Broadcom: Predicting Network Congestion in the Era of 5G, Cloud, and SDx

Executive Summary

Broadcom is a recognized industry leader in semiconductor connectivity solutions. Combining this leadership with high-scale operations monitoring software, Broadcom offers a comprehensive solution for rapidly expanding and complex modern network architectures. The end-to-end software to silicon portfolio leverages unique hardware and software synergies to drive next-generation software-defined networks and cloud networks, fulfilling the promises of modern 5G infrastructure.

The Challenges Created By the New Hyper-Connected Business

Businesses today are hyper-connected, relying on complex infrastructures, multi-cloud environments, and billions of devices across a mesh of networks. These networks face immense pressure and are more susceptible to failures and outages. The exponential growth of SD-WAN, IoT, and faster networking technologies like 5G, which support real-time applications, generates significant traffic at the network edge. Traditionally, data computation occurred at the core, but to avoid latency issues, IoT and 5G require processing and storage closer to the data source. Projections indicate that by 2025, 5G networks will carry nearly half of the world's mobile data traffic.

This traffic, coupled with constant consumer access to digital applications, necessitates resilient yet massive infrastructure. This presents an expensive and complex challenge, as current networks are not optimized for this high-speed demand. Consequently, network operations teams face increased congestion, noise, and reduced visibility for triaging digital experience delivery. Gartner predicts that by 2024, 50% of network operations teams will need to rearchitect their network monitoring stacks due to the impact of hybrid networking, a significant increase from 20% in 2019.

The current data deluge demands innovative networking and monitoring approaches to address congestion. Organizations require visibility across current, modern, and edge communication paths to ensure reliable customer experiences. Figure 1 illustrates how AIOps network monitoring captures real-time data at the chip level through telemetry, providing edge-to-edge, AI-powered triage and visibility.

Introducing a Software to Silicon Solution Powered by Automation.ai

Broadcom pioneers next-gen AIOps analytics for hybrid cloud, SD-WAN, IoT, and 5G, powered by Broadcom merchant switch silicon, Open-Source SONiC, and Broadview Telemetry. This initiative offers a differentiated monitoring solution that captures real-time packet and flow data directly from Broadcom chips and other devices.

The solution provides unparalleled fine-grain, per-packet, and flow-level network monitoring by ingesting rich network data from Broadcom chipsets. It enables advanced predictions, correlations, and automated network remediation. The system analyzes network activity to automate, repair, and tune the network, reducing operational costs and accelerating triage. This allows enterprises and service providers to deliver consistent and exceptional digital experiences. AIOps from Broadcom is a core component of their software to silicon capabilities, delivering AI and machine learning, automation, and comprehensive ecosystem observability. It correlates data across users, applications, infrastructure, network, and security services.

AIOps is powered by Automation.ai, the industry's first AI-driven software intelligence platform. Automation.ai leverages advanced AI, machine learning, high-scale operations monitoring, and open-source frameworks to transform enterprise data into actionable insights. The AIOps solution delivers cross-domain contextual intelligence by analyzing diverse data sets including metrics, topology, text, and log data, offering features such as:

• Service analytics: Expedite root-cause analysis with end-to-end visibility across key business or IT services.
• Performance and alarm analytics: Reduce false and redundant alerts, decreasing Mean Time To Repair (MTTR).
• Contextual log analytics: Automatically correlate security and application events with performance data.
• Predictive analytics: Identify network bottlenecks before disruptions and optimize resources by identifying waste.
• Contextual automation: Deliver closed-loop, service-driven autonomous remediation.

Software to Silicon Capabilities from Broadcom

AIOps from Broadcom, powered by merchant switch silicon, drives innovation across the networking stack. It enables new use cases like using telemetry for granular, real-time insights into hop-by-hop latency, packet drops, and network congestion. Key silicon technologies include:

• Operational visibility into switch health and fabric topology.
• Real-time Broadview congestion alerting and queue/congestion analytics for each port.
• In-band Flow Analytics (IFA) to view per-hop latency impact from micro-bursts.
• Automation.ai for deep insights and automated remediation of SONiC and Broadview control plane APIs.

These capabilities support high-scale operations monitoring use cases such as:

• Monitoring critical operational KPIs in context with alarms and syslog events.
• A switch topology view with alarms and health indicators.
• Real-time Broadview congestion alerting and historical buffer analytics.
• Silicon-driven In-band Flow Analytics (IFA) providing real-time hop-by-hop latency information within switch fabrics.

Silicon-level telemetry enables use cases like packet path tracing with route change detection alarms, microburst detection with proactive congestion monitoring, and mirror-on-drop for deep visibility into packet drops and elephant flow monitoring. This telemetry is particularly valuable for the financial sector, where low latency trading can result in significant financial gains or losses. Solving network congestion within nanoseconds can save customers millions of dollars.

A Software to Silicon Solution That Delivers On the Promise of 5G and Modern Network Architectures

Autonomous vehicles, a highly anticipated technology, require not only 5G speed but also low-latency network connectivity. 5G promises vehicle response times 10-100 times faster than current cellular networks. Figure 2 highlights microburst detection with proactive congestion monitoring across modern technologies like 5G, enabling nanosecond-level visibility, which is the differentiated AIOps value Broadcom provides.

Comparing 4G and 5G latency: A car traveling at 30 mph needing to avoid an object would travel approximately 1.2 meters (4 feet) with 100ms 4G latency. With 10ms 5G latency, the vehicle would travel only 12 centimeters (5 inches). This difference is critical for safety.

Broadcom's software to silicon solution helps deliver on the promise of 5G by enabling network operations to detect, predict, automate, and fine-tune modern network architectures for critical communications and business services. The solution supports 5G use cases like dynamic traffic engineering, requiring real-time, granular visibility at flow and packet levels. Broadcom combines the Inband Flow Analyzer (IFA) with SONiC NOS metrics and AIOps out-of-band telemetry to offer a unique network monitoring and intelligent remediation solution for NOC operators and network engineers.

Deployed at the network edge to minimize latency, the solution captures 5G network packets traversing Broadcom chipsets in real-time. This data is analyzed by Broadcom's AI-driven high-scale operations monitoring solution, resulting in an industry-first packet capture and flow analysis tool. This tool enables enterprises and service providers to predict and minimize network congestion, crucial for customer safety.

Wearables, trackers, and sensors will be a significant market for massive IoT supported by 5G networks, enabling billions of devices to operate seamlessly. Network congestion could interrupt or delay critical data, such as live data from a wearable heart monitor to a hospital. Broadcom's software to silicon solution scales to support IoT and enable uninterrupted application experiences mandated in the 5G era.

Broadcom also offers a broad range of semiconductor solutions addressing key 5G design challenges, including speed, latency, density, and capacity, spanning wireless, RF, Wi-Fi, fiber optics, broadband, and networking. The highly scalable new series of chips from Broadcom significantly reduces system power and costs, lowering operational expenses for customers building and operating modern network architectures.

The SONIC-based qualified, packaged cloud datacenter solution from Broadcom simplifies deployment and operations, offers a predictable roadmap of enhancements, and is backed by a support team of over 200 engineers with deep network operating system experience.

Conclusion

Software-defined networks, 5G, and edge computing technologies enhance network agility but introduce unique monitoring challenges, complicating problem detection, debugging, and remediation. By integrating hardware, software, and silicon technologies, Broadcom provides a unique industry solution that helps operations overcome challenges posed by modern network technologies and reduce associated deployment and management costs. Broadcom delivers the industry's first silicon-integrated, AI-driven analytics platform that accelerates network operations decision-making and automation, ensuring the resiliency of modern network architectures.

For more product information, visit broadcom.com/aiops.

About Broadcom

Broadcom Inc. (NASDAQ: AVGO) is a global technology leader in semiconductor and infrastructure software solutions. Its category-leading products serve critical markets including data center, networking, enterprise software, broadband, wireless, storage, and industrial. Solutions include data center networking and storage, enterprise and mainframe software focused on automation, monitoring, and security, smartphone components, telecoms, and factory automation. For more information, visit www.broadcom.com.