At the recent Telco Cloud conference in Amsterdam, Oliver Cantor, Associate Director at Verizon, delivered an insightful presentation that outlined many key considerations and challenges in analytics, offering words of caution and optimism.
Omdia view
Summary
The presence of CSPs actively working to deploy analytics, automation and AI greatly strengthened the educational component of the recent Telco Cloud conference in Amsterdam. Oliver Cantor, Associate Director at Verizon, delivered an insightful presentation that outlined many of the key considerations and challenges in analytics, offering words of caution and optimism. Our main takeaway was that the industry has a lot of work to do but can also reap great rewards from those efforts.
The challenge with data today
Rather than look at the issue from the CSP point of view, Cantor instead started from that of the application and end-user behavior. In both cases, the discussion focused on what each needs: what resources the application needs (e.g., processing power and bandwidth) and what access the users need for those applications (e.g., time of day, frequency). The goal, he said, is to “capture data and share with the network so it understands the user’s needs.”
Of course, this is easier said than done. People want tools that can ingest and make sense of every type of data regardless of the source even though formats and naming conventions are not consistent. This makes it difficult to uncover relationships and determine which data impacts application performance/behavior. What is needed is standard formats and common naming conventions within a company and, more importantly perhaps, across the industry. Just having access to data is not enough, though. Data users need to know what problem they are trying to solve, that is, have a hypothesis that the data can help them prove or disprove.
Furthermore, because applications and data reside in a mix of clouds, data centers, and edge systems, and each vendor provides varying levels of access to equipment data, it is difficult to coordinate data sources and applications. Traffic patterns in this environment are such that data can flow bi-directionally or unidirectionally. All of this is to say that getting end-to-end visibility of applications is a daunting challenge, but critical, Cantor said, to ensure service availability, reachability, performance and reliability.
Delivering network intelligence
A Linux Foundation white paper published in November 2021 titled “Intelligent Networking, AI and Machine Learning” (see Further reading) defines intelligent networking as “a network empowered by AI technologies and systematic integration of AI and communication network on hardware, software, systems, and processes.”
To make the problem statement real, Cantor showed the combination of application, security, and network systems and operational systems multiplied by the number of regions involved in service delivery, the process of “collecting and coordinating network data for actionable analysis is hard.” What’s needed, he said, is a more agile approach.
Using IT Service Management and APIs for monitoring and reporting of the edge, connectivity, security, and application domains, with orchestration supporting closed-loop assurance and controller management APIs, CSPs can deliver a “consistent management and end-user experience.”
Cantor warns, however, that CSPs should be wary when introducing closed-loop automation and assurance because processes could create chaos due to unforeseen impacts, such as the potential conflict between assurance systems. Especially in the early days, CSPs should keep a close eye on automated actions to ensure they are working as they should.
Figure 1 shows the diagram Cantor presented, which captures the elements and flows involved in enabling an intelligent network to allow customers to “access the information they need to improve application performance.”
Figure 1: Network data transformed into actionable information
Source: Verizon
Artificial intelligence (AI) and machine learning (ML) are critical enablers of this vision. Cantor explained the process of how data is collected from network devices and applications, run through processing and modeling tools to generate insights that translate into new actions for the network devices.
He encouraged CSPs to “start simple” with the “low-hanging fruit” and “learn to trust the data and results.” Because most CSPs face similar revenue growth and cost savings challenges, they should work together on building and training models that everyone could use. He also posited that “a shared understanding of intelligent networking is needed to optimize interoperability.” And while there are “some tools in place, more R&D is needed to establish best practices.” Cantor closed by saying, “the open-source community can play a key role in furthering the development of frameworks and best practices.”
Appendix
Further reading
AI Market Maturity Survey 2022: AI Taking Root with CSPs (November 2022)
Oliver Cantor,“Where is My Data? And I Want it Now!,” October 2022.
Lingli Deng, Kaixi Liu, Yuhan Zhang, Massimo Banzi, Steve Casey, Beth Cohen, “White Paper: Intelligent Networking, AI and Machine Learning,” November 2021.
Authors
Roz Roseboro, Principal Analyst, Service Provider Transformation
