Manage user performance, not the network, with ML-based tools
16 November 2017 | 0
Over the past decade, network management tools have evolved from being fault based to performance based. This has become a critical element in running infrastructure because faults do not matter as much.
That might seem like a strange thing to say, but consider the fact that critical infrastructure such as switches, routers, Wi-Fi access points and servers are deployed in a way to protect against outages. Infrastructure is built so redundantly today that any hardware device can go down and its likely no one will notice.
A bigger problem is managing user performance. Often users call in about a certain application not working well, but when the engineer looks at the dashboard, everything is green. Performance problems are much harder to diagnose and can kill employee productivity.
Earlier this year, research looking at the impact of poor application performance found that workers are 14% less productive because of application issues. Think about that statistic. Businesses invest billions of dollars every year in technologies to make workers more productive, but if they could just get all the applications they already run to work optimally, employee productivity would increase 10% or more.
That is why IT professionals need to shift the focus of management from the network to the user. The problem with network management is that by definition, it provides a bottom-up view of the world where user experience is inferred. User performance management (UPM) is more top down where the underlying technology and dependencies are understood, so if a Wi-Fi AP gets overloaded, the network operations team can quickly understand which applications and users will be impact.
It is important to note that UPM cannot be achieved by taking a fault management tool and retrofitting or putting a new dashboard on it. Instead, UPM is achieved through a combination of data, machine learning and the cloud for scale.
One vendor trying to make UPM a reality is a start-up called Nyansa. The company has announced some updates to its Voyance UPM product that shift it from being a self-contained product to a machine learning platform that provides active remediation advice and can share data from third parties, extending Voyance’s value.
The new version now pulls in SYSLOG information from Cisco’s Identity Services Engine (ISE), HPE Aruba’s ClearPass and Free RADIUS. This is an extra data source that provides information into DHCP and DNS issues that can reveal user issues that indicate connectivity problems. This tends to the biggest source of problems on Wi-Fi networks, and Nyansa has about 70% of the market covered with just Aruba and Cisco.
The SYSLOG integration will be important as Internet of Things (IoT) deployments become more common. Voyance’s machine learning algorithms can take the SYSLOG data and alter network operations when an IoT device performs at a level that is business impairing.
Nyansa also added externally facing APIs to extend the platform. Data from Nyansa can now be exported to IT workflow products such as ServiceNow. Voyance could spot a performance issue and then actively open a trouble ticket, enabling IT to get a handle on the problem before users report it.
Alternatively, Voyance could send data directly to a communication platform like Slack where data from specific endpoints, applications or users could be sent to a particular group of people. For example, problems reported with a customer-facing application could send information directly to a Slack Room that also includes the experts responsible for the application. This could significantly shorten the “resolution ping pong” that occurs as trouble tickets get passed back and forth before the problem actually gets solved.
Another interesting feature added to Voyance is the ability to easily tag a device or person. The tagging enables engineers to constantly monitor the experience of critical assets or even people. As Animal Farm taught us we are all equal, but some people are more equal than others, now IT can know when someone who is more equal, such as C-level executives or high-performing salespeople, experience problems. This can also be used to monitor things such as factory floor equipment, heart pumps or other mission-critical IoT devices. This data can also be searched on, filtered or exported for additional analysis.
Predictive not reactive
Nyansa uses machine learning to enable IT to move from a reactive model to a predictive one. Voyance now includes a remediation engine that provides IT with a view into where and how network incidents are happening and the impact on user experience. It also recommends specific actions to fix those problems—things IT may not even know are causing problems. For example, the remediation engine could suggest the 2.4 GHz radios on APs be turned off because they cause Wi-Fi problems.
One particularly valuable feature is that the tool now shows how many lost client hours will be recovered or lost by each action taken or not taken. This can help IT operations prioritises its efforts. Think of this as Splunk on steroids where Splunk shows lots of data but the insights and actions need to be determined by smart engineers who can correlate the data. The machine learning in Voyance does that heavy lifting, so IT can figure out what to fix faster and in an order that is most meaningful to the business.
Managing user performance is not done by measuring one particular aspect of the network. It’s more about understanding all the IT elements that make up s service and the relationship between them—and then having the insight to understand how that impacts productivity. With the volume of data available today, this cannot be done manually. It is important for IT pros to start using machine learning-based tools like Nyansa to solve those tough-to-fix performance problems that kill worker productivity.
IDG News Service