Artificial Intelligence is an approach that makes a computer, a robot, or a product able to think smart as a human being thinks. The term artificial intelligence was first called in 1956, but now a days AI has become more popular due to its features: increased data volumes, advanced algorithms, and improvements in computing power and storage.
Operations, are the set of all processes and services responsible for the smooth functioning of the infrastructure and operational environments that support application deployment to internal and external customers, considering the network infrastructure; server and device management; computer operations; IT infrastructure library (ITIL) management; and help desk services for an organization.
AIOps monitoring data to look for patterns and anomalies then use computer systems capable of performing tasks, which normally require human intelligence, to make decisions and perform using automation corrective actions or configuration change.
AIOps Types:
AIOps enhances operations through greater insights by combining big data, machine learning, and visualization. AI Operations include:
- Machine learning – a type of AI that gives machines the ability to learn automatically and improve from experience without being explicitly programmed
- Deep learning – takes machine learning further by processing information in layers, where the result or output from one layer becomes the input for the next
- Cognitive computing – is based on the ability of machines to sense and reason based on learned experience, then provides the information to help a person decide
- Robotic process automation – a software tool that allows people to configure robots (computer software) to perform rules-based tasks such as accessing programs and systems, performing calculations, creating reports and checking files; within telecoms, it can be particularly useful for processes that have predictable and frequent interactions with multiple applications
- Decision management engine – runs a process or set of processes to improve and streamline action items in business processes and customer-facing applications
- Virtual agents – animated characters (usually in an appealing form) that act as online service representatives and can have ‘intelligent’ conversations with users, answering their questions
- Self-organizing networks (SON) – a technology for automating the planning, configuration, management, optimization and healing of mobile radio access networks; it was developed by 3GPP and is sometimes conflated with AI
Emergence
Of course, the processing of all the incoming on time machine-generated data is not humanly possible. Where exactly in this regards Artificial Intelligence (AI) algorithms like deep learning models excel at:
- Process ALL the data rapidly: A ML model is able to process every type of data generated by your systems and it will be doing so in the future. A new type of data is added a ML model can be relatively easily adjusted and retrained, maintaining the performance all-time high. It ensures data integrity and fidelity, products in a comprehensive analysis and tangible results.
- In-depth data analysis: When whole data is analyzed, the hidden patterns emerge and actionable percepts present themselves. The DevOps engineers can then discriminate the need for infrastructure adjustments in terms to avoid the performance backed up and can have a place at the C-suite table with specific data-based indications for infrastructure optimization and operations improvement.
- Automation of routine tasks: When done with the event patterns, the automated triggers can be settled. Therefore said, when it is displayed by the statistics that certain events always lead to a particular (negative) result and specific actions must be performed to reform the issue, DevOps engineers are able to create the triggers and automate the responses to those events.
Thus, if a monitoring result reports the increased usage of CPU due to an increased number of connections, etc., Automation can wheel up the additional app instances and use the load equating to distribute the visitor flow and bust the load. It is the simplest scenario, real-world use cases are very much complicated and allow to automate any routine DevOps task, by enabling the ML model to launch it under some specific conditions and manage the issues preemptively, not after a downfall occurs.
Business benefits of using AIOps
Deploying AIOps solutions enable the achievement of the following positive outcomes:
- Uninterrupted product availability leads to end-user’s positive experience
- Preemptive problem solving, instead, permanent firefighting
- Removal of data silos and root-cause remediation, instead of working with stripped down samples using the analysis of all the data your business generates
- Automation of routine tasks allowing the IT department to focus on improving the infrastructure and processes, rather than dealing with repetitive and time-consuming tasks
- Better collaboration as the in-depth analysis of the logs helps in evaluating the efficiency of adopted business strategies
Primary drivers for AIOps
TM Forum did a survey on AIOps drivers as part of a research report called “AI and its pivotal role in transforming operations”. In the result of that survey, they found that from the whole list displayed in the figure below there are three main drivers for AIOps:
- Delivering the best customer experience
- Reducing OpEx, by using automation and closed loop systems
- Preventing failures and outages
AIOps use cases
Use Case #1
A large North American cable operator has used Guavus Alarm IQ analytics to silence unimportant alarm noise. It applies machine learning and AI to alarm streams to classify which alarms will result in problems for customers, which are associated with open tickets and which can be discarded, with a claimed accuracy level of 99.2%. This leaves staff free to escalate alarms that indicate genuine troubles and gain greater insight into how issues that impact customers develop on the network. So-called silent failures that affect customers but don’t trigger network alarms are notoriously difficult to identify and address – NTT DoCoMo used to employ 8,000 to handle the task. Now it uses AI to classify data traffic within cells as ‘normal’ or ‘deviant’, with the latter causing silent failures. This has greatly cut costs and improved workforce productivity while improving customer experience.
Use Case #2
In October 2018 Deutsche Telekom announced a pilot scheme using AI to streamline fiber-optic rollout in Bornheim, Germany. Walter Goldenits, Head of Technology at Telekom Deutschland, said in a statement, “The shortest route to the customer is not always the most economical… The new software-based technology evaluates using digitally-collected environmental data. Where would cobblestones have to be dug up and laid again? Where is there a risk of damaging tree roots?” A measuring vehicle was forwarded to Bornheim (near Bonn) this summer, equipped with 360° cameras and laser scanners. It collects about 5GB of surface data per kilometer. Says Prof. Dr. Alexander Reiterer, who heads the project at the Fraunhofer Institute for Physical Measurement Techniques (IPM): “Such huge amounts of data are both a blessing and a curse. We need as many details as possible. At the same time, the whole endeavor is only efficient if you can avoid laboriously combing through the data to find the information you need. For an efficient planning process, the evaluation of these enormous amounts of data must be automated.” Fraunhofer IPM has developed software that automatically recognizes, localizes and classifies relevant objects in the measurement data. The neural network used recognizes approximately 30 different categories via deep learning algorithms, including trees, street lights, asphalt, and cobblestones. The trees’ root structure has a cardinal impact on civil engineering decisions. Once the data has been collected, a specially-trained AI is used to make all vehicles and individuals unidentifiable. Then the automated preparation phase follows in a series of stages. The current infrastructure is assessed to determine the optimal route. A Deutsche Telekom planner then rechecks and gives approval.
Use Case #3
Dr. Lester Thomas, Chief IT Systems Architect, Vodafone Group, notes that his company is deploying self-organizing networks in ‘emerging’ markets (the first one was Egypt) where electricity supplies can be unreliable, so sometimes cells go offline then have to optimize themselves. Using big data analytics, the surrounding sites reconfigure themselves to accommodate the changes. Thomas observes, “It’s not like static radio planning,” adding that Vodafone built the autonomous application itself using open source technology.
Conclusion
Artificial Intelligence for Operations (AIOps) is a Gartner-defined platform that joins big data and artificial intelligence (AI) functionality in order to replace a broad range of Operations processes and tasks involving availability and performance monitoring, event coordination and analysis, IT service management, and automation.
By saddling log, application, cloud, network, and metric data, AIOps makes organizations capable to monitor, flag and remove possible dependencies and problems in their systems. In terms to increase accuracy and quicker the problem recognition and streamline operations, it automates routine practices.
We at Skillfield are happy to deploy our skills to empower individuals and business to unlock opportunities by applying automation and mix it with AI on top of your data. For more details, please contact us.
References
- AI and its pivotal role in transforming operations – TM Forum Inform [Internet]. TM Forum Inform. 2019 [cited 28 January 2019]. Available from: https://inform.tmforum.org/research-reports/ai-pivotal-role-transforming-operations/
- Gartner Market Guide for AIOps Platforms Published 12 November 2018 – ID G00340492
