What are the biggest benefits of hyper-convergence? According to Ben Woo, it's all about "Predictability, repeatability and scalability."
During his session at Storage Decisions 2015, "Hyper-Converged Infrastructure and Software-Defined Storage: the Good, the Bad and the Ugly," the Neuralytix principal analyst listed some of the key storage benefits of hyper-convergence and what shops must do to see them.
The hyper-converged approach of combining compute, networking, storage and virtualization is valuable for a number of reasons. One of the major benefits of hyper-convergence is single pane of glass management, which can give administrators a better sense of the environment they're managing. Data from multiple sources, such as a data center's facility systems and IT systems, is integrated and presented in one display, which makes it easier for admins to read and interpret. In a hyper-converged system, made up of so many moving parts, this simplified management can be a big help.
But management is "the easy part," according to Woo.
Before reaping the benefits of hyper-convergence, there are three components to look for in a hyper-converged system.
"I don't care about performance, honestly," Woo said, emphasizing the importance of having predictability with hyper-converged systems. "If you can't predict the performance, it doesn't matter."
Performance itself can be a struggle with hyper-convergence, but being able to predict performance and scalability can help an administrator know how hard the system needs to work.
Repeatable instances are the second necessity, not to be confused with the previously mentioned predictability.
"Repeatable means that every time you put a module in place, it's the same process," Woo explained. Some organizations even hang on to outdated, expensive technologies at the risk of being unable to repeat the process that technology has nailed down, he said.
One of the other major benefits of hyper-convergence is its scalability, Woo said. By allowing users to add nodes as needed, hyper-converged systems are highly scalable. There are some limitations with this scalability, but that can come down to the vendor and product. Because of this, doing research on different vendors and hypervisors prior to making a purchase is vital.
Transcript - Benefits of hyper-convergence and how to take advantage of them
So what else is convergence good for? I already talked about this concept of procurement; one PO [purchase order], one service contract. The third point is about single pane of glass management. That's the easy part.
The standardization of each of the modules is equally as important because it then enables you, for the fourth point, to talk about predictability. When you think about converged infrastructure, when you think about hyper-converged infrastructure, you have to think of three things.
Number one is predictability. I don't care about performance, honestly. Because if you can't predict performance, it doesn't matter -- it's a crapshoot and you're back to where you were before. What you want is a solution that is predictable in nature. It's predictable in how it scales. It's predictable in how linear it is in terms of performance. When I add something else to it, it's got to work twice as hard.
Number two, we want to look at repeatable instances. This is not the same as being predictable. Predictable means that when you put something in, it's going to work X times more based on your current environment. Repeatable means that every time you put a module in place, it's the same process.
Why does Citibank still run a Symmetrix from 10 years ago? Why does Citibank still run an RMO 6? For those of you that don't know, it was a 1970s disk drive. For people that are small enough, like me, they would have dropped me upside down to pull the disk packs out of it. It has a whopping 6 megabytes on it. It chews up about the same power as a full-sized building, and it requires cooling that is immeasurable.
Why? Not because we can't put 6 megabytes now on the tip of your thumb, but process. It's all about process. So from a repeatability perspective, if I put something else in, can I repeat that same reliability? The answer is probably "yes," but nobody's ever pulled the plug on the RMO 6, and they're afraid if they do, and try and plug it back in, the damn thing won't spin up again. And this is a true story.
The third thing is scalability. So we have P-R-S: predictability, repeatability, and scalability. Those are the three things we look for in any converged infrastructure.