelligence (daily)
- Behavioral: batch behavioral analysis, real-time scoring
Dual-pool storage — snapshot policy
Production pool and replica pool managed by an internal tool (replaces standard open source tool)::
frequent_*– every 15 minuteshourly_*– every hourdaily_*– every 02:30weekly_*– every Sunday 03:00repl_*– rsync prod → replica every 30 minutes
Automatic cleanup: an internal cleanup tool removes expired snapshots according to retention policy. Result: 148 active snapshots, 17G used out of 127G production pool.
Heritage 2001→2026
The company was not born yesterday. The first domain dates back to 2002 and Ivan Dorna’s experience starts from 2001. In over 20 years we have seen:
- The dot-com bust (2001-2003)
- The arrival of the cloud (2006-2010)
- The commodity SaaS era (2010-2018)
- Generative AI pretraining (2020-2023)
- Multi-level prompting + agentic AI (2024-2026)
Patient construction. No bandwagon. No startup pivot. No exit playbook. Just infrastructure that works, year after year.
Working with Anthilla means accessing a living stack. Not slides, not PoC: 24/7 production. For partnerships or consulting: contacts@anthilla.com