As the network management industry continues to evolve and integrate technologies like data analysis, artificial intelligence (AI) and machine learning (ML) into everyday operations, there’s a growing debate about whether it’s better for network management systems to be proactive or predictive. As with any debate, it helps to better understand each side before making a decision.
Predictive Network Technology
Put simply, predictive network technology employs big data techniques on large amounts of historical data to anticipate network problems. Predictive systems generally require many months of data to be stored and analyzed before any predictive benefits can be observed.
Predictive network technology is decidedly not a real-time AI methodology. It’s instead based on network analytics and is trend-driven. Examples include:
- Detecting normal expected peaks and valleys for different classes of applications
- Detecting when large, abnormal networking uses kick in (extensive security scans, backups, crypto mining, BitTorrents, etc.)
- Detecting unusual and unexpected traffic patterns
Generally, predictive network technology can identify when things are “abnormal” or “unexpected.” In certain cases, actions can be taken automatically (for example, adding capacity, removing capacity, or blocking traffic) in response to what has been identified.
That said, predictive network technology nearly always requires human oversight and does not focus on the actual day-to-day network operational issues that a team member faces. Its strength is when it’s used to empower humans to make better operational decisions about when to implement network changes.
Proactive Network Technology
In contrast, proactive networking technology looks deeply into real-time network events and measures, in real time, what users are currently experiencing on those networks.
Big data-based network analytics has always been considered the “holy grail” by those in the network analytics business. However, while it’s clear that interesting observations can be made with large amounts of data-driven insights, it’s a proactive approach that enables organizations and technicians to operate exact cause-and-effect scenarios and actions.
True proactive AI/ML – that incorporates a user experience point of view and creates actions in real time – is significantly more valuable than data lakes of network analytics. In fact, understanding network capacities and events over large time frames can even create predictive models. For example, an organization may learn exactly how the network is used after the lunch break Monday through Friday, and then can take proactive actions to make sure the network is prepared for the situation.
Juniper Networks’ Mist AI technology can correlate these real-time events and experiences across an organization or all of a service provider’s customers, empowering them to make clear and decisive operational decisions, eventually leading to the automation of specific operational actions.
So, Which is Better?
Predictive networks require large amounts of data, big data processing and expert interpretations of the results to be effective. These are its biggest challenges. Transporting the data to the big data processing location can also be a complex and expensive task. In addition, site and customer-specific big data results then also require interpretation by trained humans.
Due to the very large amount of data required to establish a predictive approach, and the expensive big data skills required, the predictive approach is not a good choice for businesses that are spread out geographically (getting the data to the big data locations is expensive and hard), or that lack the ability to understand and create action plans based on the predictive network data being obtained.
Generally, predictive network solutions are good only for large organizations with dense network deployments (carriers, multinational businesses with large facilities).
Proactive networking technology solutions, on the other hand, provide valuable insight from day one and do not require massive data collection in order to get started. Proactive technologies are more efficient and cost-effective for organizations to deploy and utilize as well.
Focusing on the End-User Experience
Focusing on how human users experience network connectivity is the only approach that will work for today’s networks. Network operators need to focus solely on their customers’ experiences, and they need all of the AI/ML or big data products they employ to focus on this.
Focusing on large data trends over large periods of time, as is seen with predictive networking approaches, creates inexact anomaly detection and is difficult, if not impossible, to tie back to people’s experiences on the network.
A proper AI/ML solution must collect application performance data like Transmission Control Protocol (TCP) retransmissions and time-to-first byte for Transport Layer Security (TLS) sessions. These measure how humans “feel” when using the network. No other routers, switches or general network gear collects this information for real network traffic. Application performance monitoring solutions only measure this type of data using active network testing, as opposed to measuring what real network users are experiencing.
Juniper Can Help Drive Proactivity in the Network
Juniper is the only in the industry that truly measures how users experience their network, and then correlates that experience to actual network events. Technology such as Marvis, Juniper’s Virtual Network Assistant (VNA), is a massive improvement over the status quo and can help transform an organization or a service provider into one that puts proactivity first.
Juniper’s solutions work on widely distributed networks, as well as over WANs. The company’s proactive technology has built-in, tested and proven actions that lower trouble tickets and solve problems in an automated fashion. Proactively anticipating network issues or outages helps organizations and service providers to improve discovery time, lower mean time to repair, lower costs and improve end-user satisfaction. Some network issues will even be identified and repaired – proactively – before the end-user knew they were there.
With the power of Juniper Networks’ proactive network technologies, IT teams no longer need to rely on inefficient and ineffective predictive technologies, and can easily correlate and create actions that will automate their network management. To learn more, join one of our weekly demos here.