Microservices pack a powerful punch, and AT&T expects its adoption of these software blocks will deliver lower costs, faster time-to-market and reduced risk, especially when coupled with artificial intelligence and machine learning.
AT&T and Tech Mahindra partnered on Acumos, an open-source artificial intelligence platform designed to make AI more accessible and rapidly deployable. With Acumos, developers can edit, integrate, train and deploy AI microservices, according to the joint press release.
Internally, the communications service provider's use of AI and ML to empower each so-called block within a microservice enables it to learn from experience, Todd Carlton, vice president of Common Platforms and Technology Services at AT&T, told UBB2020. When the provider unveiled its microservices supplier program before AT&T Summit in Dallas this week, it also disclosed a partnership with IBM -- developer of AI powerhouse Watson. While the press release did not mention Watson, AT&T teamed up with IBM Watson for Internet of Things (IoT) analytics in May 2017. And Watson could be part of AT&T's microservices strategy.
Power of microservices
Microservices are not new. In 2016, more than 30% of 2,151 Java Virtual Machine (JVM) developers and IT professionals polled already ran microservices in production and another 20% were "seriously piloting" microservices with plans to ultimately run them in a production environment, according to a Lightbend survey.
Large enterprises, however, have been slower to adopt microservices -- in part because they were not as fast to implement DevOps, an almost mandated prerequisite for microservices for cultural and philosophical reasons. Having adopted these approaches for several years and embraced open source, AT&T is in a strong position to use microservices in applicable business areas, said Carlton.
Business departments, not IT, drive microservice use, he emphasized. Further, they must be products or services prone to frequent change that can take advantage of microservices' ability to rapidly alter to meet new demands or change features, Carlton said.
AT&T now uses microservices for virtual agent or chatbot field tech usage to research customers' cable problems, he said.
"They can use this chatbot -- which is powered by microservices -- to basically query specific systems about, 'Are you getting a trace? Are you getting a ping? Are you getting a response?' " said Carlton. "And eventually if they cannot come to a conclusion and the chatbot answers all their questions, they can say, 'Create a trouble ticket.' 'This is a critical issue. Dispatch immediately to this location.' They can tell the bot, 'Please bring this kind of equipment. Please notify these customers.' "
AT&T also used microservices to create a new system for contract management. It replaced a frequently modified process that forced agents to interact with two or three systems; now, agents use microservices to manage contracts in a more rapid, streamlined procedure, he said.
In addition, the CSP is reviewing high-cost areas such as workflow engines as potential microservice targets, Carlton added.
"We're using those two types of things: The events of product roadmap and what our clients need, and our own costs as well, to determine which microservices to invest in first," he said. "As far as which ones are going to change the most, it's really based on those two areas."
A number of microservices that use AI or ML are in the lab, undergoing "test and verify" processes. The intelligent triage bot -- which needs about three or four months of scrutiny -- reviews defects to determine whether fault lies at the end-user, developer or tester, and then assigns the problem to a particular team.
And there's more. "It can do appropriate follow-ups. It can do the appropriate escalations if required. It can redirect defects if it reanalyzes its probability and determines, over the course of time, the defect is still open and it made a wrong decision," said Carlton. "And then it can take it out of someone's queue and replace it in another team's queue."
The triage bot is accurate more than 90% of the time for certain root causes, he said. That means it could free up developers to train the service and work on future versions which will improve weaker areas such as handling general policies, added Carlton.
— Alison Diana, Editor, UBB2020. Follow us on Twitter @UBB2020 or @alisoncdiana.