everythingpossible - Fotolia
Advances in processing power has made AI and machine learning more viable, leading to an IT operations management strategy known as AIOps. As AIOps starts to take root and IT administrators look to maximize their capabilities, they must adopt AIOps tools, gain more perspective, clean up their data, understand the AIOps deployment process and make decisions cautiously to protect their data centers.
"It's impossible for companies to scale without using AI and related analytic technologies due to the volume, variety and velocity of data coming into virtualized operational systems today," said Stephen Elliot, program vice president at IDC.
Virtualized computer infrastructure has been expanding in all directions -- meaning these infrastructures must handle more application source code, endpoints, device diversity and system checkpoints -- so manual system management has become more challenging. This is where AIOps comes in.
Recognize the need to adopt AIOps tools
Legacy virtual computer infrastructure management tools collected data at devices' entry and exit points, compared current performance with traditional metrics and identified anomalies. Generally, a system administrator uses the information to troubleshoot and remediate any performance problems.
This approach is no longer viable in many corporations. The amount of data monitored and generated exploded after changes to agile software. David Link, CEO at ScienceLogic, a software vendor based out of Reston, Va., coined the term "the alert swamp," which refers to how the massive volume of notifications overwhelms modern technicians.
AIOps has the potential to help admins drain the swamp by automating the process of separating inconsequential noise from items that admins must act upon.
Gain more perspective
New tools can enhance system management, but they also require admins to change business processes and employees to develop new skills. Admins must move away from segmented management and support staff should understand how infrastructure elements -- such as virtualization, servers, networking and storage -- synchronize.
In addition, admins and their staff must understand how vendors build and change applications. Development responsibilities branched out from IT and into different departments, meaning business line managers increasingly write code for applications. To communicate effectively with these individuals, IT staff must gain an understanding of how the infrastructure supports different departments' business goals.
Admins must cross technology, organizational and process boundaries to obtain a broader view of operations than they did in the past. Consequently, they must also recalibrate their skills.
"One area where IT support staff needs training is gaining a basic understanding of statistics and probability, because many management tools rely heavily on analytics," said Sanjit Ganguli, research vice president at Gartner, an IT research and advisory firm based out of Stamford, Conn.
Clean up the data
Improvements in technology enable admins to monitor more system infrastructure elements, which results in vast amounts of information to analyze.
"A big challenge involves understanding your data and building a culture that promotes data science and data hygiene," Ganguli said. "A lot of enterprises are weak in this area and create bad data. As the saying goes, 'garbage in, garbage out.' The new tools will not deliver performance improvements if the underlying data is flawed."
One issue admins must face is identifying items. Organizations have a variety of system management tools, and each has its own nomenclature. AIOps consolidates such information, and organizations must provide consistent terminology to help with this consolidation.
Admins' management tools also collect event information from many different applications, which can result in a large volume of duplication. Mapping connections from events and determining cause and effect requires a lot of trial and error on the admins' part.
AIOps means deploying more than one product
Given the complexity of AIOps, admins cannot deliver a quick and easy fix.
"When examining AIOps, businesses cannot just focus on deploying one of the available solutions," explained Phil Tee, chairman and CEO at software company Moogsoft and founder of AIOps Exchange, a not-for-profit AIOps discussion forum. "People and processes are what really make these systems work."
Gartner breaks the deployment process up into to four phases: descriptive, diagnostic, proactive and root cause analysis. Admins start by understanding what their management systems do. Next, they begin to correlate the alerts flooding the data center with those management systems. Third, they shift their response posture from reactive to proactive. Finally, they build a system that understands what is happening, identifies developing problems and takes corrective actions in an automated fashion.
Protect your data centers
The AIOps market is in an early stage of development. As a result, many startups have begun to emerge. A few significant purchases have already taken place: Splunk bought SignalFX, Cisco teamed up with Perspica, and New Relic acquired SignifAI, according to Elliot.
"Customers need to protect themselves after acquisitions," he said.
Admins must rely on open interfaces whenever possible so that they can continue to use a startup's tools in the future, no matter what direction their acquirer takes. The nature of system management can change dramatically and many admins struggle with the amount of data that now inundates their data centers. AIOps offers admins a way to make sense of the information.