Who is this for?
Built specifically for systems teams where generic abstraction layers cost too much CPU time and introduce unacceptable tail latency.
Infrastructure and site-reliability engineers defending Linux servers against volumetric L3/L4 DDoS attacks.
High-frequency trading and gaming infrastructure where Linux kernel firewall processing latency (iptables) is intolerable.
Teams that need deep pattern matching (Suricata/Snort/OWASP rules) at multi-gigabit line rates on commodity hardware.
What problem does it solve?
Prevents Linux kernel network stack saturation during multi-gigabit packet floods.
Standard Linux iptables or nftables process packets only after socket buffers (sk_buff) are allocated, which causes high CPU utilization and memory starvation during 10Mpps+ floods. Aethon attaches directly to the NIC driver via eXpress Data Path (XDP), dropping volumetric floods in 15 nanoseconds without memory allocation.
Why choose Aethon instead of existing alternatives?
How Aethon compares with standard Linux approaches on latency, safety, and operational overhead.
| Approach / Technology | Verdict | Trade-off & Limitation |
|---|---|---|
| Linux iptables / nftables | High Stack Overhead | Allocates full sk_buff per packet; connection tracking chokes under volumetric SYN/UDP floods. |
| DPDK Dedicated Appliances | Hardware Inflexible | Requires dedicated NICs, claims 100% of CPU cores on spin-locks, bypassing the OS completely. |
| Snort / Suricata (Standard) | Userspace Copy Penalty | Relies on libpcap/AF_PACKET packet copies, limiting deep inspection to sub-line-rate throughput. |
| Aethon (AlgoMesh) | Kernel-Bypass eBPF/XDP | Drops volumetric attacks in driver ring; zero-copy AF_XDP to multi-threaded Hyperscan DPI engine. |
Key Capabilities
Get started
Try it in a network namespace
sudo apt-get install libbpf-dev libxdp-dev libhyperscan-dev libssl-dev clang llvm
./veth0.bash # creates veth0 <-> client_ns test network
./build.sh --release
sudo ./build/src/afxdp_control veth0On the roadmap
What is planned next. Follow progress on GitHub.
- Flow bypass: inspect once, then pass clean flows at line rate
- Hyperscan streaming mode across TCP segments
- Native-mode zero-copy AF_XDP and a hard memory budget