AIOps – GAVS’ Vision

When we experienced Windows 3.0 – the matured GUI version of Operating system three decades ago, it was a defining moment for the IT world.  Five years later, when Citrix released Win Frame – a software that enabled running desktop applications in servers using browser, it was another WOW moment for the IT industry. Then we saw another great moment in 1999 through VMware workstation which allowed us to run multiple instances of OS on a single hardware. Then came the smartphones, cloud, Software Defined Data Centers, Big Data and now Artificial Intelligence. We are at the threshold of unleashing the power of the artificial intelligence to help humans.

Are machines better than humans? Humans win on the ‘intelligence quotient’ over machines, but when it comes to doing repeat tasks, machines score over humans by doing those repeat tasks error free any time, every time. So the Artificial intelligent programs combine the accuracy of machines and intelligence of humans. We at GAVS have been nurturing Artificial Intelligence for the past 5 years and we have a vision for it.

Artificial Intelligence driven IT operations (AIOps) is performing IT operations through self-learning and self-correcting systems. It deploys Machine Learning techniques to understand the ever changing IT environment, Artificial Intelligence to detect abnormalities and intelligent automation to remediate abnormalities before it impacts.

One of the key characteristics of AIOps is ’self learning’. The platform should be capable of understanding the physical and logical relationship between assets installed in the environment and the behavior of those assets through ’self learning’. It should not be depending on CMDB (Configuration Management Data Base) and / or ADDM (Application Discovery and Dependency Mapping).  It should not require any rules or configurations. It should learn from the events occurring in the environment and the way those events are occurring.

The above characteristic requires lots of data into the AIOps platform.  The quantity of data decides the accuracy of the correlation & prediction and the quality of the data decides the number of defects. AIOps platforms should not be depending on other tools to provide the data it requires to correlate and predict. It should know what kind and type of data it needs and also have the capability to generate those data.

Not all systems in an environment behave in the same way. Every system has their own characteristic and behavior. A general rule that is applied across all systems leads to false positives. AIOps platforms help move away from ‘rule based alerts’ to ‘pattern based alerts’. It should generate alerts based on historic performance & consumption taking into account the day, time, load, related services and devices. Then it should correlate based on the pattern by which the events, logs and alerts are generated in the environment. Further it should predict the performance and health of the applications and prescribe the remediation if any. Any rules or configuration or inputs are alien to a true AIOps platform.

Needless to say, the AIOps platform should be able to read the data of any format from any source, using any method. As far as possible it should be least intrusive and eco friendly. It should be able to receive the data at the speed of generation of those data at the source, Process the data at the speed of consumption by AI programs and display the results at the speed at which the users want to see.

Core of the AIOps platform typically has three components – Monitor, Analysis and automation. These are three different agents running on the server today and they handshake between them. GAVS sees this model metamorphose into a single component consisting of all above three components. This AI agent knows what it needs to check,  when to check and what action needs to be taken when it encounters an anomaly. However this agent is still outside the application. GAVS foresees that the agent would be in built into the application in the coming years. Every application will have the AI built into it and ensures defect free running. GAVS is collaborating with Indian Institute of Technology, Madras, India in building this capability through Reinforcement Learning.

The role of AIOps in resolution of incidents should not stop at just correlating the events, but go beyond it. While the correlation brings all the alerts associated with an incident, it does not provide further insights to move forward in the right direction for the administrators. One of the things it is expected to do is to display a graphical view of the journey of transactions showing the devices they touch through, with response times of the past, present and future between the devices end to end.  This helps the administrators to quickly narrow down to the devices and the path for analysis.

The ability to predict the behavior of even the smallest physical or logical component in a system, should lead to prediction of the behavior of the servers and then the applications under which those servers come and finally the business processes which are constructed using such applications. GAVS has a methodology to predict the Application Health Index (AHI), which is a function of performances of all the components of all servers, associated physically and logically with that application. The same methodology can be extended to predict the health of the business processes.

Artificial Intelligence will play a key role even before the application is rolled into production. It can very well be deployed during QA stage to predict how the application would behave when deployed in production.  The AI engine mines business requirements, test cases, unit test cases, defects logged & xRTM from knowledge repository to build predictive patterns of how application quality would result in deployments .

We let humans do better things than what a machine can do. As machines become more and more intelligent, humans become more and more wiser.

About the Author:

Chandra heads the IMS practice and IP at GAVS. He has around 25+ years of rich experience in IT Infrastructure Management, enterprise applications design & development and incubation of new products / services in various industries. He was with Genpact earlier, where he had served as CTO for about 7 years before becoming the leader of IP Creation. He had built a new virtual desktop service called LeanDeskSM for Genpact, from ‘concept’ to ‘commercialization’.

He has also created a patent for a mistake proofing application called ‘Advanced Command Interface”. He thinks ahead and his implementation of ‘disk based backup using SAN replication’ in one of his previous organizations as early as in 2005 is a proof of his visionary skills. Chandra is a graduate in Electronics and communication engineering.