IT infrastructure: Inventory before AIOps

There is hardly a topic where digital heritage (or “legacy”) plays such an important role as in infrastructure. The inventory in your own data center is crucial when answering the question of which technologies can be used in the medium term. The awareness gained in the process often leads to a “grounding,” also in management: Those who like to talk very loudly about AI, for example, quickly become very quiet again after taking a look at their existing IT infrastructure.

In the context of infrastructure, artificial intelligence is used primarily in AIOps (artificial intelligence for IT operations). The term refers in particular to the use of AI and machine learning methods to optimize IT operations. The technology promises to make it easier to automate IT processes, detect anomalies and proactively solve problems in IT infrastructure. Overall, AIOps enables more efficient monitoring and management of complex systems, reduces manual intervention and improves response times.

The legacy challenge

It is a paradox of IT infrastructure that — unlike startups, which can simply start from scratch — large companies in particular find it more difficult to modernize and optimize, as Marc Schmidt from Avodaq knows. For him, the reason for this also lies in the “bare metal”: “A hardware-heavy infrastructure in particular presents companies with many challenges. To be able to develop future topics such as AI and observability at all, they first need modern architectures and data management platforms. Only in this way can companies manage the enormous amounts of data at all.”

So before such legacy infrastructures can be made “AI-ready,” several challenges need to be overcome. These IT landscapes are often heterogeneous and usually consist of a confusing mix of outdated and modern components, which can be difficult to integrate.

A typical example of this is the combination of a mainframe for business-critical applications and additional cloud-based microservices in which newer applications operate. The mainframe, often decades old, runs stably and reliably but uses proprietary technologies and formats that are difficult to integrate into modern data and communication protocols. At the same time, microservices require rapid scaling, containerized environments such as Docker or Kubernetes, and integration via APIs. The two worlds have different requirements in terms of monitoring, logging, and data analysis, which complicates the implementation of AIOps. An AIOps system must therefore be able to aggregate and analyze data from both environments and make intelligent decisions across the board.

However, there is currently a lack of standardized file formats, which makes it difficult to consolidate and analyze operating data. In addition, there is often a lack of clear documentation and a deep understanding of the existing architecture.

“In most cases, only parts of the infrastructure are moved to the cloud,” notes Dr. Kolja Henckel of Storm Reply. ‘These tasks are often taken over by teams that are also responsible for ongoing operations. To do this, knowledge of Infrastructure as Code must be built up within the company. This approach enables faster provisioning and maintenance of servers. The goal is to achieve more with the same teams through the resulting automation.”

Modernization at the touch of a button doesn’t exist

However, organizational resistance and a lack of expertise often prevent the path to automated infrastructure. According to Henckel, the age structure in the admin area and a lack of documentation further complicates modernization. Often, there is a lack of overview of the servers and their functions, which makes operations risky.

So modernization cannot be started at the push of a button. According to Tobias Bergs from EY, there is almost always a specific reason for this: “IT infrastructure optimization is usually only carried out when there is an incentive or a trigger. This could be, for example, problems with stability in IT operations or the potential for cost savings. It is only when such issues are apparent that decision-makers consider the widespread introduction of AIOps and the like.”

Careful planning and step-by-step implementation are therefore essential to overcome these difficulties. Bergs therefore recommends a gradual introduction of decentralized individual projects to avoid overwhelming the entire organization.

“A good database is an absolute requirement for the introduction of AIOps solutions. In addition, many companies see major challenges in the process of introduction. That is why many companies are currently introducing AIOps as isolated solutions or are piloting its use. They want to gain experience and create the basis for a comprehensive introduction.”

AIOps can put an end to finger-pointing

Even if the advantages are clear, the right story is also needed internally to initiate an introduction. Benedikt Ernst from the IBM subsidiary Kyndryl sees a certain “shock potential,” especially in the financial dimension, which is ideally anticipated in advance: “The argumentation of costs is crucial because the introduction of AIOps is, of course, an investment in the first instance. Organizations need to ask themselves: How quickly is a problem detected and resolved today? And how does an accelerated resolution affect operating costs and downtime?”

In addition, there is another aspect that he believes is too often overlooked: “Ultimately, the introduction of AIOps also reveals potential on the employee side. The fewer manual interventions in the infrastructure are necessary, the more employees can focus on things that really require their attention. For this reason, I see the use of open integration platforms as helpful in making automation and AIOps usable across different platforms.”

Storm Reply’s Henckel even sees AIOps as a tool for greater harmony: “The introduction of AIOps also means an end to finger-pointing between departments. With all the different sources of error — database, server, operating system — it used to be difficult to pinpoint the cause of the error. AIOps provides detailed analysis across all areas and brings more harmony to infrastructure evaluation.”

Overall, experts note a wide variation in the degree of maturity in the implementation of AIOps. Particularly with regard to “naturally” evolved IT landscapes, you should plan carefully and, above all, not neglect the basics to create the necessary database in the first place. A clearly defined trigger that signals the pressure to act at the decision-making level is most effective. Instead of a “big bang” approach, it is better to introduce AIOps in a targeted manner in areas where there is an acute need, to quickly achieve visible effects and generate initial benefits, for example through more efficient and secure processes. All of this not only helps to build internal acceptance but also facilitates support from management.

Florian Stocker is the owner of the communications agency Medienstürmer.