ZTE and China Telecom draft IETF requirements for congestion loss monitoring
China Telecom and ZTE Corp. authors have submitted an IETF internet-draft outlining requirements and problems for real-time monitoring of network congestion-induced packet loss. The document discusses challenges with existing measurement techniques for critical aspects like 5G, eMBB, uRLLC, and AI training, emphasizing the need for accurate and scalable solutions to maintain QOS and QOE in future networks. It highlights the difficulties in distinguishing congestion-induced loss from other loss types, capturing short-lived microbursts, and accurately identifying affected traffic flows.
Key Takeaways
- Identifies microbursts spanning 100 microseconds to 10 milliseconds as primary drivers of packet loss in latency-sensitive 5G and AI services.
- Proposes a 'protocol-independent' monitoring goal to avoid upgrading data plane forwarding chips for IPv4/6, SRv6, and VXLAN.
- Draft outlines deficiencies in existing Alternate-Marking and IOAM methods, specifically their inability to distinguish congestion from CRC errors.
- Mandates scalability for tens of thousands of concurrent user flow measurements without depleting computing or storage resources.
Why It Matters
Network operators currently rely on minute-level SNMP sampling that masks the microbursts causing buffer overflows. For industries deploying high-bandwidth AI training clusters or uRLLC 5G services, undetected congestion directly degrades training efficiency and user experience. By pushing for a standardized IETF framework, ZTE and China Telecom are positioning to solve a critical visibility gap in the 'Connectivity + Computing' architecture. Success here would move the industry away from reactive troubleshooting toward granular, automated path optimization based on real-time discarded packet analysis. Watch for the IPPM Working Group's adoption of the related 'Monitoring Arch' draft, which proposes active packet capture of discarded traffic.
Additional Context
The IETF submission follows a trend where Chinese operators are pivoting from 5G subscriber growth toward high-value monetization of enterprise services. Per RCR Wireless (April 2026), 5G penetration in China is nearing 88%, forcing a transition to 'TechCo' models where differentiated low-latency experiences drive ARPU. China Telecom has already trialed distributed Large Language Model (LLM) training with 177 billion parameters over 500km fiber links, according to Developing Telecoms (July 2025). That trial achieved 99% of single-site efficiency by using wide-area lossless technology to manage congestion during massive data transfers. Broadcom research from November 2025 further underscores the need for these monitoring standards, finding that while 99% of organizations are adopting AI, only 49% believe their networks can support the required bandwidth and low latency. The top challenge reported by NetOps teams is network congestion (46%), followed by insufficient visibility (39%). ZTE is actively addressing this gap by deploying 'AIREngine' intelligent computing boards in 5G-Advanced (5G-A) base stations to provide preventive guarantees for high-value traffic. These components use AI to estimate package admission capacity and analyze root causes for quality drops, which ZTE claims can reduce poor-quality events by up to 80% (per ZTE, January 2026). The push for IOAM and Alternate-Marking extensions is part of a broader industry effort to unify telemetry formats. In March 2026 at IETF 125, China Telecom presented enhancements to IOAM Trace Options to integrate performance measurement directly into data packets. This initiative aims to reduce chip complexity by using a single encapsulation format for both path tracing and loss measurement, a move supported by vendors including Cisco and Huawei to streamline next-generation network observability.
Read full article at datatracker.ietf.org
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source