With each new year, the automotive industry unveils another feature that leads us closer to an age of self-driving cars. New models already tell us when to change the oil, when to inflate our tires, when to stop, how to parallel park and how to stay in our lanes. And sooner than we realize, it’ll be second nature just to let the car take the wheel and drive us to our destinations.
By several estimates, more than 10 million self-driving cars will be on the road by 2020. Although some may dispute those predictions, few disagree that self-driving cars are on their way. But if we can have self-driving cars where artificial intelligence does most of the work, why not self-driving data centers where automation and machine learning handle administrative storage tasks?
Even for those who don’t believe machines can execute IT-manager tasks more effectively than humans, the efficiency gains from offloading repetitive functions—or making connections between dissimilar, often unrecognized events—gives organizations more freedom to focus on innovative, strategic projects to better serve customers.
Like self-driving cars, the data center that manages itself and calls for help when needed could arrive sooner than we expect. Even complicated IT infrastructure that’s typically difficult and time consuming to upgrade and control is becoming concentrated, automated and divided into elements managed through software, not hardware.
Similar to vehicles, data centers are increasingly employing full self-driving capability. A continuum of analytics-based solutions and automation is in place to save time, money and hassle. Using the storage industry as an example, the range of new devices and orchestration technologies continuously improves machine direction of resources and reduces IT staff engagement. No longer is storage confined by the constructs of decades-old technology.
Let’s begin with a simple analogy about how people choose their cars. Do they opt for a manual or automatic transmission? Self-driving cars obviate that decision: the transmission is automatic by definition. Similarly for storage, most legacy systems require manual intervention and management. More-recent solutions automate implementation and management, hiding the complexity but not eliminating it.
Automation and machine learning offer multiple capabilities that aid in developing the self-driving data center. One is that organizations can guarantee performance without intervention. With traditional storage, applications compete for resources from a fixed number of buckets or IOPS. Guaranteeing a set number of IOPS to a particular application prevents organizations from accessing those IOPS for other applications. In other words, returning to the car analogy, the number of lanes is limited and only one car is allowed to use the fast lane—no others are allowed even when it's unused.
Automation enables businesses to access IOPS resources when and where necessary, and it allows other virtual machines (VMs) to employ them for other purposes. So, although it ensures a clear lane for every VM, it also enables the VMs to access IOPS as necessary. This approach avoids the danger of saving and wasting unused IOPS, instead making them available when needed. Being able to drive your car at 70 mph doesn’t mean you should do so in a 25 mph zone. Automation enables organizations to deploy that performance (IOPS) where it’s most needed rather than waste it.
Also, by giving each VM its own lane, organizations can make the best use of all their performance all the time. On those rare occasions when VMs ask for more than the storage can deliver, performance can be assigned dynamically to applications that require it rather than on a first-in, first-out basis.
A Smoother Ride on the Road to Success
Looking further ahead, machine learning and automation can optimize the performance of storage arrays and predict future usage trends. Machine learning can analyze past performance to predict trends for the next 30 days, for example, giving organizations insight into what’s necessary to optimize performance and capacity for storage-array pools.
By analyzing trends, machine learning can enable organizations to move VMs from a particular array to somewhere else in the pool if performance dips. Better still, it allows organizations to predict and address poor performance on an array.
This approach is similar to the way self-driving cars can employ shared intelligence to switch lanes if one is blocked miles ahead. Machine learning can provide the same capability in how storage resources are allocated to VMs if a disruption occurs or an application suddenly requires greater performance from an application.
Machine learning can also help companies plan further into the future. Analytics enables organizations to improve predictions and make better decisions about infrastructure requirements to avoid downtime. It’s like building another lane on a highway to address additional traffic in the future.
The advent of self-driving cars provides compelling evidence of the shift from human control to automation as a means to make transportation safer and more reliable. Reducing or eliminating manual intervention in the data center through automation, along with improving decision making with analytics and machine learning, can help organizations steer a better path as well as dramatically increase performance and capacity.
Further development of apps and devices that use this learning will take place, and companies will try to find new and exciting ways to include AI. The philosophical debate about AI will always continue, but there are ways to expand its uses without giving it too much control. Automation is making self-driving data centers a reality, guaranteeing real-time-predictable performance without IT intervention.
IT teams will be able to concentrate on more-important tasks that add value to the company rather than keeping the engine running, or simply keeping the car on the road. Companies will achieve this goal when they ensure a clear lane for very every VM. These are exciting times; in the coming months, you’ll start to see how machine-learning-based intelligent automation will become a critical component of modern data centers.
About the Author
Kris Boyd is field CTO of Tintri. He formerly worked for F5 Networks and VMware.