What is AIOps?

AIOps is artificial intelligence for IT operations. It’s both an IT operations approach and an integrated software system that uses data science to augment manual problem solving and systems resolution. AIOps combines big data and artificial intelligence or machine learning to enhance—or partially replace—a broad range of IT operations processes and tasks.

Before the AI part of AIOps can work, it needs something to work on. It needs data—operational data. Specifics like uptime, downtime, processing use, network traffic, application logs, errors, authentication attempts, and firewall alerts, as well as historical data. This data collection, organization, and cleaning is usually harder than incorporating the algorithms and learning models.

With that data established, it’s time to determine service level objectives and indicators. Define that operational health using trackable metrics, which then become the baseline of an AIOps system. Many enterprise platforms come with (or connect to) operational observation components: Red Hat® OpenShift® includes Red Hat OpenShift Observability, Red Hat Enterprise Linux® uses Red Hat Satellite, and Red Hat Ansible® Automation Platform uses Prometheus and Grafana.

With operational health defined, you can apply AI. And it’s easier than ever to incorporate AI into projects. 

• Use IBM watsonx Code Assistant (including as part of Ansible® Automation Platform via Ansible Lightspeed with IBM watsonx Code Assistant), and IBM’s quantum computing fleet is even available as-a-service to process all those data sets.
• Tap into OpenAI’s ChatGPT—a generative AI system—using an application programming interface (API). 
• Use Google’s Natural Language AI as a cloud service. 

With all these opportunities, it’s no wonder natural language processing (NLP), AI, machine learning (ML), and deep learning (DL) have become part of our cultural lexicon.

Benefits

• Resolution speed: AIOps reduces downtime by detecting and reacting to emerging issues, decreasing mean time to resolution (MTTR).
• Self-healing systems: Self-healing infrastructure can significantly improve performance and uptime. 
• Big data: AIOps can put big data to use by cleaning, analyzing, and taking action on it.
• Efficiency and scale: Increase staff efficiency by using insights from AI models to identify actions and scale detection.
• Innovation: With repetitive work eliminated, IT teams can develop and deliver more strategic and higher value projects.
• Simplification: AIOps can streamline many repetitive IT service management tasks.
• Real-time data correlation and decision making: When AIOps includes an automation engine, it can respond automatically based on data—reducing human intervention and error while minimizing noise.
• Scaled data correlation and prediction: AIOps can automatically analyze every possible permutation, to degrees far beyond what humans can do manually.

Challenges

• Expertise: With extensive data science expertise required, even getting started has a high barrier of entry.
• Infrastructure: Without standardized platforms and capabilities (like those provided by Red Hat OpenShift and Ansible Automation Platform), training AIOps to your specific infrastructure can be challenging.
• Time to value: AIOps systems can be difficult to design, implement, deploy, and manage, so it can take some time to see any return on investment.
• Data: The volume, quality, and consistency of data produced by modern IT operations can be overwhelming and difficult to wrangle, and AIOps outcomes will only be as good as the quality of the data sources.
• Collective agreement: Baselining system health and setting standard operating goals requires mass buy-in from many parties—consensus that can be difficult to attain.
• Scope: The number of considerations can be overwhelming to even get started. Or the environment can simply be too dynamic to baseline.
• Failure rates: AI projects have a huge failure rate. According to IDC's AI InfrastructureView, 31% of the study’s respondents have AI in production—but only a third of that number have realized organization-wide benefits.

Let’s get into the specifics: Why would different types of professionals use AIOps?

• Application site reliability engineers (SREs) can define the 4 golden signals the AI can focus on: latency, error rate, traffic, and saturation.
• Developers can use AIOps analyses to perform their own root cause analysis (RCA), or developers could allow the AIOps engine to perform RCA without human intervention.
• Business owners can use AIOps to monitor the same golden signals used by SREs to understand applications’ performance from end users’ perspectives.
• Infrastructure operators can use AIOps to monitor hybrid cloud, multicloud, and microservice-based IT environments—from a few dozen virtual machines (VMs) to thousands of clusters—and simplify day 2 operations.

Each of these use cases illustrate that AIOps helps teams detect and react to potential issues, but we’re not at a place where AIOps systems can replace experienced IT systems administrators and other operations team members. AIOps—like most IT revolutions—just makes machines do our chores while we stay in the driver's seat.

So machines aren’t replacing humans. But data scientists and DevOps engineers alike should still take advantage of the incoming IT revolution to broaden their skills.

• Application performance monitoring (APM) will become more important as businesses look for candidates with performance-driven backgrounds.
• Automation skills will become more important to understand, incorporate, or write the underlying AI scripts, as well as turn an event correlation and alert engine into an execution engine. 
• If you’re already versed in AI, find reasons to start (securely) experimenting with network AI (think: SD-WAN, Wi-Fi, etc).

DevOps is all about making small, incremental improvements along the entire application life cycle—constantly. So the bane of DevOps is downtime, which is where AIOps comes in. AIOps augments DevOps culture by adding data science to development and operations processes. 

AIOps doesn’t replace DevOps—it’s an evolution of DevOps. AIOps is another point along the same digital transformation life cycle. AIOps and DevOps share the same responsibilities. AIOps just augments human intelligence with a mechanized brain. 

While the actual lines between DevOps and AIOps blur quite a bit, AIOps fits nicely on either end of DevOps processes:

• On the front end, AIOps can consume huge amounts of infrastructure data, alerting DevOps engineers of underlying integrated development environment (IDE) issues (or just fixing them outright). 
• On the tail end, AIOps can automatically resolve redundant IT issues in production—all while learning to remediate novel bugs that come with each new incremental release. 

Like DevOps, there’s no single AIOps tool, AIOps platform, or AIOps product. The tools you use to build DevOps and AIOps capabilities are as numerous and unique as your IT stack (hardware and software). That’s because any AIOps solution you build has to integrate, analyze, and act across everything that makes your development and production environments so unique.

AIOps has a deep presence in open source—both as upstream projects and within many communities. While no single product is a complete AIOps solution, there are many open source development, operations, AI, and automation projects that can be used as part of your unique AIOps solution. And there are also many specific open source projects being developed to provide AIOps solutions to specific AIOps problems.

Companies are releasing their downstream AI product code as upstream projects:

• Meta—the world’s largest social media conglomerate—released the Llama 2 large language model as an open source project.
• We at Red Hat are hoping the Project Thoth open source project will lead to enterprise-grade hardened products in the same way that Project Wisdom led to Ansible Automation Plaform’s Ansible Lightspeed with IBM watsonx Code Assistant component.
• We’re also contributing to AIOps projects led by other organizations, like the Artificial Intelligence Center of Excellence’s (AICoE) AIOps project.

Our automation platform combined with our partners’ AI capabilities can give your enterprise a massive head start to coding a strategic AIOps solution—pairing AI’s observability capabilities with our automation engine’s event-driven architecture. 

Use Event-Driven Ansible to take action against your AI’s findings. Pair our automation platform with our partners’ causal AI engines (like those provided by Dynatrace and other modern observability tools). And use Ansible Lightspeed with IBM watsonx Code Assistant to help developers and operations teams across all skill levels write syntactically correct code with AI-generated recommendations.