Technology is constantly evolving, and IT operations teams have recognized the value in swiftly implementing emerging technologies.
We’ve seen IT embrace big data, analytics, software defined infrastructure, cloud services, and DevOps methodologies to increase flexibility, efficiency, and availability. DevOps already leverages software defined infrastructure and cloud services, but what about big data and analytics? To pull it all together, we need to include one more booming technology—Artificial Intelligence (AI).
In this context, AI refers specifically to machine learning.
Like any artificially intelligent system, machine learning algorithms learn from experience to improve themselves and make more accurate predictions. Machine learning allows intelligent algorithms to perform computational analyses for us. When combined with other powerful technologies, machine learning can make more accurate predictions that provide additional insights and trigger automated responses.
The result of this harmonious melding of technology is a self-healing, self-improving AIOps platform—as illustrated in this diagram from Gartner’s August 2017 Market Guide for AIOps Platforms:
AIOps In Practice
AIOps takes one of the most valuable resources available in our modern world—data—and uses it to provide insights into IT operations that it can then act on. The data is already readily available in most IT organizations in the form of service desk tickets, system logs, application and network monitoring platforms, and even internal corporate social media.
We built an AIOps system in our iLab using logs from VMware NSX and VMware vSphere. VMware Log Insight aggregates the data, and then Elasticsearch provides analytics and basic machine learning. Finally, Red Hat Ansible automates responses to anomalies and makes changes in the NSX environment, mitigating risky behavior. Our system is a Security AIOps platform.
Other possible use cases are optimizing workload placement, identifying trends in service tickets, and optimizing operations processes.
The Value of AIOps
You may be asking, “This is great, but how do I pitch it to a CIO?” With AIOps, you can:
- Maximize the utilization of existing tools, without requiring additional head count. (Monitoring and logging systems are often underutilized because they require a dedicated person assigned to the system.)
- Reduce the number of alerts by automatically acting on many of them and suppressing ones that are redundant.
- Be proactive. Performance and availability is a common focus, but this could apply to cost control in the cloud.
Ultimately it comes down to doing more with less. By increasing insights and reducing noise, operations teams can focus more on innovation.
Some recognizable vendors have already begun investing in AIOps for their own platforms, but you don’t have to fork out big bucks for pre-canned solutions. If you have data, a way to aggregate it, a system to provide analytics and basic machine learning, and an automation platform to take actions, then you can create your own AIOps platform.
However, achieving AIOps is a large undertaking with many manual steps, requiring a wide range of knowledge. If your environment is missing a key component, you may need guidance on choosing the best tool to fill it. In its report, Gartner explains why you might work with a partner to implement AIOps:
I&O teams must take a step-by-step approach incrementally deploying AIOps functionality starting with accessing and analyzing historical data and then, at some later point, accessing and analyzing streaming data as well as applying machine learning functionality. It should be noted that both historical and streaming data analysis will require the construction and refinement of models describing the IT environment capable of generating such data.
[…] Gartner has found that it is best to begin with mastering the use of large persisted datasets ingested from a variety of sources. Only after the IT operations team has become fluent with the big data aspect of AIOps should it attempt mastery of the capability categories. Hence, when selecting tools or services, an enterprise should prioritize those vendors that allow for the deployment of data ingestion, storage and access independently of the remaining AIOps components, but nonetheless support the gradual addition of those other functionalities.
You can download the full Gartner report here.