ONES Rule Engine details AI-Fabric queue, PFC, and link alert monitoring

The ONES Rule Engine update expands AI-Fabric monitoring by adding alertable metrics for queue behavior, PFC activity, traffic rates, and link or node failures. For data center operators and security leaders, the change affects how quickly congestion, loss risk, and link instability can be detected and escalated.

Research Overview

The blog describes the ONES Rule Engine as an integrated alert and notification system that provides monitoring at both device and interface levels. It also describes the engine as supporting rule creation tied to AI-Fabric metrics.

The update broadens the metric set to include queue counters, PFC indicators, traffic rates, and link and node failures. The intent is to improve visibility into network performance and help identify potential issues affecting RoCE-based traffic.

Key Findings

The blog states that the engine can monitor queue counter inputs such as packet transmit and receive rates, dropped packets, and ECN marked packets. It also describes PFC as an alertable metric based on priority pause frames related to congestion.

In addition to queue and PFC counters, the blog states the engine can support monitoring for link failure through link flap detection. It also describes automated alert generation when monitored conditions are met.

Technical Breakdown

For queue counters, the blog defines Packet Transmit Rate as the packet transmit rate from the RoCE queue and Packet Receive Rate as the receive rate on the RoCE queue. It describes Dropped Packets as packets dropped due to queue overflows or other reasons, and ECN Marked Packets as the count of ECN-marked packets used by RoCE to signal congestion.

For PFC, the blog describes PFC as pausing traffic by sending priority pause frames per traffic class when buffer thresholds are exceeded because of congestion. It states that counts of priority pause frames sent and received are available in the PFC counters, and that the rule engine can create alerts based on configured monitoring conditions.

Operational Impact

The blog describes how administrators can set alerting parameters for PFC counter monitoring using configurable attributes. It lists time interval options (5 min, 10 min, 15 min, 30 min, 1 hr), threshold breach operators (Greater/Lesser/Equal), and count occurrence ranges from 1 to as high as millions.

When conditions are met, the blog states the rule engine dispatches alerts through configured channels including Slack, Zendesk, and the Watcher – Alerts page. It says the generated alert includes details on the device, interface, and queue, and for link events it includes additional payload fields such as optics information, device location, and layer.

Conclusion

The ONES Rule Engine update expands alertable AI-Fabric metrics to cover queue counters, PFC activity, traffic rates, and link and node failures, with configurable threshold logic and multi-channel alert delivery. This “Blog Signals brief” is a fact-based summary of the vendor blog.