More Isn't Always Better
First of all, VMTurbo (Turbonomic’s old name) was a presenter at my first-ever Tech Field Day event back in early 2014. Although I’ve talked with them since then, I was excited to get a full-on Tech Field Day update roughly 3 years later and hear about what’s new. Secondly, my friend Eric Wright (@discoposse) was also at that VMTurbo presentation many moons ago and he took an interest in the technology and now works for Turbonomic as their Technology Evangelist! Any opportunity to see and hear from Eric is a welcome one.
The presentation commenced and I heard about what I expected to hear – the overview of what Turbonomic does, how the virtual supply chain and resource economy model makes sense, and so on. I still think the story is compelling, just like I did three years ago.
There was one statement during the presentation that stood out to me though. Mor Cohen-Tal, Product Manager at Turbonomic, said something wise that I think demonstrates the clarity of vision and the maturity that Turbonomic has at this stage. She said,
“We don’t think that the more metrics we collect, the better we are. Our goal is to deliver application performance as efficiently as possible. Collecting more and more metrics that we can’t control is not what we’re trying to do. Monitoring is more of a byproduct.”
What Do You Really Want?
In the ever-bigger, ever-faster world of enterprise IT, I think we can sometimes get confused about what we’re really after. Speeds and feeds are attractive, especially to quantitatively minded people.
But at some point, the obscenely fast storage array becomes – to the average IT organization – a bit like buying a McLaren for a Sunday driver. Yeah, it’s got a 650 horsepower V12 engine, yes it does zero to 6o in 3.4 seconds, and yes it can do a 11.6 second quarter mile. But you’re going to drive it up to the grocery store once a week and take it to visit the relatives.
Mor was reacting to a question about how metrics are collected (agent-based vs. agentless) and what sort of metrics are collected, and I think her answer was spot on. In essence, she said that the only metrics that matter are the ones which Turbonomic can directly act on to assure application performance. Which is what we all actually care about anyway, right?
I believe this outcome-oriented mindset is key to the value proposition of Turbonomic’s platform. A few minutes later, Mor gave an example of how focusing on the result is so much more important than focusing on the metrics themselves.
“Giving more CPU to a VM might mean that it’s going to have higher ready queues. Sometimes taking CPU’s off is going to increase performance.”
Looking back at the Tweets, I see that I commented on this during the presentation, too.
I can count on 1 hand the # of admins I’ve ever met that would remove vCPUs to increase performance. Even though it would’ve worked #TFD14
— James Green (@jdgreen) May 11, 2017
The takeaway from this article is this: when you’re considering a product or solution, consider what it is that you’re ultimately trying to get rather than how many nerd knobs it has an how fast it goes, or in this case, how many metrics it can collect. Look at how well it can deliver what it is you really want. In the case of Turbonomic’s Autonomic Platform, that would be the most efficient use of resources possible while assuring application performance. My personal opinion is that they’re doing a good job of delivering on that.