In-Depth

Choosing Between Scaling Up and Scaling Out in Enterprise Storage

Making a game plan before purchasing your storage solution will help to point you in the right direction.

• By Scott D. Lowe
• 03/19/2014

Two terms can often create confusion in the storage buying process: Scaling up vs. scaling out.  Some vendors that can't do one or the other will give you all kinds of reasons as to why their method is the only one you should consider and why the other one is wrong.  However, in reality, both have their place in the enterprise.  It all comes down to what you need to accomplish.

In a scale up scenario, you buy a unit that features controllers, storage, networking and CPUs.  As you need to add additional capacity, you simply buy additional disk shelves, but these disk shelves do not include additional supporting resources -- they provide just capacity and not additional networking or processing.  So, there may come a point at which adding new disks has the potential to overburden those networking and processing resources.

However, just adding disks is less costly than adding all of the brains and guts, too.  So, it's a tradeoff and one that requires careful consideration.

Scaling Out
As you might expect, scaling out is a bit different than scaling up.  With scale-out systems, products take more of a "linear scalability" approach to the equation by adding storage, networking and processor resources in tandem.  This helps to ensure that storage performance will continue to grow along with storage capacity.  Also, never forget about latency that could be introduced as you add more disks without adding more supporting resources.  In modern data centers, latency is one of the most significant issues facing the environment.

Which Is Right for Your Enterprise?
The choice to be made with regard with scale up vs. scale out storage is fully dependent on the kinds of workloads you're running and your expansion needs.  For general purpose workloads and file/print needs, a scale up environment will often be the most cost efficient option, but you need to make sure that your workloads remains within the confines of the head unit.

When you are running particularly I/O intensive workloads or ones that will result in excessive network traffic, you may want to consider scale out storage so that you don't risk running into unexpected performance issues due to storage.

You will also find that scale out systems are gaining in popularity today as these kinds of systems provide organizations with the most predictable performance outcomes.  Today, to reduce latency and provide a better overall experience, many new vendors are combining storage and compute into a single appliance that provides the best balance of performance and capacity while also significantly simplifying the data center.