- Artificial intelligence (AI) and machine learning (ML) are being developed for networking and network management, with the first iterations providing deep analytics and augmenting IT’s root cause analysis workflows.
- There are a number of gaps that need to be closed before the full capability of AI and ML systems is realized.
Artificial intelligence and machine learning techniques continue to be improved as companies and researchers develop and test products. In networking, established vendors and startups are developing management systems that promise to automate the time-consuming, low-value work of data collection and correlation as well as the AI and ML techniques to provide actionable information to network IT, such as predictive trouble alerts and recommendations to resolve problems. Vendors’ products such as the recently announced Cisco DNA Center Assurance, Cisco Network Analytics Engine, and Mist Systems Virtual Network Assistant are also using sophisticated UI elements to help IT better understand networking problems and assist with root cause analysis. We’ve heard these promises in the past, and for one reason or another, these advanced techniques failed to deliver. Before you go the AI route, here are some key questions to ask.
- With what products do you work? I think the value of any management and analytics system drops precipitously as more product gaps are found, because any model will be incomplete, equating to a blind spot. If there are products in your network that are not part of the model, then you or your vendor will have to determine how to cover the gaps.
- How long does it take to learn new products? While having 100% product coverage at the start is best – and in time, as more products are integrated, that likelihood increases – chances are there will be product gaps. So, what does it take to have new products added? What is that process and what are any associated costs? Unless you have some niche networking product, vendors have an incentive to build integration capabilities only once and make them available to everyone.
- What can you do today? Just like any employee, understand what AI can do now. AI development is very much ‘crawl, walk, run.’ In the case of the products mentioned above, and from other vendors such as Apstra and Forward Networks, there is some AI automated analysis at work and some tooling that aids IT in its own analysis, such as contextually relevant breadcrumb trails, a collection of recent events, and workflow funnels from general to specific cases, as well as specific to general cases. When these products work, they work well.
- Where do you see yourself in five years? The dreaded interview question no one wants to answer, and perhaps five years is too far out, but it’s important to understand the vendor’s roadmap and product direction. Integration with third-party products is table stakes, but understanding how the AI system will integrate with existing management products and how it will impact your management strategy and workflows allows for planning. Knowing how future analytic research will align with your organization’s needs lets you better select a product that matches your future requirements.
AI and ML are coming to IT and network management with useful and practical outcomes. I see these technologies only growing more sophisticated over time as vendors pour more intelligence into their analytic engines. The current crop of products is focused on specific use cases in order to be immediately useful, and those use cases will grow over time.
I welcome our robot overlords, but be careful if they ask you to play a game.