What Kind of Storage Buyer Are You? (Part One)

In case you haven’t noticed (and I am betting you have), something has not-so-secretly occurred, which has shifted the center of the IT universe: storage has become the gravitational constant. Attend almost any industry trade show in 2016 and you’ll see proof of this. The shows are all ostensibly about some central theme, brand, company or product … But 60%-80% of the floor space is all about where you put your data, and how you get at it when you want it.

The array of choices is dizzying, and can be confusing as you wade through the long list of “best thing ever” products. This post is the first of a series designed to help you – the new storage buyer – make sense of it.

It’s not just the sheer number of companies, but the scope, focus and diversity of solutions. There are entire companies devoted to marketing storage for VMware, for instance … besides VMware themselves, who also market Virtual SAN, for VMware. There are solutions designed exclusively as add-ons to Microsoft Windows Server, or various flavors of Linux, and general storage products which work – at least in theory – with anything and everything. There is convergence, hyper-convergence, scale-up, scale-out, server-based and software-defined to contend with (and more), with often confusing combinations and overlaps.  This is important because data is hard to move, and has a large influence across all of IT.   So let’s make sense of these concepts, with a view towards your needs and those of your company. First up: Scale-Up vs. Scale-Out, and how you might best implement one, the other, or both.

“Scale-Up” (sometimes called Vertical Scaling) generally refers to traditional (Storage Area Network) SAN implementations. It is typically characterized by dual controllers which centrally manage the storage resources of one or two standard racks of drives, or sometimes a large “frame” housing the same equipment.  Depending on the system the controllers are either contained in a single chassis or implemented as two separate controller heads in different collocated chassis.  The controllers may have large quantities of DRAM, and multiple, very-fast processors. Reliability, Availability and Serviceability (RAS), is enabled through a doubling of resources including the controllers, power and cooling. They are often referred to as having “no single point of failure”, “five 9s”, and 7x24 operations.  Performance is primarily enabled though very highly optimized proprietary controller designs including hardware and operating system software.  Separate physical networks are used for data and management traffic.  The data network is typically provisioned over high-speed fiber optical links to enable fast transmission of data traffic.

You “scale up” by adding more disks for capacity, or adding more DRAM or faster processors to increase performance. At some point you reach hard limits on expansion, and need to replace the storage frames with new equipment. This approach is typified by vendors like EMC, NetApp, Tintri, Nexenta and others.


Like a skyscraper, scale-up storage is powerful, but can only grow so large

Scale-Up Advantages Scale-Up Disadvantages
  • Often exhibit lower power consumption
  • Central location of hardware leads to easier HVAC solutions
  • Has historically performed better with low-latency applications (OLTP, Decisions support, Simulation)
  • Very predictable performance
  • Potentially uses less networking hardware than in scale-up, depending upon specific application
  • Proprietary hardware and system design can lead to higher costs
  • Fault tolerance usually limited within a single system
  • Changing or upgrading can be more difficult or costly
  • May not lend itself well to cost-effective provisioning of remote-office of SMB implementations
  • Requires careful forward-looking planning.
  • Does not easily accommodate unplanned growth or variable changes in storage demands


Like an office campus, scale-out storage can expand with multiple, smaller and less costly units

“Scale-Out” (sometimes called Horizontal Scaling) refers to implementations using a larger number (but at least two) of smaller, less individually powerful servers which each possess CPU and storage (drive) assets. Management, provisioning and fault-tolerance are decentralized, which is leverage when upgrading or changing individual systems. The server/storage nodes themselves may be geographically separate as well, sometimes spanning multiple buildings, cities or even continents. Network traffic may be flexibly provisioned according to business and performance needs, and typically supports a variety of technologies and protocols. Theoretically, you can “scale out” infinitely by adding additional storage nodes, and don’t’ necessarily need to replace older equipment until it becomes cost prohibitive to service. This approach is typified by vendors like Nutanix, Atlantis Computing, DataCore, and StarWind.

Scale-Out Advantages Scale-Out Disadvantages
  • Historically much lower overall solutions costs.
  • Theoretically infinite ability to scale performance and capacity, on a “pay as you grow” basis
  • Fault-tolerance is distributed over multiple nodes, and can be distributed geographically.
  • Distributed resources may require fewer data center class facilities
  • Decent performance gains with low latency applications
  • (OLTP, Decision Support, Simulation etc.)
  • Common networking requirements
  • Some solutions may lack mature management tools
  • Physically distributed systems may require a distributed support model, which can lead to higher management costs
  • Some solutions may result in greater HVAC costs.
  • Some solutions may increase licensing costs
  • There are still performance challenges with some mainstream line-of-business applications.

So which is “better”? That depends upon the needs of your business, including planning for future growth. The “challenges” noted for each solution type need to be researched to see how they affect your outcomes. Some vendors have made significant advances in these areas. If your application needs are well understood, unlikely to change, and you can plan growth effectively, there is a good case for traditional Scale-Up, SAN systems. If you expect more dynamic and unpredictable circumstances, are implementing virtualization or hyper-convergence solutions (more on this in a future post) and want to carefully match requirements to cost, Scale-Out solutions might be a better bet.  As Scale-Up systems adapt to integrate Scale-Out storage, and Scale-Out storage matures, the distinction will become blurred and fuzzy, (if it isn’t already). Hopefully this entry-level post (and those which follow) will help you navigate to the best solution for your needs.