With 2018 now in full gear, I wanted to take time to discuss some emerging challenges IT infrastructure is now facing as the demands around computing evolve. Today's data center is like a well-running power plant; it is able to meet its specified needs during peak data flow - that is, until it isn't. When factors, such as power, increase demand - like once everyone's air conditioner is powered on during a heatwave - blackouts roll over a region.
But in the mega-data space of data center infrastructure, peak demand is not enough. With Internet of Things (IoT) deployments and other data-heavy applications, that peak is a moving target, and billions in capital are on the line. Blackouts are not an option if a company aims to survive in this highly demanding, and always "on," sector.
So, how does the industry respond as consumer, business, academic, and industrial appetites for data are compounded by the second and as workloads continue to increase into a future that is not yet quantified?
Consider this information: humans generate hundreds of petabytes of data per year through personal activity. Today's data center infrastructure can handle the demand for people's digital lives for now with the storage, agility, and performance capabilities that the cloud offers.
In an increasingly connected world, however, limitations to the cloud model are becoming clearer. These most immediately include the practicality of moving the increasingly enormous data volumes generated by connected devices communicating with each other in the field. The growing scale at which data must be interpreted and acted on in real-time, with computers communicating directly without human intervention, demands that it be managed and analyzed in facilities closer to where it is being generated. Migrating data from IoT devices located throughout the physical environment to traditional, centrally located cloud facilities creates latency issues that slow computing activities. In addition, maintaining static infrastructure to effectively answer the multivalent, inconsistent data flows is cost-prohibitive, if not impossible for any organization outside the largest ISPs.
To further this point, innovation continues to introduce such disruptive amounts of data into the ecosystem that many of today's most cutting-edge architectures can quickly become overwhelmed. The upcoming revolution of self-driving cars is an example of this eventuality, generating mountains of image, telemetry, and diagnostic data. Beyond this, the pace of growth of Industrial IoT will create a "new normal" - generating hundreds of zetabytes per year, several orders of magnitude more than is generated by personal activity. This new normal for handling "peak needs" around zetabyte data management and response will be sooner than expected: in 3-5 years, not 10.
Infrastructure allocation will evolve to triage the management of data on device, at the edge, and in the cloud according to quality of service requirements. Composable Infrastructure is the most flexible and responsive solution to this coming revolution in the industry of data. Composable can respond to workload according to demand, re-prioritizing physical access to data. It can optimize the edge to meet peak demand without being overbuilt, allowing IT administrators to deploy assets where they are needed. Resources can be automated to accommodate data spikes when they occur then directed toward other applications when not in use. In addition, with composable GPU -- a Liqid breakthrough with far-reaching use cases -- the most data-intensive applications using artificial intelligence to interpret these data can now utilize this once-static data center element at scale to accelerate applications in deep learning, analytics, and engineering.