The New Requirement: Advanced AI-Driven Network Triage and Remediation

Broadcom is a proud sponsor of  ONUG Fall 2019 in NYC. Join us as we showcase how to simplify Hybrid-Cloud and SD-WAN deployments by combining AI and ML with rich granular data captured at the chip level for one-of-a-kind AIOps analytics and automation.

The highly complex nature of today’s modern hybrid IT architectures continues to present increasing challenges for teams relying on traditional, siloed or domain-specific monitoring approaches to deliver a reliable digital experience. With these disparate tools, it will remain extremely difficult and time consuming for teams to establish integrated, end-to-end service monitoring, efficient root cause determination and actionable intelligence that drives automated triage.

Today’s AIOps solutions promises to break down these monitoring silos by enabling advanced service-centric analytics and automation across applications, infrastructure, and networks. They should be built on an open data lake with applied AI and ML that delivers Full Stack Observability, Service Driven Business Intelligence, Smarter Insights from Algorithmic Root Cause Analysis, Intelligent Automation and Proactive Remediation.

Additionally, network teams are experiencing challenges today with managing complex network architectures like SDN, SD-WAN and NFV to deliver an innovative, reliable, and responsive digital experience.

Costs is one major challenge. It is expensive to run network operations today and requires many personnel to architect, monitor, understand patterns and triage these new architectures.  Recent studies show that an enterprise loses an average of $9k every minute during a data center outage. Some of our customers have reported up to $20M in losses to data center outages per year.

Speed is another challenge. Today’s digital experience requires a new level of operational responsiveness to be able to identify, diagnosis, react and solve problems quickly to avoid bad customer sentiment and mounting operating losses.

It is critical to bring AI and ML through an advanced AIOps solution to solve for reducing costs from operating today’s networks, avoiding outages and to speed up root cause analysis for faster triage and operations.

AIOps can eliminate hours, days and weeks spent on manual pattern identification of trend charts looking for any unusual activity in the network. It can enable advanced root cause analysis and anomaly detection through intelligent thresholding and alarm noise reduction. Furthermore, AIOps enables advanced predictions, correlations and automated network triage. AIOps can learn from analyzing network activity to automate, repair and tune the network for reduced operational costs and faster triage to deliver consistent and exceptional digital experiences.

We all understand no one solution fits all problems and network monitoring is no exception. Some next-gen use cases such as dynamic traffic engineering require granular visibility at flow or packet level in realtime. AIOps solutions should be flexible, scalable and enable real-time monitoring that leverages in-band and out-of-band Telemetry technology to deliver granular telemetry at flow or packet levels. It should also support several key out-of-band telemetry use cases such as streaming telemetry directly from the ASIC. AIOps solutions should help solve for popular use cases such as Packet Path Tracing with Route Change Detection Alarms, Microburst Detection with Proactive Congestion Monitoring, Mirror On Drop providing deep visibility into packet drops and Elephant Flow Monitoring.

Telemetry network data captured at the chip level provides timed stamped packet path tracing enabling per hop latency visibility.

AIOps solutions combined with chip level telemetry data enables disruptive advances in Enterprise data center and campus networks including.

  • A scalable architecture, with bandwidth from 2.0 to 12.8 terabits per second
  • Extensive compile and run time programmability, allowing changes to behavior during switch operation
  • An extremely rich feature set with a high degree of feature concurrency
  • Large, fungible databases and an industry-leading on-chip packet buffer”
  • 10 Tbps in a single device, delivering hundreds of Mbps (Megabits per Second) per subscriber
  • Elastic Pipe, delivering advanced packet processing, programmability, virtualization and investment protection
  • Large scale, carrier grade, expandable IP-MPLS forwarding
  • Hundreds of milliseconds of deep buffering
  • Advanced Telemetry and Instrumentation

Today’s AIOps solutions are operationalizing hybrid-cloud environments by combining AI and ML with rich, granular packet data captured at the chip level. These solutions enable you to use Business Service KPIs to track the overall health of your infrastructure, identify potential outages and gain insights and suggestions on probable root cause; along with targeted visibility down to the packet-level for one-of-a-kind advanced network analytics and automated remediation.

See an example of AIOps for Networks in action here:








Author's Bio

Jeremy Rossbach

Sr Product Marketing Manager, Broadcom