As 5G, network slicing and edge computing accelerate, network performance is no longer defined by bandwidth alone but by precision, predictability and intelligent control. Modern infrastructure must guarantee deterministic performance for mission critical workloads while efficiently managing best effort traffic, especially at the edge where latency and SLA enforcement are critical.
iWave’s iW-Fibre SmartNIC addresses this need by integrating Hierarchical Class of Service and Push In First Out packet scheduling directly into the FPGA data path using P4 programmability. This hardware accelerated, policy driven architecture enables line rate traffic management, protects high priority flows during congestion and delivers scalable, SLA assured performance for next generation networks.
The Challenge: Why FIFO Falls Short in Modern Networks
Legacy NICs built around First-In-First-Out (FIFO) queuing were designed for a simpler era. In today’s multi-tenant, cloud-native and latency-critical environments, FIFO is no longer sufficient.
With FIFO, traffic is treated equally, regardless of priority or service requirements. A sudden burst of low-priority flows can quickly saturate buffers, forcing mission-critical packets to wait or, worse, be dropped due to tail-drop. The outcome? SLA violations and degraded application performance.
For demanding environments such as:
- 5G User Plane Functions (UPF)
- Cell-site gateways
- Edge data centers hosting diverse, mixed workloads
best-effort fairness simply isn’t enough. Operators require deterministic bandwidth control, strict traffic isolation and guaranteed performance.
The Solution: P4-Driven HCoS with PIFO Scheduling on iW-Fibre
Moving beyond FIFO requires more than simple priority queues. It requires Hierarchical Class-of-Service enforced with deterministic precision in hardware.
The iW-Fibre SmartNIC implements a fully programmable PIFO scheduler directly in the FPGA data plane using P4. By combining structured HCoS policy control with rank-based packet insertion, iW-Fibre transforms QoS from static configuration into real-time, hardware-enforced intelligence.
Why HCoS + PIFO Matters
Hierarchical Class-of-Service defines what performance guarantees must be enforced:
- Committed Information Rate (CIR) ensures mission-critical services always receive reserved bandwidth.
- Peak Information Rate (PIR) prevents noisy neighbors from exceeding allocation.
- Multi-level hierarchy enforces isolation across tenants, slices and services.
PIFO enables how those guarantees are enforced at line rate.
Unlike traditional tail-drop queues, a PIFO scheduler inserts packets based on computed rank while always transmitting from the head, enabling precise, policy-driven ordering aligned with HCoS.
How it works
- Traffic Classification
Incoming packets are classified into forwarding classes such as VIP and Best Effort using P4 match-action pipelines. - Hierarchical Policing
Traffic is metered and marked (Green, Yellow, Red) based on configured CIR and PIR policies. For more information on QoS policing, refer to the following article: [Enabling P4 on SmartNICs: QoS Classification and Rate Limiting] - Intelligent Scheduling
Each packet is assigned a rank. High-priority packets move closer to the front of the queue, effectively skipping ahead of lower-priority packets, while lower-priority traffic waits or is dropped during congestion.
The outcome is predictable latency and guaranteed performance even under heavy load.
SLA-Aware Traffic Prioritization
To illustrate this capability, a congestion scenario was evaluated with two clients operating under different SLA policies.
In this setup, a QSFP loopback connects to the iW-Fibre SmartNIC running the P4 pipeline and mapped to the physical functions.The Server and Client VMs are connected to the physical functions via Linux virtual network interfaces.
The Server VM generates traffic for both clients, distinguished by MAC address. The iW-Fibre SmartNIC acts as an inline processor, applying Hierarchical Class of Service policies in real time and forwarding packets to the appropriate client interfaces based on their classification.
Congestion Scenario: Protecting the VIP Flow
Guaranteed Packet Delivery for High-Priority VIP Traffic
Best-Effort Traffic Behavior Under HCoS and PIFO Enforcement
As shown in Figure 1 and Figure 2, Client_1 is assigned high priority while Client_2 operates at low priority, enabling deterministic traffic differentiation during congestion. When server traffic exceeds available egress bandwidth, the hierarchical policer classifies and marks packets by rate compliance and the PIFO scheduler reorders them in real time. High priority packets are placed at the front of the queue for minimal latency, while lower priority traffic is deferred or dropped only when buffers reach thresholds.
This approach ensures mission critical traffic remains protected during link saturation, while best effort flows are efficiently paced to use remaining bandwidth, delivering predictable QoS and optimized link utilization.
The Intelligent Edge Starts Here
By integrating HCoS enforcement and PIFO scheduling directly into the iW-Fibre SmartNIC, iWave elevates the NIC into a fully intelligent traffic management platform. Offloading traffic control to the FPGA based P4 engine eliminates CPU bottlenecks, sustains line rate performance and provides flexible, programmable control over how network resources are allocated. As a result, the iW-Fibre SmartNIC delivers deterministic, SLA driven traffic control at the network edge, ensuring that critical applications consistently receive priority without compromising overall network efficiency.
For platform evaluation or additional information, contact mktg@iwave-global.com
