Can this all-flash storage really deliver autonomous operations?

Autonomous operation is the new buzzword in all-flash storage but can your infrastructure really run without manual intervention? Tintri has the answer.

KEY TAKEAWAYS

  • Why all-flash storage players are using the word autonomous
  • Autonomous operations require multiple components to work together
  • Why autonomous operation matters when building an enterprise cloud

We operate in a noisy market, so naturally we pay close attention to the competition. So we’ve been interested to see a term we’ve used for years begin to diffuse into the wider market: autonomous operation.

When you hear the term autonomous, you probably think about autonomous vehicles (especially if you live in the San Francisco area and see Google’s cars crawling about town). These are cars that are navigated by cameras and computers through all sorts of situations and conditions.

In terms of all-flash storage, autonomous operation refers to infrastructure that works without manual intervention.

So, why would a provider that has been touting all-flash storage performance for years suddenly make autonomous operation their headline? It’s simple. Enterprises are seeking cloud capabilities—and hardware isn’t the solution; it’s all about software.

Autonomous operation matters a great deal when you’re building out an enterprise cloud. The primary reason is scale—if you’re thinking cloud, you’re likely planning to grow your virtual machine count from hundreds to thousands … even one day, millions. As you scale your footprint, if your management burden grows in direct relation you’re adding a lot of operating complexity and expense. But if your all-flash storage works autonomously, you can add petabytes and manage it all with the same staff you have today.

But buyer beware—when you hear claims of autonomous operation—you’ve got to look under the hood. Autonomous operation isn’t a feature—it’s the outcome of multiple web services working together.

Delivering autonomous operations starts with the right level of abstraction. Abstraction is the level of granularity at which infrastructure operates. In all-flash storage terms that can be LUNs, volumes or individual virtual machines. Tintri is an example of storage that operates in individual virtual machines—the same units used by compute, network and other cloud components. If the components don’t support the right abstractions, it’s not possible to operate autonomously. That’s the case with the irrelevant abstractions of traditional storage – LUNs and volumes.

Once your storage stack has the right level of abstraction, it needs other web services that can deliver performance, predictability, control and native policy based management.

A traditional all-flash storage can at best deliver peak performance but it lacks:

  • Predictability because it can’t deliver VM granular Auto-QoS or use VM granular analytics to know when and where to move VMs and workloads to optimize performance and capacity
  • Control as it can’t guarantee AND limit performance to a VM. Controlling BOTH maximum and minimum performance is essential—especially when auto-QoS throttles the wrong workload
  • Native policy based management as it can’t automatically apply and retain policies on VMs, across all major hypervisors, as they are added or moved around your footprint

Let’s try and be illustrative with a metaphor. Think of Tintri as providing a fleet of autonomous limousines, and conventional storage as providing a city bus. Each limousine provides each passenger a personalized experience that is independent of other passengers—it autonomously works to get the passenger from A to B as quickly and efficiently as possible.

On the other hand, the city bus is shared by a group of passengers, who must follow a single route that is not optimized for any individual passenger. All passengers on a bus also need to share whatever resources are available on the bus. You can make that bus ‘autonomous’ by tying a brick around the gas pedal (look—it just goes!) but it won’t give the passengers—your virtual machines—exactly what they need. And unless you’ve got Keanu Reeves running your footprint, then you’ve got the wrong kind of Speed.

The bottom-line for you is simple. You’re going to hear more all-flash storage players talking up autonomous operation. You don’t even have to take it for a test drive to know how it works—just ask the salesperson if the level of abstraction is volumes and LUNs or individual virtual machines. If it’s volumes or LUNs, it ain’t a limo—it’s a city bus.