Dear Prasad, Thanks for the feedback and confirming that that path delay measurement IPFIX entities are useful.
Indeed visibility into queuing, https://www.ietf.org/blog/banishing-bufferbloat/, in particular FQ-CoDel https://datatracker.ietf.org/doc/html/rfc8290#section-3 dimensions would be very useful. Allow a network operator to understand which flows are ending up into which queue and how the CoDel algorithm is invoked on that queue. With traditional per packet queuing https://datatracker.ietf.org/doc/html/rfc7806 techniques, a mapping between queue to flow is not achievable with RFC 7015 Flow Aggregation. Therefore the scalability and applicability would be limited. You are proposing additional congestion related IPFIX entities. From an implementation perspective, could you please describe the metrics from a network plane perspective more detailed, to which IPFIX dimensions and measurements they are applicable to and which use case your are trying to cover. I am willing to help how this could be at IETF OPSAWG in a dedicated document best addressed to. Best wishes Thomas From: Prasad Yadati <ypra...@juniper.net> Sent: Wednesday, January 22, 2025 7:12 PM To: opsawg@ietf.org; Graf Thomas, INI-NET-VNC-E2E <thomas.g...@swisscom.com>; alex.huang-f...@insa-lyon.fr Subject: Subject: Proposal to Augment draft-ietf-opsawg-ipfix-on-path-telemetry-14 with Throughput and Congestion Metrics Be aware: This is an external email. Subject: Proposal to Augment draft-ietf-opsawg-ipfix-on-path-telemetry-14 with Throughput and Congestion Metrics Dear OPSAWG Members, I want to express my appreciation for the excellent work on draft-ietf-opsawg-ipfix-on-path-telemetry-14. The detailed IPFIX Information Elements (IEs) for delay performance metrics provide significant value to network operators and represent a strong foundation for real-time performance monitoring. I am writing to propose an augmentation to the draft by incorporating throughput and congestion metrics. These additional metrics would complement the existing delay-focused IEs and enable network operators to correlate delay performance with throughput bottlenecks and congestion points, providing a more holistic view of network behavior. Proposal Overview The draft currently focuses on delay performance metrics such as pathDelayMeanDeltaMicroseconds and pathDelayMaxDeltaMicroseconds. While these are critical for delay monitoring, they do not address throughput or congestion, which are often the root causes of delays. I propose defining unique Information Elements (IEs) for throughput and congestion metrics to complement the existing delay metrics and provide operators with actionable, real-time insights into network performance. Proposed Unique IEs To extend the draft's IPFIX template, I propose adding the following unique IEs: | IE Name | Description | Type | Units | |-----------------------|--------------------------------------------------|-----------------|--------------------| | PacketsPerSecond | Number of packets transmitted per second | Float/Integer | Packets/second | | BitsPerSecond | Number of bits transmitted per second | Float/Integer | Bits/second | | CongestionDetectedFlag| Boolean flag indicating if congestion detected | Boolean | N/A | | CongestionStartTime | Timestamp when congestion began | DateTime | Seconds since epoch| | CongestionEndTime | Timestamp when congestion ended | DateTime | Seconds since epoch| | CongestionDirection | Indicates congestion direction: ingress/egress | Integer/Enum | 1 = ingress, 2 = egress | Advantages of the Proposed Enhancements By incorporating throughput and congestion metrics into the draft, operators will benefit from: * Enhanced Correlation: Correlate delay performance with throughput bottlenecks and congestion issues, offering deeper insights into the root causes of performance degradation. * Holistic Network View: Gain a comprehensive perspective on network performance, enabling operators to monitor and troubleshoot delay issues alongside throughput and congestion in real time. * Reduced Collector Overhead: Pre-calculated pps and bps metrics exported directly from devices reduce the processing burden on collectors, improving scalability in high-traffic environments. * Improved Troubleshooting: Real-time access to these metrics enables faster diagnosis and resolution of performance issues. * Simplified Export: Consolidating delay, throughput, and congestion metrics into a unified IPFIX template reduces operational complexity and export overhead. I believe that incorporating these additional metrics into the draft aligns with its objectives and significantly enhances its utility for network operators. I would be honored to collaborate further on refining these enhancements and contributing to the draft as a co-author. Please let me know your thoughts on this proposal. Regards yp Juniper Business Use Only
smime.p7s
Description: S/MIME Cryptographic Signature
_______________________________________________ OPSAWG mailing list -- opsawg@ietf.org To unsubscribe send an email to opsawg-le...@ietf.org
