Preparing the Network for AI and Machine Learning

The integration of artificial intelligence (AI) and machine learning capabilities are poised to drastically change the modern business landscape. A recent survey of Fortune 500^® CEOs showed that in 2017, 81 percent believe these technologies will have a positive impact on their organization, up from just 54 percent in 2016. Furthermore, the Forrester^® recent technology predictions report forecasts that “2017 will be the year the big data floodgates open, driven by a voracious appetite for deeper contextual insights that drive customer engagement via mobile, wearables, and [the Internet of Things (IoT)].”

The possibilities for impactful business transformation are endless. More and more organizations are realizing that potential competitive advantages are locked in their customer and business data, and are beginning to invest in AI and machine learning to unlock that value. But while AI and machine learning offer infinite possibilities to business leaders, they present new challenges for network management that you should begin preparing for.

AI and Machine Learning Fundamentals

The benefits of these analytics and predictive capabilities are certainly far reaching. A retailer, for example, can collate and process customer data to pinpoint noteworthy trends, such as fashion interests for next season, and use that insight to inform their inventory decisions. Other organizations can leverage business data to drive data-informed project management, allowing business leaders to more accurately determine how long certain operations may take and will cost.

But these capabilities also put added responsibilities on network administrators’ shoulders, starting with the need to increase familiarity with various methods of algorithm scripting and mathematics in order to achieve the business outcome the organization desires. Are you implementing a neural network, supervised, or unsupervised algorithms? What type of task do you need the algorithm to execute? Decision trees? Classifications? Cluster analysis? Pattern recognition? In most cases, your organization will be leveraging a programming specialist to do this, but it falls within your purview as the network administrator to have a fundamental understanding of what these processes are so you can prepare your network and security processes accordingly.

There are a number of different outcomes that can result from each unique algorithm and data process, so you should be prepared to follow six fundamental steps in order to appropriately define your problem and map to a specific formula.

Thankfully, technologies like software-defined networking (SDN) and IoT have, at least to some extent, already given IT professionals a taste for how machine learning deployments will impact the network and management best practices. With SDN, you create a template and associated behaviors and the software-actuated network implements changes as necessary. As an offshoot of this technology, machine learning takes one step beyond the fundamentals of SDN by looking for action X (perhaps that priority network traffic is getting clogged) in order to auto-remediate by doing Y (adjusting bandwidth resources).

Implementation Considerations

For all their benefits, AI and machine learning capabilities do require paying attention to certain considerations, the first being cost. Leveraging these technologies is not cheap. The fundamentals of these technologies are rooted in data-driven algorithms that enable machines to develop learned responses or predictive capabilities. As a result, with AI and machine learning comes data—big data—that requires resources to be allocated, not only specialists like programmers, but additional on-premises resources such as storage, server CPUs, networking bandwidth, and cloud-hosted storage services.

At the same time, while it’s tempting to rapidly adopt a new technology in anticipation of its transformative benefits, it’s critical you and your organization set realistic deployment expectations—it may be a bit ambitious to assume an AI or machine learning deployment will get off the ground and generate actionable insights within three months. If not careful, you may actually hurt your business rather than help it. To most efficiently and accurately realize the benefits, your organization must set expectations up front, run through the various deployment preparation steps, know the problem you want to solve, and ensure you have the data to solve it.

Preparing Your Network: Best Practices

Although deployments take time and not every organization is ready to leverage this technology right now, IT departments must begin preparing to support these new trends today to ensure that not only is the anticipated ROI achieved, but that your networks can support these efforts long term. Here are several best practices that will help you begin to prepare your networks for AI and machine learning:

Know the problem you want to solve

What kind of machine learning or artificial intelligence capabilities does your organization need? Before deploying these capabilities, your organization should follow six steps to help you define and begin to solve your problem:

1. Classify the problem you want to solve. What business outcome do you need it to address?
2. Acquire the data needed to achieve a solution. Does it exist?
3. Process the data. How are you going to cleanse it and prepare it for various algorithms? Will you be running supervised or unsupervised algorithms?
4. Model the problem. What algorithms will your business be using? There are different supporting suites, but you need to align to the problem you’re trying to solve.
5. Execute and validate the results. Do you need to refine the process at all?
6. Deploy the updated algorithm and process to generate real business insights.

Educate yourself

Just like how you prepared for the shift to the cloud, you need to start skilling up for the onset of AI and machine learning. You must have an appreciation for the art of algorithm development and machine learning implementations in order to successfully prepare your network for these workloads. Much of the fundamental math behind machine learning deployments is often left up to a programmer or data scientist, but you should supplement that with additional research into the mathematics behind these algorithms and research into the specific type of machine learning you’ll be using. This insight will help you, as a network administrator, better assist in the development of strategies for the added demands on your networking resources.

Establish performance baselines

This should be the network administrator’s main task when preparing a network for AI and machine learning. Can you support the added bandwidth needed for the data traffic? If not, what do you need to do to enable your company to eventually support these processes? By leveraging a comprehensive monitoring and management tool, you can establish performance and resource utilization baselines that will allow you to make informed decisions about how your network will be impacted by the introduction of AI and machine learning big data. Such tools should ideally also include the ability to monitor and maintain visibility from your on-premises data center to the cloud, where this data is often stored, to ensure you’re aware of any potential vulnerabilities or performance bottlenecks.

Create a solid security foundation

This certainly isn’t a best practice unique to AI and machine learning, but it bears repeating for every data-based technology or capability you introduce into your environment. You must make sure that only the people who truly need access to the information get access to that data. You should look to establish a comprehensive compliance policy that includes real-time change notification and an approval process that will help avoid covert data pulling and breach or sabotage.

Final Thoughts

As businesses look to develop their digital transformation strategies and create unique competitive advantage, AI and machine learning are increasingly considered the keys to unlocking the value of an organization’s accumulated data. The C-suite and IT leaders will continue to push to make this accumulated data a value-added reality inside your organization, but it’s incumbent upon you to implement and manage it responsibly. These new capabilities require new approaches to management and monitoring that you and your organization can prepare for by leveraging the above best practices.

About the Author

Destiny Bertucci is a Head Geek at SolarWinds^®, and is a Cisco^® Certified Network Associate (CCNA), CIW Masters, INFOSEC, MCITP SQL and SolarWinds Certified Professional. Her 15 years of network management experience spans healthcare and application engineering, including over nine years as SolarWinds Senior Application Engineer.

Sign up for the free insideBIGDATA newsletter.