Assessing the Impact of AI to Cloud Networking


Ubiquitous cloud access has become more business-critical than ever and monitoring, managing and trouble-shooting have increased in complexity. SD-WAN allows organizations to seamlessly bring together diverse circuits like MPLS, DIA Internet, broadband Internet, WiFi, satellite and any other network transport with an IP address.

We can already see some of the ways in which AI can optimize WANs to improve business outcomes. For example, ubiquitous networking has become a critical layer in the network landscape to ensure reliable, cost-effective and highly scalable connectivity. The challenge is in managing the ubiquitous WAN’s real-time decision making, based on network conditions, locations and application types.

Ubiquitous Cloud Networks need AI

With artificial intelligence, IT automation will determine what circuit traffic is using or which cloud to access. AI is already playing a major role in fulfilling automated, real-time intelligent decision-making of moving traffic among clouds. AI and machine learning, with its autonomic-oriented algorithms, are likely to act as the brains that orchestrate the enterprise WAN of the future.

Rather than responding to adverse network conditions, like latency, packet loss and jitter, AI will take advantage of its predictive attributes to move traffic to the best circuit and paths, understanding the intent of the application and always delivering a reliable user experience. AI will also evaluate elements beyond conventional quality measures such as network responsiveness (e.g. TCP window-size, handshakes, SYN-ACK exchanges, retransmissions) correlated to specific applications. When AI is integrated into SD-WAN, it can associate quality data with underlay data to learn a more accurate and detailed representation of the application experience by the user.

Beyond ensuring WAN reliability, AI can protect traffic in the same way. With real-time data collection and network analytics, AI can anticipate and predict traffic patterns and anomalies to protect WANs from cyber-attacks. Today’s WAN security platforms use algorithms to find abnormalities, such as traffic build-ups and malicious activities. An AI-powered WAN will use more intelligence-based algorithms that quickly anticipate threats to avoid attacks.

IT operations and analytics administrators clearly understand the complexity that network evolution has created. Using SD-WAN centralized orchestration to manage network underlay devices and systems, helps simplify management. However, as WANs become ever more expansive and distributed, the solution to managing them will come from analyzing real-time data collected from all circuits, devices, sensors, applications, devices and services.

It is unrealistic to expect ITOA to manually correlate and analyze omnibus data. And that’s ok, because this is where AI excels. Utilizing the distributed power and virtually unlimited scalability of cloud resources, AI can bring together a confluence of correlated data from diverse systems and network layers, analyze it, make recommendations, or take policy actions to improve network performance.

AI Will take Intelligent Networking to Another Level

The many elements that affect WANs, and their immense volumes of data are overwhelming IT. AI will be a welcome contribution to managing everything connected to the network, taking the immense burden off of IT.

These are important times for enterprise IT teams making big decisions about SD-WAN. As we move forward, timing is everything, and the decision to utilize SD-WAN driven by AI is sure to play an integral role in bringing business connectivity and communications forward. Believe me, no enterprise will want to be late on making these decisions when they swing the proverbial bat (or WAN) for the fences.

Multi-Cloud Imperative

Most organizations today have a mix of enterprise software/hardware needs on various cloud platforms. Research by IHS Markit found that enterprises are expecting to use on average of 11 different cloud service providers (CSP) by 2019 to meet all of their off-premises cloud needs.

This new paradigm of a “cloud of clouds’ within the enterprise, or Multi-Cloud, is creating its own set of management and security challenges.

  • Application Clouds: SaaS applications that run in their own cloud, e.g., Salesforce
  • Infrastructure Clouds: Infrastructure-as-a-Service such as Google or Azure
  • Telecom Clouds: Verizon Cloud for mobile network operations
  • Enterprise Clouds: MPLS-based private networks incorporating branch offices
  • Partner Clouds: typically leveraging co-location facilities to create virtual POPs

In the context of SD-WAN, the enterprise “cloud” thus can be viewed as the “hub” or interconnection point for the mother cloud of multi-cloud connectivity.

Connecting branch locations to each CSP creates another level of complexity on a network landscape that already is a management and operational nightmare. The situation worsens when IT teams are unable to monitor application performance in real time, ensure that the network is secure yet agile and resilient, while automating performance tuning to assure SLA attainment between enterprise sites and cloud.

Organizations turn to the cloud because business users and customers today demand a more flexible, open, anytime anywhere access model for enterprise applications. And above all, application performance and user experience predictability have become a critical factor for IT teams today.

However, conventional network architectures don’t support the cloud-driven application consumption patterns of today. Enterprises need to re-architect the network for multi-cloud and SaaS applications. But how and with what?

Versa envisions elements of AI for intelligent application detection and underlay performance monitoring to make dynamic adjustments (infrastructure) that steer application traffic according to a perceived experience. While such functionality, in part, represents core SD-WAN value, Versa accelerates the beginning stages of AI implementation by being able to discern the user’s level of satisfaction and user experience by learning from the app how to best automate infrastructure changes.