Partner POV | Pioneering a user-centered approach to AI-powered network operations

This article was written and contributed by our partner, HPE Aruba.

With the integration of diverse technologies, proliferation of IoT devices, and rise of data-intensive applications, network operations have become highly complex. And, incorporating the additional challenges of security and extensibility presents a strain on already limited resources.

Let's consider an example. Many universities are trying to augment the academic experience with a modern, in-campus student experience. One interesting use case is to provide personalized network services that enable students to use and selectively share gaming devices in their dorms. Consider how this drastically alters the access control, security, and performance requirements of the dorm networks, and increases the network complexity for an over-burdened network operations team.

Meet new HPE Aruba Networking Central

Announced at Atmosphere '24, new Central maximizes the efficiency of Day 0 to Day N network operations with an intuitive, operator-centric experience that is designed to harness the value of data and take advantage of deep AI capabilities.

With new Central, we democratize access, interpretation, and understanding of network data across diverse operator profiles and skill levels with the power of human-centered design.

The design foundation makes it easier for domain experts and L1/L2 engineers to troubleshoot complex issues with diverse datasets and also configure and maintain complex networks. Let's examine three primary recurring customer pain points to illustrate how we address them.

Manual correlations and inefficiencies

Network issues typically originate as innocuous "Wi-Fi is not working" or "my internet is slow" tickets. First-line support then begins with questions on:

• Was it specific to a student or did it impact others?
• Was it a WLAN, LAN, or WAN issue? Was it on their private network or university network?
• Was it a critical business application?
• Was it due to a security incident?

These are complex datasets spread across multiple screens which can overwhelm the operator and slow down analysis, increasing human error. To tackle this, data must be correlated, interpreted, and presented using a dynamic, multi-dimensional framework – which doesn't exist today.

The design foundation for new Central addresses this through our multi-dimensional solar-system view. The focus of the investigation (the sun) could be the Site/ Network / Device / Client / Application. All relational elements that can determine the outcome of the investigation are depicted as planets around the sun (see Figure 1).

These include:

• Networks (WLAN, LAN, WAN, events)
• Clients (Users and devices - metrics on access, usage, classifications, behavior)
• Applications (Institutional and private – metrics on resource consumption, performance scores, user/device relational pivots)
• Security (Authentication and authorization policy, IDS/IPS, Security & QoS Policy conformance, behavior anomaly, threat scores)
• Alerts (AI assisted issue diagnosis on connectivity, performance, configuration, and health)

Each planet also has a composite health score that quantifies the specific attribute, offering a holistic view at every level of investigation. This design contextualizes massive amounts of data and offers unparalleled flexibility for operators to analyze the network at varying levels of detail.

"...when you can demonstrate an interface like this that's intuitive, saves time and it makes it easier to do your job, those are all things that resonate with customers and users," said Joel Grace, senior vice president of IT infrastructure and cloud practice for Sayers, a Vernon Hills, Ill.-based solution provider. The updated interface is less about dashboards and more geared toward the end-user perspective that fits well into "the day in the life of a network administrator," Grace said.

Rising MTTRs and support costs

Another challenge faced by network operators is troubleshooting intermittent issues that occur once and are hard to solve, driving up the mean time to resolution (MTTR). Until now, solutions in the market did not have adaquate toolsets to troubleshoot transient issues on the network.

Typical troubleshooting would involve pouring over voluminous SIEM logs, application logs, network diagnostic commands, triggering a dynamic packet capture, searching through alerts to triangulate a possible event in past and even jumping into the CLI for deeper analysis. Even AI-driven alerts do not capture every intermittent issue if a training pattern hasn't yet been developed to detect those issues.

New Central introduces a pioneering concept called network time travel which allows operators to move back in time to troubleshoot. Think of it like a favorite 1980s movie, The Terminator, where rather than hunting John Connor, you are trying to find out where/what/how or when a network issue occurred.

The feature builds a series of snapshots with complex datasets that records events, states, configs, logs, reports, and alerts. A network operator can mouse over the period of time, much like hitting a rewind button, to piece together information and determine the root-cause of the issue (see Figure 2). We believe this has game-changing potential to define the future of network troubleshooting techniques.

In addition, AI-powered insights for client experience and device health assist operators in proactively managing the network. Not only does new Central rapidly analyze the root-cause but it also comes up with specific impact and recommendations for remediations reducing time-to-fix by steering the operator through a noisy heap of event triggers and reducing alert fatigue.

Monitoring the operational state of the network

Another common pain point from customers is to visualize the operational state of the entire network in a single view. For example, a network administrator deploys a new site and wants to make sure that the network is working as expected. 

Using the innovative sunburst view (which provides a bird's eye view of the network visualizing physical and logical connectivity with layers of network performance), attributes are interspersed in a single transformational snapshot. This form of interactive visualization is attested to be an ideal design model for the presentation of large, complex, and hierarchical data.

The sunburst view scales for networks of any size and allows engineers to have the most efficient operational graph on which attributes such as segments, link-connections, health, location, presence, heat-maps can be represented as dynamic layers (see Figure 3).
 

Much like Google Maps, the operator can quickly navigate by searching, zooming, pivoting, and applying multiple layers of visualization to network elements. 

Innovations in HPE Aruba Networking Central are happening every day. Stay tuned for more insights and details.