Ignore the Scale-Up vs. Scale-Out Debate:
It’s Time for Something Better
Written by Lynn Orlando
Published on February 20, 2020
Whatever you’re doing with your storage, you’re likely going to need more of it over time, so what’s the best way to expand? Historically, you have had two choices: scale up or scale out.
Each has its advantages, but your best bet now is to avoid the argument entirely because neither model is particularly great at meeting real-world needs. Why commit yourself to one or the other if you don’t have to? Until recently, you didn’t have a choice.
Today, you have another choice. The Stellus “scale-through” technology can give you the best of both worlds while avoiding the costs and drawbacks of yesterday’s solutions. To understand why that’s such a big deal, let’s review how we got here and why traditional approaches aren’t getting the job done.
Scale-up vs. Scale-out
The scale-up model is basically as it sounds: go bigger. If you’re talking about servers in a data center, scaling up means replacing a size “Large” server with size “XL” to handle more requests. In storage architectures, scaling up means you have a storage array with a single controller, and you increase capacity by adding more drives to the array.
Historically, this was the simplest and least expensive way to scale, but there are obvious limitations. For example, when you’re using the same storage controller your array’s throughput stays the same as you add capacity. Given that you needed to add more capacity, you’re probably managing more data, so at some point you’re going to max out on performance. What do you do then? Enter scale-out storage.
Scale-out is the “newer” approach pioneered by the big webscale companies (Google, Amazon, Facebook) for their hyperscale data centers serving millions of application requests. The idea is that instead of getting a bigger platform you just add more platforms and spread out the job across them. In storage this means adding new nodes (each being a server with attached storage).
The advantage here is that you’re no longer limited by a single controller; you’re increasing capacity and performance as you scale. The tradeoff is that with scale-out architectures you’re adding both capacity and performance, whether you like it or not, most often paying for resources you don’t need.
A Smarter Approach: Scale-Through Storage
The only reason the scale-out vs. scale-up debate continues is that both models have big downsides. You’re basically arguing about whether you’d prefer technical headaches or runaway costs. If neither sounds appealing, yet you’re still asking whether you should scale up or scale out, you’re asking yourself the wrong question. What you should be asking is this: Isn’t there a better way?
Today, the answer is a definite yes. It’s called scale-through storage, and it delivers the best attributes of scale-up and scale-out—without the limitations. Available in the groundbreaking Stellus Data Platform, the scale-through technology represents a brand-new approach to scalability. It combines scale-out principles with a simpler (much higher-performing) file system and the flexibility to scale capacity and throughput independently. Seriously.
Here’s what Stellus scale-through storage delivers:
- Software-defined storage: Dynamically scale storage as needed without having to worry about limitations or dependencies in the underlying hardware.
- Ultra-fast key-value-over-fabrics performance: Believe it or not, even the latest storage systems are based upon fie systems that are about 16 years old. As a result, performance gets bogged down by processes designed for spinning disk drives—processes totally irrelevant to modern flash media. Taking a cue from the big webscale pioneers, the Stellus software is optimized for the solid-state drives and massively parallel architectures of the present.
- Decoupled capacity and performance: Stellus lets you scale how you want, when you want, without paying for resources you don’t need. Add a Key-Value Store (KVS) when you need more raw capacity. Add a Data Manager to increase application throughput. Do one or the other, or both, as needed, up to 1.457TB of storage with up to 80GBps consistent throughput.
Get smart About Scaling
Let the scale-out versus scale-up arguments rage. While other vendors try to convince you that their tradeoffs are less onerous than someone else’s, you can ignore the debate entirely. It’s time to demand storage architectures based on technology from this century. With scale-through storage, you can ensure that no matter how your requirements evolve, you’ll continue to meet them—paying only for what you need with performance that legacy approaches can’t touch.
Ready to learn more? Visit https://www.stellus.com/products/stellus-data-platform/