As I covered in my previous blog, AIOps is an exciting new area with potential for addressing network troubleshooting, remediation, and performance optimization. Potential, however, does not equate to a solution: approaches to AIOps vary by vendor, and the variations have led to confusion about precisely what AIOps requires, how it operates, and concern with the benefits that it can deliver. NOC teams adopting AIOps are often disappointed in the actual improvements realized. This general confusion begs the question, “where is the AI in AIOps?”
To help address this concern, Brillio and Blue Planet/Ciena have taken a pragmatic approach to building a framework for AIOps designed to remediate most-common network problems with automation control and with built-in validation designed to accelerate AI/ML benefits for a specific network. The building blocks of this approach include:
Providing IT administrator validation of remediation outcomes to help fine-tune root-cause diagnosis along with the effectiveness of the steps in remediation, as they apply to the organization’s unique network deployment
This approach to AIOps is the basis for a new solution for network automation, available now from Brillio and Blue Planet/Ciena in the Blue Planet Enterprise (BPE) Automation Suite became available on Sept 3. For more details on the solution and to request a demo or free trial, check out our new Brillio microsite on BPE providing Dynamic Configuration and Change Management (DCCM) and Intelligent NetOps (INO) AIOps capabilities.
In a typical approach to AIOps, data scientists will examine large sets of historic data attempting to identify insights that can help IT teams to operate their networks with higher performance. In practice, however, the data is so cluttered with edge conditions and exceptions, that the analysis produces a smattering of micro-improvements that are considered largely “noise”. It is rare that this approach produces significant results. In fact, this approach to AIOps offers little improvement of traditional rules-based tools.
In contrast, the BPE approach is to derive root cause symptoms for most-common network problems in non-cluttered pristine network conditions. ML is then used to determine best how to recognize and interpret those conditions when they occur in a real customer environment. Root cause analysis (RCA) using this approach can be efficient and accurate. BPE then provides detailed, step-by-step remediation that is proven to address the exact root-cause condition – and presents this for push-button selection in a closed-loop automation approach which is simple for IT administrators to work with.
In fact, ease-of-use is critical to ensure that AIOps is practical and delivers real benefits when it is needed most.
As another key step, administrators using BPE are asked to validate that the remediation worked as expected to fix the problem, enabling the BPE knowledge base to constantly learn the specific network – with opportunities for predictive and preemptive remediation with much higher accuracy than other AIOps methods.
Enabling closed-loop automation in complex enterprise environments requires precise and fine-tuned remediation, and constant evolution as the network changes over time. Let’s consider a typical WIFI environment, which includes many heterogeneous devices – all of which must work together to enable connectivity.
For example, a typical WIFI system will include a mix of access points (APs), wireless controllers (WLCs), policy managers and Active Directory (AD) for network access control. All of these components must be monitored and then considered in RCA, to provide an accurate diagnosis as well as to enable effective remediation.
BPE’s patent-pending approach uniquely applies ML to identify root-causes for user authentication issues in this multi-layered WIFI network, delivering significant benefits to cost savings reduction, operational efficiency, and improved connectivity service levels.
In selecting the use-cases to examine, we begin in the BPE solution by examining high-impact, common root-cause conditions in the categories of network authentication, configuration settings, performance, device access, and circuit loops. Effectively, this approach takes advantage of Pareto logic which tells us that 80% of impacts will most likely come from 20% of the root causes that exist in the network. By solving high-impact, most-common root-cause problems first, this approach to AIOps enables NOC teams to solve the large majority of most commonly occurring impacts.
This innovative approach will continue to be applied to more of these most common use-cases over time, constantly expanding the practical use of AIOps to solve the next set of key problems while constantly reducing disruptions and optimizing performance for networks under management.
By combining NCCM capabilities along with AIOps-driven monitoring and analysis, BPE delivers a uniquely beneficial set of capabilities that cover the entire Acquire, Aggregate, Analyze, and Act continuum. The impact is easier to appreciate when you recall that 80% of unplanned network outages result from problems with configuration settings. So, by combining NCCM together with AIOps, BPE enables NOC teams to more easily standardize and control network configurations and to avoid misconfigurations from human errors that occur during network change management.
In short, NCCM is a natural fit for AIOps because gaining control of configuration in a network eliminates the first 80% of the most common problems.
This combination is exactly what the BPE solution offers:
First, the NCCM solution – called DCCM – alerts administrators when network device configurations change unexpectedly, such as when someone on-site at a remote location might change settings on a device. This type of visibility and control immediately helps IT teams to avoid the most common root-cause problems known to impact network operations adversely.
Second, by enabling AIOps closed-loop automation – in our INO product – to address root-cause problems that include NCCM changes, the approach enables a simpler, faster network operations that is AI-driven for maximum effectiveness.
I hope that I’ve caught your attention on AIOps – and the innovative approach that Brillio and BP/Ciena are taking, to make networks perform better, to simplify and automate troubleshooting, and to reduce the cost of network operations in complex, large-scale, multi-vendor networks. Stay tuned for my final blog entry on this topic next month.
