Where is the “network” in AIOps? No where in most AIOps discussions. Case in point this page summarizing AIOps descriptions. What quickly becomes apparent is the term emerged from a set of problems in compute/applications software architectures: micro services, cloud compute, etc. These shifts “broke” traditional IT operations, requiring a new approach.
The figure below depicts two columns, common AIOps capabilities and areas that differentiate AIOps implementations. A great deal of industry discussion focuses on the first column, the new technology approaches. While there can be great differentiation in the first column, what defines the class a product/service is in, comes equally from the second column: what is being modeled, who is the consumer, what kinds of anomalies are being detected, domain-specific algorithms, specific data types, and more.
Figure 1. Two ways to think about AIOps, the technology and the implementation focus.
If you compare ITOps AIOps solutions, to Augtera’s Network AIOps solutions, through the lens of the first column, it may seem like there are a number of similarities, especially without an in-depth technology discussion. If you compare the average ITOps AIOps solution to Network AIOps through the lens of the second column, significant differences appear immediately.
It may be practical to label product categories by a central technological approach, for example AI/ML, however, operations teams do not buy technology, they buy answers to their pain points. As a result, an Application & Compute AIOps product/service is very different than a Network AIOps product/service.
Up until recently, there has been no “network” in AIOps. In designing, developing, and deploying the first AI purpose built for networking, Augtera Networks has changed that.
Where is the “network” in AIOps? To provide just a few examples, ask the following questions:
- Which AIOps solutions model all layers of the network from the physical layer to the TCP layer and above?
- Which AIOps solutions detect optical signal degradation in time to replace the optics or automatically move traffic somewhere else before a failure occurs?
- Which AIOps solutions understand BGP state machines and how that impacts anomaly detection and incident mitigation?
- Which AIOps solutions assess the impact of data center fabric congestion on application experience?
Many similar use-case centric questions could be asked. The point is clear. Networking is different, and it requires a purpose-built approach.
