dcsimg
 

Poor Visibility: The Speedbump on the Road to Automated Networks

Thursday Oct 18th 2018 by Arthur Cole

As they deploy automated — and, eventually, intelligent — network technology, enterprises need better monitoring tools.

Automation is probably the leading technology initiative in networking circles these days. Whether it is through software definition or advanced, even intelligent, management systems, the goal is to automate network provisioning and operation to better match the service-driven data environments of the emerging digital economy.

But automation, by nature, is highly data dependent. Regardless of whether a human operator is programming the system to perform rote functions or the system itself is becoming increasingly autonomous, the need to gather data and interpret it quickly is paramount. Many networking pros are calling for increased visibility into disparate architectures, because the one sure way that automation can foul things up, over and over again, is to give it an incomplete picture of what is happening out there.

Network analytics firm Kentik’s latest State of Network Management report, in fact, lists visibility as one of the key enabling technologies for advanced automation, particularly now that most network infrastructure these days extends to third-party clouds. While there has been a large proliferation of tools aimed at network visibility, this can be a curse as well as a blessing. Many organizations are still trying to devise the magic formula that will give them the ability to trust their connectivity to automation, but this may take a while. The leading approach, network traffic analytics, is only employed by about 28 percent of network managers.

Strong visibility will become increasingly important as the enterprise evolves from simple automation to full-blown intelligence, says Savvius’ Jay Botelho. Algorithms driving AI and machine learning are only as good as the data they are given, and unfortunately most flow-centric approaches lack the detail to fully support an intelligent management stack. Tracking speed and pathway metrics are fine for trouble-shooting basic problems but more complex issues require deeper examinations into network packets and other elements. Automation, after all, isn’t just focused on repairing bottlenecks and anomalies but on continuously optimizing resource consumption and overall operational efficiency.

Even some of the largest networking vendors have come to realize they cannot deliver this level of visibility without help. Juniper recently expanded its partnership with Nutanix to improve multi-cloud architectures by improving automation, security and other tools. Part of the effort involves integrating the Contrail Enterprise Multicloud platform with Nutanix APIs to enhance visibility into virtualized workloads. In this way, the companies hope to support the development of fully automated fabric management solutions.

Gaining visibility into cloud infrastructure becomes even more challenging once you connect it to the Internet of Things (IoT), says Solarwind’s Patrick Hubbard. Once data has moved beyond the firewall, it becomes very difficult to see where it goes. At this point, visibility needs to extend beyond mere hardware health and performance to encompass the inner workings of cloud infrastructure and the cloud-facing components of individual applications. A key asset in this transition is a single interface for the entire monitoring stack, whether the target is the status of VPN tunnels between sites or the health and readiness of the firewall.

The fundamental question that network managers face when contemplating visibility challenges is, how much visibility is enough? If experience is any guide, it’s not the things you can see that cause the biggest problems, but the things you can’t. So no matter how deep or how wide your visibility solution extends, won’t there always be some tidbit of data somewhere out there that can bring your network to its knees?

More than likely, yes. Neither visibility nor automation is foolproof.

But this much is certain: an automated network requires much better visibility than exists today. It only makes sense to plan for the known requirements today, and then hope the unknowns will reveal themselves before they can do too much damage.

Arthur Cole is a freelance journalist with more than 25 years’ experience covering enterprise IT, telecommunications and other high-tech industries.

Home
Mobile Site | Full Site