Virtualization Technology News and Information
VMblog's Expert Interviews: Tintri Talks Enterprise Clouds, Machine Learning and Real-Time Analytics


Tintri, if you aren't familiar with them, offers an enterprise cloud infrastructure that's built on a public cloud-like web services architecture and a set of RESTful APIs.  Organizations use Tintri all-flash storage with scale-out and automation as a foundation for their own clouds-to build agile development environments for cloud native applications and to run mission critical enterprise applications.  Their cloud infrastructure is designed to deliver the same agility as the public cloud, but in the enterprise data center.  And they recently announced the extension of its enterprise cloud platform to Amazon Web Services and IBM Cloud Object Storage.

While at VMworld 2017, I received an update on the company and learned how Tintri was continuing to extend its enterprise cloud platform over the year ahead.  To learn more, I spoke with Kieran Harty, CTO & co-founder of Tintri.

VMblog:  What does it mean to enable "enterprise cloud" benefits?

Kieran Harty:  First, let's talk about the elephant in the room. Everyone likes to say they can enable "enterprise cloud" benefits. But I just don't think that's true.

According to NIST, most enterprises will operate multi-cloud environments. So this August, we released the Tintri Cloud Connector, which tightly integrates customers' Tintri footprint into their AWS or IBM COS instances. And since Tintri enables VM-level data transfer, dedupe and compression, customers can use Cloud Connector to transfer and restore individual applications in just clicks.

That slashes administrative costs by 72%, lowers cloud storage costs by about 80-88%, and gets you up to 4x lower RPO or RTO. Those are the results of enterprise cloud benefits you can only get with application-level operations.

VMblog:  How does machine learning provide meaningful benefits for organizations?

Harty:  Well, it all comes down to what organizations need. You need software that can analyze data on the application level. You need real-time analytics that can calculate beyond correlations and averages. And you need powerful machine learning algorithms that help you stay on the right track.

At Tintri, we use machine learning algorithms powered by Elasticsearch, Apache Spark and Amazon Machine Learning. So let's say you need to roll out a big SQL deployment, and need to see how it would affect your performance and capacity. Well, thanks to Tintri Analytics, all your VMs, their sizes, space savings and growth rate have been recorded and analyzed already-up to 3 years' worth.

Tintri would feed all that information into our algorithms, which in turn would take millions of data points from your applications, simulate scenarios, and output what-if analysis that predicts how your deployment will go.

And the best part is it would take literally seconds. Seconds to see how your deployment would affect you up to 18 months in the future.

VMblog:  How is the scale-out process improving for organizations?

Harty:  Traditionally, going into the data center is just a headache, right? You've got to think about rack space. You've got to think about cabling. You've got to think about adding 8 drives at a time even if you only need one or two.

And that's just the physical stuff. When your organization starts working on thousands or even tens or hundreds of thousands of applications, keeping things simple starts becoming, well, not simple. How do organizations on conventional storage manage all those applications efficiently?

They don't, at least, not well. Which is why at Tintri, we make it easy to start small-with just one 17 TB system-and scale out to more than 20 PB, keeping all the policies you set as you scale to hundreds of thousands of VMs. We keep storage and compute separate, so you can scale either as you need them.

And those same machine learning algorithms I talked about earlier automatically see what the best place for each individual VM is. All you have to do is review their recommendations, hit "execute," and watch your VMs migrate. And with that application-level data, we can also predict what-if situations, like how another 10 development servers would impact your footprint, based on how your 90 previous development servers have acted.

As for the physical stuff I talked about earlier, Tintri works as a loosely coupled federated pool of storage. So when you want to add more capacity and performance, you just add another device, which is automatically added to the pool. In other words-the data center is no longer a headache.

VMblog:  Everyone says they provide "real-time analytics."  How does Tintri differ?

Harty:  Everyone does say they provide "real-time analytics," but with some exceptions, most of those "real-time analytics" are more like real-time correlations or averages. Which is an obvious issue: after all, Bill Gates raises the average net worth of a room whenever he enters it. It doesn't mean everyone there is actually a millionaire.

That's the fundamental difference between Tintri's real-time analytics and the competition's. Our storage is architected for enterprise cloud, meaning we can capture data about the exact behavior of every VM's individual behaviors, and then act accordingly. Only Tintri's built on the currency of cloud.

This past August, we also extended our analytics capabilities even further. Before, customers could use Tintri Analytics to see when they would run out of capacity, performance and working set. And now, just as we provided real-time analytics into compute, moving forward we're extending predictive analytics into compute.


Published Wednesday, October 04, 2017 7:33 AM by David Marshall
Filed under: ,
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
<October 2017>