4 x 10G Enterprise Router Powered by SONiC and VPP
VAT excl.
- Pre-installed SONiC + VPP router OS — deploys with zero configuration.
- 8-core ARM64 Neoverse N2 @ 2.5 GHz with 16 GB DDR5 (up to 48 GB).
- 4×10GE SFP+, 4×2.5GE, 8×1GE RJ45, with PoE+/PoE++ (150 W budget).
- 50 Gbps for routing, firewall, IPSec, and SSL/TLS on a single 1U box.
- BGP, OSPF v2/v3, RIP, MP-BGP, MPLS, VRF, VXLAN/EVPN, and HQoS.
- Stateful firewalling, IPSec/WireGuard VPN, ACLs, GeoIP/GeoSite filtering.
- Optional 160 TOPS INT8 AI inference engine, M.2 SSD up to 4 TB, 5G/LTE modules.
The RT2508 is an enterprise-grade router that pairs Asteraix's compute-and-networking hardware RT2508 with AsterNOS-VPP
a production router operating system built on the proven SONiC framework and the high-performance VPP (Vector Packet Processing) data plane. Shipped pre-integrated, the two work as one product: an enterprise router that delivers carrier-grade routing, security and VPN performance out of the box, while remaining fully open and free of vendor lock-in.
At its core is the Marvell OCTEON 10 CN102XX,
a fusion chip that integrates an 8-core ARM 64-bit Neoverse N2 processor, programmable Ethernet switch ports with a total throughput of 58 Gbps, and an embedded encryption/decryption engine with the same 58 Gbps processing capacity. AsterNOS-VPP binds directly to these hardware engines, so routing, stateful firewalling and IPSec/SSL all run at line rate rather than in software alone. The result is a 1U compact box, that consolidates the router, firewall, VPN gateway and traffic-control functions of a traditional rack of dedicated appliances.
| Computing | |||
|---|---|---|---|
| CPU | 8 Core ARM64 N2 @ 2.7 GHz | Cache | L2 8MB, L3 16MB |
| RAM | 16GB DDR5 SODIMM, up to 48GB | Flash | 64GB eMMC 5.1 |
| NVME SSD (Option) | up to 4TB, M.2 M key, share slot with AI accelerator | ||
| AI accelerator (Option) | |||
| Compute performance | up to 160 TOPS up to 160 TOPS | Memory | 24GB LPDDR5/LPDDR5X Memory bandwidth: 153.6GB/b |
| Network interface | |||
| 10GE (SFP+) | 4 | PoE | 2.5GE ports POE++, 1GE ports POE+ |
| 2.5GE (RJ45) | 4 | WiFi (Option) | WiFi6E/7, M.2 E key |
| 1GE (RJ45) | 8 | 5G/LTE (Option) | 2 SIM cards, M.2 B key |
| Antenna | 6 | ||
| Network performance | |||
| L3 Switching capacity | 58Gbps | Firewall capacity | 58Gbps |
| Routing capacity | 58Gbps | Encryption and Decryption capacity | 50Gbps |
| PTP/SyncE accuracy | 20ns | PTP/SyncE holdover time | > 8hours |
| Misc interface | |||
| USB | 1 x USB 3.0 | Console | 1 x Console RJ45 |
| Electrical characteristics | |||
| Fan | 2 | Power Module | 1 x 150W (w/o PoE); 1 x 270W (PoE) |
| Maximum power consumption | 60W (FULL configuration and workload) | Input voltage | 100~240VAC |
| PoE budget | 150W | ||
| Mechanical | |||
| Operating temperature | 0 – 45℃ | Dimensions (H x W x D mm) | 44 x220 x 310 |
| Relative humidity | 5% - 95% (non-condensing) | Height | 1U |
Loading…
Loading…
| Memory Size | 16GB, 32GB, 64GB |
|---|---|
| PoE Capability | None, 150Watt |
| PTP Module | None, Class C |
| NVME Storgae | None, 1T, 2T |
| Role | Router |
| Speed | 50G |
| Throughput | 50Gbps |
58Gbps
58Gbps
50Gbps
SONiC Router OS
SONiC-VPP Routing OS
AsterNOS-VPP, A next-gen SONIC-based routing OS combining robust management with a high-performance VPP software data plane. Running on ARM/x86 architectures, it replaces standard libraries with libsaivpp.so to deliver line-rate L3 routing, CGNAT, PPPoE, and secure IPsec/WireGuard VPN at up to 100Gbps per license.
Layer 2 Features
Layer 3 Features
Security
MPLS
IP Service
Tunnel
High Availability
Quality of Service (QoS)
Monitoring & Management
High-Performance Routing for SMB & Enterprise Networks
In modern Spine-Leaf topologies, RT2500 acts as a powerful SMB Router/Border Leaf. Powered by the Marvell OCTEON 10 CN102 chip and Enterprise SONiC Router OS with VPP/DPDK acceleration, it delivers 58Gbps of intelligent data processing across flexible interfaces 4 x 10GE, 4 x 2.5GE, 8 x 1GE. It supports IGMP, MLD, PIM, IPsec VPN, Wireguard VPN, PPPoE, GeoIP, HQoS and so on.
Option Modules
Edge AI inference
Optional 160-TOPS module runs 7B–30B LLMs locally. No cloud round-trip. Slot the optional M2-AI-INFERENCE-24GB module into the same chassis, and the gateway becomes an inference node — 160 TOPS of INT8 compute and 24 GB of LPDDR5 at 153.6 GB/s, enough to host quantized models up to the 30B class. Run intelligent operations, automated troubleshooting, anomaly detection, and private assistants on-site, with prompts and data that never leave the box.