0 0

VM-level QoS Redefines Quality

Today, Tintri announced the release of a set of new capabilities that showcase the benefits of managing storage at the VM-level. QoS is a great example—everyone does it, but no one really agrees on what it means. Well, today we can show you the simplicity and power of VM-level QoS. Watch this four-minute video, and read on for more details. 


This is the behavior for a single VM in Tintri’s UI. Not a LUN or a volume, but one VM. That’s an important distinction. When you set QoS for an entire LUN or volume, you are giving every component VM—hundreds or thousands—the same QoS setting. So, you either have to set IOPS for the most demanding VMs (which is inefficient), or average out the IOPS (which is ineffective).

In the above screenshot, check out the top graph, where you can toggle the minimum and maximum IOPS. You’re not punching in numbers for min and max… you’re dragging thresholds up and down—that’s a lot more intuitive.

And in the bottom graph you can SEE the immediate impact on latency. Conventional storage puts you in an awkward spot—you set QoS policies and then wait to see if users complain about performance. Tintri removes the guesswork. For example, toward the right side of the image, we’ve toggled down the maximum IOPS, and right below you can see that throttle latency immediately climbs.

A few other vendors let you set min and max IOPS, but no one lets you see how it affects your latency, and no one but Tintri lets you do that at the VM level.


Enterprises – Have a critical VM that needs a certain amount of performance to satisfy users? Dial up the minimum IOPS. Have a rogue VM whose burps and bursts are having adverse effects? Dial down the maximum IOPS to contain its bad behavior. Stop shuffling LUNs around to tune performance… do it with a single toggle instead.

Service Providers – If you have service tiers that you need to control, here’s a simple way to create differentiation. You don’t have to spend time planning and fixing – now you can deploy and move on. If a customer ever asks for visibility into performance you have fully transparent chargeback (or an opportunity to upsell).


On the theme of visualization, let's talk latency. Getting to the root cause of latency shouldn’t be a plotline for Indiana Jones. Tintri lays it bare in one simple visual. You can see whether latency for an individual VM stems from host, network or storage… and if it comes from storage, whether the source is contention, flash or disk. Now you don’t have to guess, you’re sure to ‘choose wisely’.

VMstore Performance

They say a picture is worth a thousand words. We believe these pictures are worth a thousand hours (of your time). That’s the power of managing storage at the VM-level. Learn more about Tintri’s approach to QoS in this simple one-page summary.

In our next post we’ll explore time travel. Now, will we post it in the future, or the past?


Scott Buchanan / Apr 09, 2015

Scott leads Tintri's Marketing function. His team is responsible for Tintri's brand, content, and social media. Prior to Tintri, Scott led marketing teams at Microsoft, Medallia, and Merced (all t...more