• Government network automation paves the way for artificial intelligence and machine learning. Credit: Shutterstock
     Government network automation paves the way for artificial intelligence and machine learning. Credit: Shutterstock

Network Automation Brings Federal Agencies Closer to AI

The Cyber Edge
July 3, 2018
By Bob Nilsson

Automation enhances network efficiency.

It has become increasingly evident that artificial intelligence (AI) and machine learning (ML) are poised to impact government technology. Just last year, the General Services Administration launched programs to enable federal adoption of AI, and the White House encouraged federal agencies to explore all of the possibilities AI could offer. The benefits are substantial, but before the federal government can fully take advantage of advancements like AI, federal agencies must prepare their IT infrastructure to securely handle the additional bandwidth. At the top of the list for the federal government to prepare for AI technology is the network.

Network automation introduces two essential elements: operational efficiency and a programmable interface into the network. Both have significant impacts on IT maintenance spending, a challenge of particular importance to federal agencies. Internet of Things and cloud-focused digital transformation are pushing the limits of today’s networks, and with so many unique data sets, automation could be the difference between network outages and network connectivity. With that, the federal government should look to automated network solutions that will give it greater visibility into usage and also provide the infrastructure for future technologies like machine learning. Below are examples of how this looks in practice.

Increased Visibility

As the volume, velocity and variety of data in the network expand, comprehensive visibility into the network becomes critical. Network visibility enables agencies to quickly identify problems, accelerate mean-time-to-remediation and improve overall service levels. Additionally, without pervasive visibility, companies eventually run the risk of automating into trouble by taking action using insufficient or inaccurate information and expending already constrained resources.

To better illustrate how agencies can benefit from this technology, consider a distributed application in a data center where an operator gets a call from an end user who shares that an application isn’t running or is slow.

The network operator would log the ticket and determine the origin of the problem, whether it is the application, insufficient or overloaded compute resources, a misconfigured overlay or a faulty network connection. Today, the operator must gather data from different places in the network with multiple tools and process that data before taking action, which relies mostly on manual effort to complete the process.

Alternatively, a federal agency using an automated tool that offers the right level of information at the right time and location enables action on an individual switch or router, and can send the data to an external tool. This type of visibility into common issues and processes ensures that automation is tailored to common agency events and is applied to functions that are the most cost-effective. Automation should not be handled using a one-size-fits-all approach. Instead, network administrators should equip their networks with the appropriate tools that ensure both automation and visibility at the appropriate level.

Machine Learning Capabilities

Once the proper networking infrastructure is in place, federal agencies can leverage machine learning technology to further adapt to changes in technology. Through ML, IT has the ability to keep a record of and recognize different network events, including failure, congestion, security anomalies and other network problems. From there, teams can create models to forecast where to apply resources or take other actions.

Fundamentally, ML methods are an inversion of the traditional computer programming paradigm. Classical programming approaches are based on the premise that programs are developed to produce data.  ML reverses that relationship—data now produce programs.  The accuracy and reliability of ML is completely dependent on the data that it trains or learns from.

Because of this, one of the key drivers for successful ML is data acquisition. Pervasive visibility allows for the necessary data collection that is foundational to enabling ML. Subsequently, ML can find those unique patterns in your data and classify it in a meaningful way to unlock your ability to make inferences for better decisions. The result ensures more informed decisions based on your data.

Another key driver is a programmable network. A programmable network completes the life cycle of decision making at software speeds. For this to occur, the network must have automation. Without a structured interface to the network, managing the final stage of the life cycle can be excessively burdensome, resulting in increased effort instead of a reduction in effort. ML can engage the network through that structured framework, which continues decision making at the speed of software. This speed is mandatory to maintain operational efficiency and to combat threats that are operating under ML control.

Federal agencies can take advantage of network automation technologies today, without hesitation. By identifying strategic areas in the network where automation and visibility can be injected, agencies can begin cutting costs and creating opportunities to implement administrative efficiencies as they work to meet their mission. It is critical that government invests in network-enhancing solutions today, so it can realize the benefits predicted for AI tomorrow.

Bob Nilsson leads the Extreme Networks strategy and programs for vertical markets, including healthcare, higher education, K-12 education, federal government and hospitality.

Enjoyed this article? SUBSCRIBE NOW to keep the content flowing.


Share Your Thoughts:

I am a fan and long time follower of AI and ML. As a cyber security person, the line in this article "The accuracy and reliability of ML is completely dependent on the data that it trains or learns from" gave me pause to think about the integrity of ML programs. How to prevent being fooled by an attack that intentional poisons data? This is certainly a factor that requires thought to ensure the effectiveness of generated programs. I understand we are on the cusp regarding AI and ML and am interested to understand how the DoD is approaching the risk factors and potential threats in these early stages.

Share Your Thoughts: