Compute Analytics: extend predictive analytics from storage to compute

Tintri Analytics has always given you insight into your individual VMs. Today, that power extends to compute, and uses machine learning to boost predictive power even more.

Here’s a favorite anecdote. We competed in a deal with another vendor that bragged about their analytics. In a proof of concept, as soon as virtual machines were motioned onto the competitive system, latency spiked. Since those virtual machines were stuck inside volumes, it was very hard to understand the source of the problem—the prospect naturally blamed the storage.

So, the Tintri rep suggested deploying our platform and motioning the VMs on to Tintri (done in 20 minutes). Our analytics instantly showed that the root cause of the latency was compute—a server gone haywire. Of course, when the prospect saw that Tintri’s analytics were far more accurate and useful, they chose our enterprise cloud platform.

Predict resource needs for storage AND compute

Today’s launch expands our predictive powers even more. For a long time, we’ve helped customers determine exactly how they use resources, and when they will run out of storage capacity, performance and working set. Now, we’ve extended those capabilities from storage to include compute.

Other analytics solutions involve mathematical gymnastics, assumptions and guesswork. Tintri is complete, accurate and easy to use. Case in point …

See latency across compute, network and storage

Tintri Analytics shows you the real-time behavior of every virtual machine or container. When an application experiences latency, you only need to hover over the virtual machine in the Tintri UI. Then, Tintri shows you the real-time root cause. Practically speaking, it puts an end to finger-pointing.

Other providers in our space have been pretty loud about their analytics offerings. But louder doesn’t always mean better, especially if they still lack real-time predictive analytics.

I’ll be blunt in stating that Tintri offers the most complete and accurate analytics in our market. Now, let me substantiate that claim and highlight what’s new.

Analytics at the virtual machine and container level

Our fundamental difference is the ability to provide analytics on EVERY individual virtual machine in your footprint. That’s because of how our CONNECT architecture is designed to take every action on virtual machines and containers.

That’s an important distinction from other solutions that tout VM-level visibility. Those providers take analytics about a LUN or volume and then divide by the number of VMs inside the LUN or volume. What that spits out is analytics about AVERAGE VM behavior, when the distribution of VM behavior inside the LUN or volume could vary wildly. The high-profile solutions that take this approach burden you with inaccurate analytics that could prompt bad decisions.

Meanwhile, Tintri analyzes every individual virtual machine. That also unlocks the real power of prediction. For example, Tintri captures data about the exact behavior of your 90 development servers. So when you want to know the impact of adding 10 more development servers, Tintri can leverage 3 years of history about the first 90 to tell you precisely how adding 10 more will affect your capacity, performance and working set.

We are using Apache Spark and Elasticsearch to crunch more than one million data points every second. Machine learning algorithms use historical data about how you use resources to predict your need for compute CPU and memory up to 18 months into the future. That level of precision puts over-provisioning to an end for our customers.

Excited to learn more? Read our launch blog about machine learning. Then ask your Tintri rep for an analytics demo or contact us directly—we’ve got something to shout about.