When each is the best choice?
Organizations are striving to optimize their operations and put on autopilot everything that doesn’t need a dedicated human operator. It boils down to cost and efficiency. It is a matter of saving expensive human resources and using people for what they do best: thinking and problem solving instead of tedious, repetitive tasks. In this context, Artificial Intelligence offers a natural evolution of DevOps. It promises to help organizations monitor their operations and detangle some of the exponentially increasing IT complexity.
What is Artificial intelligence for IT operations
AIOps is not a single technology, but rather a group of processes, including machine learning, big data analysis, artificial intelligence, all working together to help IT remain up and running and provide tech teams with valuable insights about performance, issues, event correlation, and resolutions. The platform uses both historical data from an organization as well as real-time logs.
The final goal of an AI-powered solution is to analyze large volumes of data to help the IT department manage, optimize, automate and protect while also performing security alert triage.
The Digital Transformation of IT Operations
All companies face two challenges: increased IT complexity and pressure from stakeholders to process everything faster and accurately. The reality is that at the current rate, humans will no longer be able to offer real-time support without consistent help from automated solutions.
AIOps aims to be a total digital transformation of current operations. Still, reality shows that it will have a gradual penetration into existing systems, for well-defined, easy to automate tasks at first, slowly growing towards full-scale AI applications. NOC and SOC teams are already testing AI-powered process automation platforms at least at a project-size scale.
Domain-centric and domain-agnostic tools
Depending on their affinity for specific data types or data sources, the deployment of the AI-powered platform can be classified as domain-centric if they prefer only specific inputs and, respectively, domain-agnostic if the platform can process a wider variety of diverse data points.
There is no one-size-fits-all. Each has pros and cons. After careful analysis of the project’s needs and constraints, a decision will be made only after considering the ideal system’s capabilities.
Domain- centric
These are a laser-focused subset of platforms. Their goal is to provide in-depth performance analysis for a specific topic or a particular use-case. One can expect that a domain-centric solution is less complex and less likely to fail, also providing a quick ROI.
These types of AI platforms are expected to have success first, according to Gartner’s recommendation to “Increase the odds of a successful AIOps platform deployment by focusing on a specific use case and adopting an incremental approach that starts with replacing rule-based event analytics and expands into domain-centric workflows like application and network diagnostics.”
The downside of domain-centric platforms is that they lack the ability to correlate operations data and events with the entire IT environment, thus having limited applicability.
Domain-agnostic
As the name suggests, these types of platforms integrate well with various data sources – across all types of industries, as well as across all segments and functions of an industry (such as IT, manufacturing, security, and so on).
The domain-agnostic tools are beneficial for a broad range of applications and data sources, including pre-processed or data from other monitoring tools and domain-centric systems.
The domain agnostic AI-powered platforms are useful for making correlations between different parts of an extensive IT system and performing root cause analysis for identifying the failure causes across the entire system.
Which one to choose?
There was some discussion at the beginning of AIOPs in 2018, if this new technology aims to replace domain-centric monitoring tools. The short answer is no. It provides better analytics and can work in tandem with nonAI domain-centric tools by retrieving data streams from these and processing them in a centralized manner.
AI-powered Tools Vendors
Since domain-centric and domain-agnostic serve very different purposes and are tools that can work together, we can expect mixed solutions in the future. This is not meant to be a top of vendors, just a list of examples currently available on the market.
Domain Agnostic platforms
Companies choose domain agnostic when they need disparate data in a single place to automate processes, including careful analysis and alert triage. The platform should be able to handle any type of data and in any format.
Some commercial tools which are useful for this work include:
- Siscale’s ai Alert Triage – probably the first really AI-powered process automation platform, is fully designed for IT systems to perform alert triage and integrate with various monitoring and detection tools. Look into this solution if you are willing to reduce alert fatigue in your organization drastically.
- Big Panda deals with preventing IT outages which could lead to money and credibility loss
- Moogsoft – a tool that helps IT teams analyze their data streams by independently creating a data collector. Best for monitoring external services.
Domain Centric platforms
These are more focused and specific tools, and their role revolves around processing data coming from a particular source or a special type of data. Operations teams rely on such tools if they only need to adopt AIOps for answering specific business questions which can’t have multiple causes.
There are hundreds of dedicated such AIOps platforms, some of which are open source, but we will focus on a few:
- AppDynamic belongs to Cisco and provides data collector facilities, integrations with the most popular systems, and its signature Central Nervous System platform.
- DataDog with WatchDog a platform created for anomaly detection, outlier detection, forecasting to alert the right team in real-time without the need to set alerts. It just learns what is expected.
- StackState promises to pinpoint root causes as fast as possible with autonomous anomaly detection. It can upgrade the existing incident management process without replacing underlying tools.
The importance of an open AI data platform
The need for AI-powered tools is dictated by the failure of existing monitoring tools to correlate different data types and offer a bird’s eye view of the organization’s problems. As systems become more complex, they need to be scalable, and manual reporting is not feasible anymore.
While domain-centric solutions might offer quick wins, a domain agnostic choice can change the workflow completely, offering improved security and better response time.