5 Steps to Get Up and Running With AIOps

Bob Friday, CTO and Co-founder, Mist
Originally posted on CMSWire

It’s a bit of a mouthful as a tech buzzword, but AIOps is starting to gain attention — and traction — across the industry. And with good reason. AIOps is the stepping stone to something that has never before existed: a true network AI assistant that can answer questions on par with a human network expert and take proactive action on problems it detects.

AIOps (artificial intelligence and operations) refers to algorithms being adopted by IT operations to address the growing data and complexity in networking and, at the same time, the continuing pressure on IT budgets. AIOps solutions use the same kind of machine learning and advanced analytics technologies behind Google Maps or Uber’s predictive ride pricing models to help IT departments anticipate and fix problems before users even realize they’ve happened.

In the AI self-healing network of the future, users will enjoy steady performance and companies can stop asking their precious IT resources to scurry around with blinders trying to solve a pile of end user support tickets.

AIOps holds especially exciting promise in assuring speed and reliability of wireless networks. Wi-Fi has taken its place alongside water, power and light as a must-have technology on which businesses are building mission-critical services for consumers and employees in today’s highly mobile, app-driven world. Thus, it must be more predictable, measurable and easily managed than ever.

All of this means enterprises need more visibility into and actionable insight from the petabytes of data flowing through their wireless networks and, in real time, proactively and automatically tuning their infrastructure to avoid user impacting failures and optimize performance.

It’s important for CIOs to understand the right steps to leveraging AIOps as a foundation for an AI-powered self-driving network. Here are five priorities they should keep in mind as they get started.

1. Set a Course for True AI

As in AI systems for self-driving cars or diagnosing medical conditions, true AI for IT operations gets more intelligent as it analyzes more data over time and is always improving its capabilities for autonomously monitoring and healing the network.

Some solutions on the market purported to be AI are really just fancy data collection or number-crunching products. They do a good job gathering and analyzing statistics, but they lack the foundational AI technologies to you to a true self driving AI network. Make sure to look under the hood of your AIOps platform and make sure it contains these elements:

  • Data pipeline: it can support data from multiple data sources
  • AI primitives: the ability to add domain expertise to the data
  • Library of data science algorithms: deep learning is in the toolbox
  • A user interface that allow everyone in the business to use the solution, thus democratizing data

It’s important for CIOs to know the difference and set a strategy that embraces AIOps as the path to a self-driving network.

2. Harmonize Your Data

Just as it takes great grapes to make great wine, clean, integrated data is required to do AI right. After all, an AI system needs to analyze and learn from data to perform its function. It can’t do that if the myriad devices spread across the enterprise — WLANs, WANs, routers and firewalls — aren’t able to share and correlate data.

Thus, it’s essential for companies to eliminate any data siloes across the IT stack and bring together the many disparate systems into a cohesive, data-sharing whole.

3. Make Sure You Have the Right Skills

AIOps requires a dramatic re-orientation of the IT staff, from a command line interface paradigm for configuring boxes to an API programming model paradigm for getting actionable insights from data.

Traditionally, the training of most network administrators has focused on configuring boxes. That’s not enough for AIOps, which demands a broader skill set capable of taking data from a variety of boxes and then applying AI in a common, interoperable format.

It’s important for CIOs and other company leaders to understand this distinction and act accordingly in recruiting, hiring and re-training. Strong development skills are necessary.

4. Understand AIOps Is all About a Distributed Software Cloud Architecture

The first generation of cloud managed endpoint technology, still often used today, essentially consists of on-premises controllers using embedded software architecture. While these products made deploying and managing wireless networks easier by using an embedded software architecture, a self-driving AI network of the future requires a distributed cloud software architecture that allows you add new AI models to added quickly.

Companies that want to get into AIOps need a distributed microservices architecture that can apply algorithms across various parts of the network, providing cross-enterprise visibility and allowing repairs and new features to be implemented quickly.

5. Embrace the Cloud

While organizations have been moving to the cloud for sales, HR, finance and other lines of business, networking has been slower to adopt the cloud.

CIOs need to recognize that the cloud is AIOps’ best friend. The cloud provides the scalable infrastructure needed to gather actionable insight from the data coursing through wireless networks.

By following these five steps, companies can start taking advantage of the exciting opportunity that AIOps offers to build self-driving networks that are easier to operate and more enjoyable to use.