5 ways data science can boost network operations

dandelion-2295441_1280By Vivek Kumar

The usage of data science tools and techniques has grown tremendously over the last couple of decades owing to the ever-increasing data deluge. Today, it’s impossible to find any domain of services that are not leveraging the sheer power and might of big data and analytics, or, as you call it, Data Science.

From e-commerce to streaming platforms, healthcare to education, government agencies to non-profits, organizations across domains are trying to make sense of their data. For this very reason, Data Science has evolved from a mere buzzword to being an absolute savior. Organizations striving to stay ahead of the curve are working hard to get the most skilled data scientists on-board. There’s a significant gap between demand and supply of qualified data professionals. Therefore, if you’re a budding data scientist, you should equip yourself with the required knowledge and get a hold of data science certifications to stay ahead of your peers.

Over the years, communications service providers, too, have grown to embrace the beauty of Data Science. Many service providers are rapidly shifting to SDN/NFV technologies, which stands for software-defined networking and network function virtualization respectively, to power their services. These technologies make it possible for organizations to access the network capacity on-demand using a self-service portal. The CDN shuns proprietary hardware for an open, programmable global network infrastructure that can is managed centrally, and the NFV enables features such as acceleration and firewall/proxy to be delivered either from the network or from customer premise equipment – which allows zero-touch provisioning in need of additional functionalities.

As more and more service providers are adopting these technologies, the underlying architecture that binds the network together is becoming more complex and distributed. The operations teams of these service providers have to deal with a system that is utterly dynamic regarding scalability and complexity. In such a dynamic environment, it is always challenging to predict what can go wrong.  

Data science to the rescue!

Data science has the potential to transform the way SP network operations are done, including the reduction of manual efforts required for network monitoring, troubleshooting, and optimization.

Let’s look at how SDN/NFV tools powered by data science can overhaul the operations of Customer Service Providers:

Reduce alert fatigue

The number of components that need to be monitored and managed has increased exponentially ever since SPs have switched to SDN/NFV.  One of the most alarming issues for these service providing organizations is the overwhelming amount of data/information they have from the distributed components in the form of logs and alerts. With so much information and no prioritization and an extremely high false-positive rate, it is impossible for organizations to focus on the critical information. Data science makes it possible to understand the context of these errors and neglect the irrelevant ones – which can lead to a prioritized list of alerts for SP operations team to review and take action.

Improve network visibility, performance, and management control

The introduction of SDN has brought forth many benefits such as network-wide visibility, analytics, and control via a simple dashboard. There’s a centralized controller that determines the best route for the traffic flow of each application. It gauges the real-time congestion levels, link health, the priority of workload to the business, and the quality of service required. This ability to quickly assess route traffic via multiple paths through a network increases redundancy.

Using data science and intelligence at the core as well as the edge of this complex network can come in handy while executing tasks that are susceptible to latency faster, for example – traffic acceleration. This ensures that cloud applications are not only responsive and easy to use but are also helping increase employee productivity and improved customer experience, all the while minimizing network costs.

Enhance security

As surveyed by the publishers of eWeek, security is one of the major attractions of SDN for 45 percent of SPs interviewed.  The centralized SDN controller in the core network controls the end-to-end traffic flows and emerging threats. Using data science and algorithms, these centralized SDN controllers can be trained to adapt to the threat landscape and make decisions when something is malicious and provide reports to the experts. SDNs can be trained to push security updates out to central sites regularly, while a virtual switch can be set up to filter packets at the edge of the networks and redirect malicious traffic to higher layers of security.

This multi-layered approach to security has been made possible only by the use of Data Science. Such granular insights into the traffic and the ability to react in real-time cannot be matched by the traditional hard-wired networks having stringent security policies.

Proactive network optimization

The operations teams of these service providers often struggle in providing the right balance of good performance and high availability. These teams need to identify and resolve the crises in their network proactively.

Using data science, it is possible to quickly process the enormous quantities of monitoring data that these network devices generate, find repeating patterns in the data, and build accurate models of their performance. Anomaly detection techniques can also be used to spot distractions from normal system behavior, that could eventually lead to network failures.

Reduce costs

SDN collates various computing, storing, and processing functions into cheaper commodity servers that reduce the capital expenditure hugely. At the same time, data science and virtualization help in manual network configuration and automating various management tasks – thereby reducing the overall operations costs. Because of this, the need to physically visit branch office sites is reduced drastically.

Most of the giants like Facebook, LinkedIn, Netflix, etc. have already switched to self-healing for some basic operational tasks. Over the years, more and more service providers will move towards “management by exception,” wherein most common errors and performance degradations will be addressed via automated self-healing using Data Science.

Vivek Kumar is President of Consumer Revenue at UpGrad.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.