Don’t let anyone tell you hardware doesn’t matter. Hardware is everywhere. In every mobile phone. Every Fitbit and techno-gadget we own. In our cars. In our laptops and tablets. Increasingly, it’s in our appliances. In our watches. And apparently, in some cases, in us. It’s smaller, faster, and prolific.
Apps, no matter how big or small, need hardware resources. CPU. Memory. Storage. Network. Without these, none of them work. None. Whether we’re talking in the data center or in a cloud (data center), hardware is omnipresent. It provides the horsepower necessary for apps to make our lives more connected, more collaborative, and more convenient.
But that doesn’t mean that the data center isn’t changing with respect to hardware. It is. While hardware is still the foundation upon which we build every technology, the frameworks and technical houses (the apps) we build atop it have changed. In some cases, quite dramatically. Moore’s Law is no longer a constraint when you can pool resources and deconstruct applications into multiple services, using the underlying mesh of compute and memory like a massively distributed system itself.
We see these data center changes in terms of generations based on the model of how we provision, manage, and consume those resources.
In the longest-lived generation of application infrastructure – generation zero – we booted machines and started applications and pounded our chests proudly about sequential years of uptime. Then came virtualization, and we booted machines and resumed apps running virtual machines. Uptime of a single system became passé; availability of an app was the metric du jour. In the cloud, we launch services with the click of a button and scale up and down without ever lifting a finger. And in the post-cloud generation, containers clone themselves automagically with a speed that would make pit crews at Daytona green with envy*.
The time between app infrastructure generations is compressing. The time between generation zero and generation one spans decades. Virtualization has been around a long time, but it didn’t “come of age” until nearly 2010. Cloud? 2015. Containers? They’re moving quickly and it’s not even 2018. Each generation advances on the lessons learned from the previous generation, noting its failings and disadvantages while trying to improve upon them. At the heart of these improvements lies speed and scale. Speed to market and of delivery. Scale of business and of the apps upon which it now relies. Both have contributed to the evolutionary changes impacting every facet of the data center.
All those apps and services still need hardware. Resources don’t appear magically, even if they seem to. That was part of the reason virtualization and cloud grew at the pace it did. Because it seemed magical. But magic is, after all, just an illusion. In the case of computing, that means the hardware (and the resources it provides) is still there. What’s changed is how we provision, manage, and consume it. And those changes are having a profound impact in every rack in the data center.
And it’s causing a generational gap that requires NetOps to close.
The scale and speed with which organizations are now being pressured to deliver new applications and APIs introduces the need for automation – whether via cloud or containers, on-premises and off. The shortened life expectancy of the use of hardware resources by apps and app services means recycling at a phenomenal rate. Manual processes and lengthy boot times cannot keep up with the frenetic pace set by such systems. Which means containerized and cloud (or at least cloud-like) frameworks that enable the fast, frequent service creation and destruction necessary to scale at speed.
Automation, consumption of resources, and service creation speed delineate the data center generational lines. Even moving rom generation 2 (cloud) to generation 3 (containers) drastically changes these characteristics. Each generation moves the needle further toward full automated sub-second responsiveness thanks largely to resource consumption profiles that last minutes instead of months or years.
That’s why NetOps is (or should be ) such an important focus right now. Because it is the network – the traditionally hardware network – that is impacted most by expectations arising from generation two and three app infrastructure. IT is often tasked with supporting application infrastructure spanning all four generations which means costs and infrastructure in the network must often be shared. That often means purpose-built appliances. But even if an app service is hosted on a traditional appliance, it still needs to provide the means to be consumed by a generation two or three app infrastructure model.
Self-service and automation must be used to meet the demands of the latest generation while maintaining the reliability required to support the first generation. That means integration with ticketing systems and exposing analytics via APIs and dashboards. It means that no matter what the underlying hardware – purpose-built or COTS - the network must present itself as a consumable, automatable resource that responds with the same alacrity as generation two or three application infrastructure.
In some cases, that means network and app services must adopt the same style of infrastructure as the applications it delivers. App services like load balancing and web application security are increasingly being embraced by developers and baked into app architectures and infrastructure. Others – particularly in the realm of security – simply don’t work as well (or at all) in generation two (cloud) environments. That means those services must become cloud-consumable, containerized, and easily accessible.
The existential generational gaps inside data centers can be problematic for organizations. They can inhibit or outright stall digital transformation efforts. They definitely increase friction between development and IT, between IT and the business, between business and its customers. That friction can be reduced by NetOps efforts to meet the consumption models required to support generation 2 and 3 infrastructure models with APIs, self-service, and automation.
The complexity and sometimes competing requirements that exists in mature data centers between applications in different generations means it is up to NetOps to rise to the challenge. The network (and all that entails) must transform if it is going to succeed in simultaneously supporting four generations of app infrastructure.
*NASCAR pit crews aim to complete a pit stop (changing all four tires) inside 12 seconds. The fastest recorded pit stop was a mere 8 seconds. The speed with which a crew works is key to a driver’s success.