Skip to main content

Blog post · November 19, 2026 · aviznetworks.com

Aviz Networks details Service Node R2.6 release features

Summary

Aviz Networks has released Service Node R2.6, adding operator-defined application identification, runtime Kafka topic configuration, and an AI inference parser. The changes matter to IT and security leaders managing data center and telco traffic analysis.

Traffic Classification and Export Controls

Custom App-ID lets operators define application identification rules based on inner-flow fields and payload signatures, exported as custom app_name and app_group attributes. Rules are stored in a YAML file, match on IP, port, transport protocol, and payload byte signatures, and hot-reload without an ASN restart. Flows that match no custom rule continue to be classified by PACE2, and invalid rules are rejected while the previous rule set stays active.

Kafka export topic names, previously fixed in ASN, can now be set per export stream through the ONES UI. The streams include U-Plane user sessions, EPC and 5GC control plane, RADIUS, Diameter, SIP, and PFCP. Changes take effect without a restart, and deployments without custom settings keep existing defaults.

DC Session Records and HTTP/2 Processing

The DC export model now produces one combined record per TCP session instead of separate uplink and downlink records. The record carries uplink and downlink bytes, packets, and latency side by side, along with per-interval packet loss and retransmission counts for each direction. It also adds TCP status indicators for SYN, SYN-ACK, FIN, RST, and ESTABLISHED states, and all fields are available over Kafka and IPFIX without changes to existing consumers.

The N11 HTTP/2 partial-session table holds up to one million entries and fills in about one hour under production load. Previously, a full table was flushed inline on the control core, stalling packet processing for 50 to 60 seconds. Each table slot now has a permanent 32 KB memory arena with a dedicated nghttp2 allocator, so flushing a slot is a pointer drop and HPACK state persists for live connections.

AI Inference Traffic Visibility

The ASN Core AI Parser passively dissects OpenAI-compatible chat-completion traffic in transit. It reconstructs requester identity, model name, prompt and completion token counts, and end-to-end response timing for each call. The parser runs on an ASN appliance in DC mode using tap, mirror, or inline capture, and no instrumentation of AI clients or inference servers is required.

According to the blog, engine-side metrics from LLM-serving platforms such as vLLM do not identify requesters, models, or on-wire response performance. Parsed records are exported through the standard ASN Kafka path alongside existing session data.

Storage Management

Custom Disk Path redirects all ASN-managed files, including session state, export buffers, diagnostics, and configuration, to a single operator-specified directory. The path can be set at install time or at runtime without reinstallation. An invalid or inaccessible path falls back to the previous setting with a warning, and the configured path persists across restarts.

This Blog Summary is a fact-based summary of the vendor blog. For enterprise decision-makers, R2.6 changes operator control over traffic classification, export behavior, session data, AI inference visibility, and storage layout in ASN deployments.

Our summary. The original keeps its own byline.

Read the original on aviznetworks.com

Companies in this article

Topics