MENU ↓
       Datasheets Download
         Catalogs Download
           Product & Service FAQs
             Company Policy
               Privacy Policy
                 Distribution Products


                keyword: 12V 1A EU adapter



                  Main Content
                Starlink Enterprise Network Integration: Firewall, VLAN, SD-WAN & CGNAT Guide | SatNet Power Supply | KRECO
                Publish Date:2026-09-30    Click:920

                KRECO | Starlink & SatNet Power Supply FAQ Series — Article 26

                Starlink Enterprise Network Integration: Firewall, VLAN, SD-WAN & CGNAT Guide

                Short answer: Starlink Enterprise ships with NO router — by design. The PSU exposes a standard RJ45, so you connect your own firewall / router / SD-WAN appliance directly: dish → Enterprise cable (50 m) → IP66 PSU (100–240V ~6.3A, dedicated earth terminal) → RJ45 → firewall WAN → managed PoE switch → VLANs + APs. Set WAN to DHCP (no PPPoE), expect a lease in the CGNAT 100.64.0.0/10 range, and clamp MSS ≈1360–1400 on tunnel interfaces (WireGuard ≈1420 MTU) to kill fragmentation. CGNAT blocks all unsolicited inbound: no port forwarding, no inbound IPsec, no hosted servers — fix with hub-and-spoke VPN (every spoke dials out), reverse tunnel, cloud relay, a public/routable IP option (confirm with your plan), IPv6, or an SD-WAN overlay, plus ≈25 s NAT keepalives. Segment with VLANs (management / staff / guest / IoT / voice / CCTV) + inter-VLAN ACLs + guest isolation. Tune SD-WAN for LEO (RTT ≈25–60 ms, higher jitter, brief handovers): probe 1–3 s, require 5–10 consecutive failures, add hysteresis and 5–15 min failback dampening; never per-packet bond Starlink with a terrestrial link. Shape upstream ≈90–95% with fq_codel/CAKE. Finally, power the whole connectivity layer (≈215W → ≈2.0 kWh for 8 h, ≈6.1 kWh for 24 h) on battery-backed UPS/DC-UPS — Starlink's 15–20 min reboot gap makes this the highest-value item on site.

                1. Reference Architecture & Why Enterprise Has No Router

                Layer Device / Interface Enterprise-Specific Note
                Antenna Standard V4-class dish, IP67, 110° FOV, snow melt ≤40 mm/h Fed by proprietary passive PoE 56–57V on pins 1236+ / 4578- (not 802.3af/at)
                PSU 100–240V AC ~6.3A, IP66, −30 °C to 50 °C, wall-mount, dedicated earth terminal Larger PSU than Standard; bond the earth terminal per local code
                Cable Enterprise Cable 50 m (164 ft); PSU RJ45 → your network Longest official dish cable in the range; do not extend with generic Cat5e/Cat6 or couplers
                Gateway YOUR firewall / router / SD-WAN appliance No Starlink router in the box — you supply NAT, DHCP, DNS, firewall, IDS/IPS, Wi-Fi
                Distribution Managed PoE switch → ceiling APs, phones, cameras, NVR Cat6 ≤100 m per run; size the PoE budget for all APs
                Power protection UPS / DC-UPS on battery-backed outlets for PSU + firewall + switch + APs Bridges the 15–20 min Starlink reboot/re-acquire gap

                Why "no router" is a feature: IT teams get full control — your own NAT, DHCP scopes, DNS, firewall policy, IDS/IPS, SD-WAN, VLANs, guest isolation, logging and compliance posture. Standard 4 X can also do this, but only in Bypass Mode, and Router 3 must remain powered because it is still the dish's PoE source. Enterprise removes that dependency entirely.

                2. WAN Interface Setup (DHCP, MTU/MSS, Diagnostics)

                Setting Value / Action Why
                WAN addressing DHCP client (no PPPoE, no username/password) Starlink hands the WAN address by DHCP
                Expected WAN IP Typically 100.64.0.0/10 (CGNAT shared space) unless you have a public/routable option Explains why inbound access fails (see §3)
                LAN MTU Leave at 1500 on LAN; do not "jumbo" anything Standard Ethernet; mismatches cause silent breakage
                Tunnel MTU / MSS WireGuard ≈1420 MTU; clamp TCP MSS ≈1360–1400 on IPsec/overlay interfaces Overlay headers plus a 1500-byte underlay cause fragmentation → black-holed HTTPS, stuck VPNs, broken apps
                Dish diagnostics Add a route on the WAN interface toward the Starlink diagnostic address, commonly 192.168.100.1 (confirm for your kit/firmware) Lets NMS/monitoring scrape dish telemetry; the Starlink app also provides obstruction view and stats
                NAT / firewall Enable outbound NAT (masquerade) on WAN; default-deny inbound Your firewall is now the only NAT layer you control
                NTP & DNS Configure NTP and DNS explicitly; consider filtered/DoH DNS Certificate validation, VPN establishment and logging all depend on correct time and DNS

                3. CGNAT: What It Breaks and the Six Ways Around It

                Broken by CGNAT Symptom Fix
                Inbound IPsec / site-to-site HQ cannot initiate the tunnel to the site; only outbound works Make the site the initiator dialling a hub with a public IP; add NAT-T and short keepalives
                Port forwarding / hosted services Web server, NVR, camera, telemedicine endpoint unreachable from outside Reverse tunnel or cloud relay; or a routable/public IP option if your plan provides one
                Remote admin / SSH / RDP Cannot reach site devices directly SD-WAN overlay, mesh VPN (WireGuard/Tailscale/ZeroTier-class), or cloud relay
                SIP / VoIP One-way audio, dropped registrations (double NAT + ALG) Use a SIP-aware SBC or cloud PBX; disable/avoid ALG where it misbehaves; short keepalives
                NAT mapping expiry Tunnels "up" but idle, then silently dead Set VPN keepalive ≈25 s; enable DPD/dead-peer detection
                Geo/IP reputation Some services mistrust shared CGNAT space; captchas, blocks Route that traffic via a cloud egress/relay with a clean IP

                Six accepted solutions (pick per use case):

                1. Hub-and-spoke VPN — every remote site initiates outbound to a hub/data-centre with a public IP. The most robust general fix.
                2. Reverse tunnel / reverse proxy — the site keeps an outbound connection to a cloud relay that brokers inbound sessions.
                3. Cloud relay / SD-WAN fabric — vendor overlay handles NAT traversal and remote access (best for multi-site fleets).
                4. Public / routable IPv4 option — ask Starlink whether your plan offers a public or static IP; where offered this restores true inbound capability. Confirm for your plan and region.
                5. IPv6 — where Starlink offers a delegated prefix, IPv6 restores end-to-end reachability for dual-stack endpoints. Confirm availability.
                6. Mesh / overlay VPN (WireGuard-class) for small teams and rapid deployment — simple, NAT-traversing, short keepalives.

                4. VLAN Segmentation Design

                VLAN Purpose Policy
                Management Switch, AP, firewall, UPS/DC-UPS, PDU, dish telemetry Admin access only from here or over VPN; never from guest or WAN
                Staff / corporate Employee devices, domain resources Full internal access per policy; consider 802.1X
                Guest Visitors, shelter residents, café users Internet-only; client isolation; captive portal; content filtering; bandwidth caps
                IoT / OT / medical Building systems, clinical IoT, cameras-adjacent devices Isolated; explicit ACLs; no lateral access to staff VLAN (HIPAA/PCI scope reduction)
                Voice IP phones, intercom, incident comms Priority DSCP EF; dedicated QoS queue
                CCTV Cameras and NVR Block outbound internet for cameras; allow NVR + management only
                • 802.1Q trunking from switch to APs (SSID-to-VLAN mapping) and to any virtualised firewall.
                • Per-VLAN DHCP + DNS, DHCP snooping, and (where supported) ARP/DHCP guard to stop rogue servers.
                • Inter-VLAN ACLs default-deny, with explicit allows — the ACL list is the security design, not an afterthought.
                • Diagnostics: keep dish telemetry reachable from the management VLAN via the WAN-side route (§2).

                5. SD-WAN & Failover Tuned for LEO (Not Fibre Timers)

                Parameter Fibre-style (wrong for LEO) LEO-tuned (recommended)
                Probe interval 5–10 s 1–3 s to several diverse targets
                Failure threshold 2–3 misses (flaps constantly on LEO) 5–10 consecutive failures before path-down
                Latency threshold 30–50 ms Allow 60–120 ms+; LEO RTT is typically ≈25–60 ms with occasional spikes
                Jitter / loss Often ignored Add explicit jitter and packet-loss thresholds, not just reachability
                Hysteresis / dampening Immediate failback Require 5–15 min of stable health before failing back
                Load sharing Per-packet bonding Per-flow / policy-based steering only — per-packet bonding reorders packets and destroys voice/video
                Policy Route by destination Application-aware: voice/telemedicine/incident comms to the lowest-latency path; bulk backup to the cheap path

                Why the numbers differ: LEO constellations hand traffic between satellites every few minutes and the link is subject to weather, obstructions and terminal attitude. Short, occasional spikes are normal, not an outage — fibre-grade timers will flap the link all day and make the site less reliable than a single well-tuned WAN.

                6. QoS & Bufferbloat Control (Upstream Is the Bottleneck)

                • Measure first: run a baseline speed and bufferbloat test; record down/up and loaded latency.
                • Shape upstream to ≈90–95% of measured upload. Saturating the uplink is what makes calls stutter and VPNs crawl.
                • Use modern AQM: fq_codel or CAKE on the WAN interface — far more effective than classic priority queues alone.
                • Prioritise by DSCP: voice and telemedicine EF; interactive/incident comms AF31-ish; bulk backup and updates to a scavenger class.
                • Per-VLAN guarantees: reserve a floor for voice and critical VLANs so guest traffic cannot starve them.
                • Schedule heavy jobs (backups, image sync, cloud upload) outside business hours — LEO capacity is shared and deprioritisation may apply after plan thresholds.
                • Re-test after changes: loaded latency is the metric that matters, not raw throughput.

                7. Security Hardening Checklist

                • Default-deny inbound on the WAN; no management interface exposed to WAN.
                • Admin only from the management VLAN or over VPN; MFA on firewall and controller accounts.
                • IDS/IPS + DNS filtering + content filtering (mandatory for schools and shelters).
                • Segmentation to keep PCI / HIPAA / clinical scope small and auditable.
                • Firmware update policy for firewall, switch, APs and the terminal; subscribe to advisories.
                • NTP + centralised logging + alerting (link down, WAN IP change, UPS on battery, disk/CPU health).
                • Physical security: lockable rack, cable management, no exposed PSU or dish cabling, secured outdoor enclosure.
                • Guest isolation and wireless client isolation; rotate PSKs; prefer 802.1X for staff where feasible.
                • Backups of firewall/switch config, stored off-site, with a documented restore runbook.

                8. Network Power Budget & Backup (The SatNet Power Supply Layer)

                Device Typical Draw On Backup?
                Enterprise dish + PSU 75–100W avg (plan 100–110W; peaks ≈195W-class) Yes — battery-backed outlet
                Firewall / SD-WAN appliance 30–40W Yes — without it there is no gateway
                Managed PoE switch 25–60W (plus PoE load) Yes
                Wi-Fi APs (PoE) 10–18W each Yes (critical APs at minimum)
                Small server / NVR / telemedicine endpoint 30–80W Yes, if business-critical
                Typical connectivity layer ≈215W —

                Battery (Wh) = Load (W) × hours ÷ 0.85, then add 15–20% cold margin. For a ≈215W connectivity layer:

                Autonomy Battery (before cold margin) With 15–20% Cold Margin
                30 min (Tier 0 bridge) ≈127 Wh — any small UPS Covers the 15–20 min reboot gap — highest-value item
                8 h (school day) ≈2.0 kWh ≈2.3–2.4 kWh
                24 h ≈6.1 kWh ≈7.0–7.3 kWh
                72 h (shelter standard) ≈18.2 kWh ≈20.9–21.8 kWh
                • Battery-backed outlets only — never surge-only (surge-only gives zero runtime). This is the #1 UPS mistake.
                • W = VA × PF (≈0.6–0.7) — a 1500VA UPS may deliver only ≈900W; VA-rated units give minutes, Wh-rated LiFePO4/DC-UPS give hours.
                • Standard 4 X instead of Enterprise? Router 3 must stay powered even in Bypass Mode — put it on backup power too.
                • DC-UPS preferred off-grid: one protected DC bus feeding dish (via 56–57V DC-DC/APSU) + firewall + switch + APs, avoiding inverter loss; require auto-restart after depletion.
                • Inverter route: pure sine wave ≥500W (800–1000W recommended). Modified/square wave will not run Starlink PSUs properly.
                • POE_INSUFFICIENT is a power fault, not a network fault — check supply wattage, cable length (≤50 m Enterprise) and connectors before touching VLAN/QoS config.

                9. Redundancy & Uptime Design

                Layer Redundancy
                Power UPS/DC-UPS bridge + LiFePO4 bank + solar/MPPT + generator with ATS (public-safety sites); dish and network on separate protected paths so one failure cannot kill both
                WAN Secondary WAN (fibre / LTE / 5G) with SD-WAN failover; or a second Starlink terminal for critical sites
                Gateway HA firewall pair where uptime justifies it; otherwise a cold spare with a restored config backup ready
                Power-path monitoring SNMP/telemetry on UPS/DC-UPS (on battery, SoC, runtime) and dish telemetry; alert before the bank is empty
                Process Monthly 10-minute failover test, quarterly full-duration test, documented runbook, spare cables/fuses/PSU on the shelf

                10. Common Mistakes (Corrected)

                • Expecting a router in the Enterprise box. Wrong — no router included; you supply gateway, DHCP, DNS, firewall and Wi-Fi.
                • Assuming the WAN IP is public. Wrong — default is CGNAT 100.64.0.0/10; inbound is blocked. Confirm whether your plan offers a routable/public IP.
                • Planning inbound site-to-site IPsec from HQ. Wrong under CGNAT — make the site the initiator into a public hub.
                • No MSS clamping on tunnels. Wrong — clamp ≈1360–1400 (WireGuard MTU ≈1420) or suffer black-holed sessions.
                • Using fibre-grade SD-WAN timers on LEO. Wrong — probe 1–3 s, require 5–10 consecutive failures, add hysteresis, dampen failback 5–15 min.
                • Per-packet bonding across Starlink + terrestrial. Wrong — reorders packets and ruins voice/video; use per-flow/policy steering.
                • No upstream shaping. Wrong — shape to ≈90–95% with fq_codel/CAKE; upload saturation is the real bottleneck.
                • Plugging the PSU into a surge-only UPS outlet. Wrong — battery-backed outlet only.
                • Protecting the dish but not the network. Wrong — a powered dish with an unpowered firewall/switch/AP is still offline; budget the whole ≈215W layer.
                • No VLANs / flat network. Wrong — guest devices on the same broadcast domain as staff and clinical IoT is a security and compliance failure.
                • Unplugging Router 3 (Standard 4 X) in Bypass Mode. Wrong — it remains the dish's PoE source and gateway; keep it powered.
                • Extending the 50 m Enterprise cable with generic Ethernet. Wrong — use remote power (IP66 enclosure near the dish) and Cat6 ≤100 m for data.
                • Treating POE_INSUFFICIENT as a network problem. Wrong — it is a power fault; check wattage, cable length and connectors.
                • No config backup or restore test. Wrong — a firewall failure with no backup turns a 10-minute fix into a site-wide outage.

                KRECO / iPSKRE SatNet Power Supply Solutions

                KRECO Group — an IEC 60601 medical & industrial power manufacturer since 2008 (Equity Code 891359, Medical Reg. 20200118#, FIME Silver Sponsor) — is developing Starlink-compatible SatNet Power Supply OEM/ODM solutions under the iPSKRE Power Ecosystem, designed so the dish and the whole network stack share one protected DC bus:

                • Wide-input DC-DC boost modules — 10–60V DC input (12V / 24V / 48V and 51.2V LiFePO4) → regulated 56–57V DC output at ≥200W continuous for Enterprise / Standard 4 / 4 X, and 340W-class modules (≈54V / 6.3A) for Flat High Performance / Performance Gen 2 — the clean alternative to APSU + inverter at an IT-managed site.
                • DC-UPS for the whole connectivity layer — LiFePO4-compatible DC-UPS with auto-restart after depletion, MPPT solar input, monitoring/telemetry for NMS (on-battery, SoC, runtime), and 15–20% cold-margin guidance; keeps dish + firewall + switch + APs alive as a single protected subsystem sized by Wh = W × h ÷ 0.85 (≈2.0 kWh for 8 h, ≈6.1 kWh for 24 h at ≈215W).
                • Proprietary-pinout passive PoE injectors — matching 1236+ / 4578-, regulated 56–57V, 195W / 150W+ / 200W / 340W classes, with surge/ESD protection and gigabit pass-through; correct spares for enterprise closets.
                • IP66 weatherproof remote-power & grounding enclosures — IP68 cable glands, drip loops, venting/shading and an earth-bonding point for the PSU's dedicated earth terminal; for remote-power designs that respect the 50 m Enterprise cable and the 100 m Cat6 data limit.
                • Medical power (clinical layer, separate by design): KRECO medical series KRE006MPS–KRE072MPS (wall-mount 6W–72W) and KRE048MPS–KRE480MPS (desktop 48W–480W, e.g. 48V 10A), built to IEC 60601-1 4th Edition, 2×MOPP, ≤100 µA leakage, Medical Reg. No. 20200118#; plus medical backup power, UPS and energy storage for healthcare facilities.
                • Compliance & build: SatNet range designed to IEC/EN 62368-1, IP66/IP67 options, full OVP / OCP / OTP / SCP + reverse-polarity protection; medical range to IEC 60601-1. B2B OEM/ODM customization (input/output voltage, connectors XT60 / barrel / RJ45 PoE / Anderson, DIN-rail, wall or rack housings, cable length, labelling, packaging, enclosure branding) for enterprise IT distributors, MSSPs, SD-WAN vendors, system integrators and SatNet brands.

                Inquire Now: sales@kreco.com.cn | iPSKRE Power Ecosystem | Starlink-compatible SatNet Power Supply OEM/ODM — wide-input DC-DC, DC-UPS for the full connectivity layer, passive PoE injectors & IP66 enclosures under development, plus IEC 60601 medical power in production.

                Based on Starlink official Specifications & Help Center (Enterprise: Standard V4-class antenna, 75–100W average, 110° field of view, IP67 Type 4, snow melt up to 40 mm/h, wind operational 96 kph+, Enterprise Cable 50 m / 164 ft, larger power supply 100–240V ~6.3A 50–60 Hz, IP66, −30 °C to 50 °C, ≈1.5 kg, wall mount, dedicated earth terminal, NO router included, RJ45 Ethernet for direct network integration; proprietary passive PoE 56–57V with +V on pins 1,2,3,6 and −V on pins 4,5,7,8; Flat High Performance / Performance Gen 2: 110–150W average, 200W max, in-motion with Mobile Priority / Maritime plans; Standard 4 X: Router 3 is the dish PoE source and gateway and must stay powered, including in Bypass Mode; DC-DC 12–48V in → 56V/3.57A ≈200W; APSU UTP-252C AC 100–240V + DC 12–57V → 57V IP68; pure sine wave inverter ≥500W guidance; Starlink reboot/re-acquire ≈15–20 min) plus general enterprise networking practice as of 2026 (IEEE 802.1Q VLANs, DSCP QoS, fq_codel/CAKE AQM, IPsec/WireGuard NAT traversal, RFC 6598 shared address space 100.64.0.0/10). CGNAT behaviour, public/routable IP options, IPv6 availability and the dish diagnostic address vary by plan, region and firmware — confirm with Starlink and your own kit before finalising a design. Sizing uses battery Wh = Load (W) × hours ÷ 0.85 with a 15–20% cold margin; the ≈215W connectivity layer and device wattages are illustrative templates — measure your own site. KRECO is not affiliated with or endorsed by Starlink/SpaceX; Starlink-compatible SatNet Power Supply modules are under development. SatNet modules are IEC/EN 62368-1 industrial products and are NOT medical devices — clinical equipment must use IEC 60601-1 medical supplies. Always use licensed installers for mains, grounding and lightning-protection work, bond the dedicated earth terminal, and keep original Starlink parts for warranty.