BlogData Center
Blog • • 6 minute read

Streamlining Your IT Toolbox: A Guide to Vendor Selection

Navigating the plethora of choices in IT infrastructure can be as overwhelming as selecting a simple sandwich spread. This article explores strategies for optimizing enterprise storage solutions, balancing simplicity and specialization, and emphasizes the importance of a well-defined strategy to streamline decision-making and management.

In this blog

  1. Myriad choices
  2. Optimize for the 80 percent
  3. Product catalog
  4. OEM portfolio matrix
  5. Snowflakes will exist
  6. On the management front
  7. Conclusion

Nut/seed butter seems so straightforward until you are in the aisle at the grocery store, trying to select one to purchase.  Do I have peanut or tree nut allergies? Do I want crunchy or creamy?  What about no-sugar-added or sweetened varieties?  Do I want jelly in the same bottle as the nut butter, giving up modularity for simplicity?  Once those choices are made, it's time to choose from the multiple brands available. We now live in a world full of choice, where every needed item has fifteen different options to choose from and all I wanted was a simple sandwich.

 Myriad choices

In the enterprise infrastructure space, there is some advantage to utilizing as few vendors and products as possible.  To that end, I have seen customers move from disparate OEMs to a single compute and storage stack to simplify support when something goes sideways.  Consolidation simplifies the purchasing, operation, and support of the environment. However, that approach is generally limited to specific areas like mainframes or other specialized infrastructure stacks.  For everything else, most customers utilize at least two OEMs for much of their infrastructure for pricing, platform features and supply chain reasons.  How do you go about choosing the items in your toolbox? 

 Optimize for the 80 percent

If you have a passing interest in industrial design, you may have heard that the kitchen tool company, OXO, optimizes for the edge case users, with the standard user needs being taken care of in the process. However, when choosing IT platforms, my general rule is the opposite: optimize for the bulk of the workloads and deal with the edge cases individually. I mean, you could take the all-in-one approach but you're probably providing more service than is needed for most workloads, and the commensurate cost. Most of your environment likely fits into some small number of T-shirt sizes, so determine what those need before dealing with the outliers.  There are probably a few response time objectives, replication (local and remote) requirements, and availability needs around which you'll build your platform strategy.  Additionally, there are things the storage admin team will care about like scale-up versus scale-out and how those affect upgrades and end-of-life migrations.

 Product catalog

So you have decided to investigate what the market has to offer. Let's take a gander at a few of the more common storage OEMs.  In the enterprise storage space, each protocol (block, file, etc.) has options from several OEMs.

If I need a block platform, here are my options:

  • Dell - 5 options (PowerMax, PowerStore, Unity, PowerFlex, PowerVault)
  • Pure - 1 option (FlashArray)
  • NetApp - 2 options (AFF/FAS/ASA, E-Series)
    • Since AFF, FAS, and ASA all run ONTAP, I consider them a single platform, just with different enabled features.
  • Hitachi - 4 options (VSP 5K, VSP One, VSP One SDS, VSP E)
  • HPE - 4 options (Alletra 5K/6K, MSA, Alletra MP, Alletra 9K)

Now, if I want file, here are my options

  • Dell - 5 options (PowerMax, PowerStore, Unity, PowerScale)
  • Pure - 2 options (FlashArray, FlashBlade)
  • NetApp - 1 option (AFF/FAS)
  • Hitachi - 2 options (VSP One File, HNAS 5K)
  • HPE - 3 options or 7 if including HPC/AI (StoreEasy, Qumulo, Greenlake for File, and a slew of HPC/AI options)

Note: I am doing some model consolidation since I view things like QLC or hybrid models the same as the core system, so things like FlashArray//C are the same as FA//X or //XL.  As far as market categorization, those differences can matter, I don't think they do for this particular use.

 OEM portfolio matrix

First up, that's a lot. Now then, the OEMs above have different strategies for covering the storage market segments.  Some of them, like Pure and NetApp, have a limited platform count, so the choice seems simple, but their products must cater to the spectrum of requirements in all market segments.  On the other hand, you have HPE and Dell with portfolios of products to address specialization within the market segments and cater their individual needs.  They each have ways to handle various needs, from the slowest to the fastest on-premises needs to public/hybrid cloud scenarios. There are a lot of table-stakes features like replication, an AI ops SaaS portal, and simple management. The nuance lies in the details of those things along with the aforementioned things like scaling and migration.  It's always worth a conversation with your technical team to delve into those subtleties as you build your strategy.

 Snowflakes will exist

Taking care of the inevitable outliers is a per-need task. Given there are only so many platforms out there, it may be possible to maintain a highly scalable, highly performant platform for all snowflakes, protocol-dependent, of course.  A common approach is to utilize dual controller arrays for the bulk tier and scale-out systems for the workloads with more comprehensive needs.

 On the management front

Management of platforms is a critical consideration; point-and-click web GUIs, while approachable, do not scale well.  To keep the management team sizes from increasing with each new storage system, automation becomes critical for management, consistency, and auditing.  Day-to-day provisioning and de-provisioning can be handled via automation used by the storage team or put into the hands of end customers with some guardrails around its use; VVOLs or assigned pools come to mind here.  Some of our customers have a goal of only interacting with systems via automation. Finally, when it comes to management, I propose that training on new platforms is an important step to operationalizing that system.  Storage platforms, with their modern simplicity, are far more intuitive than those of yesteryear, but an understanding of OEM best practices is critical for long-term fault-free operation.

 Conclusion

What am I hoping you'll take away from this?  Perhaps it's just commiseration amongst IT peers. We can be both excited about and lament the variety of offerings on the market for dealing with all needs from specialized to generalized. On the other side of the coin, my team has developed a rubric we run through in a few hours with the relevant stakeholders and can narrow down the field to options that are testable faster than you can paint the Golden Gate Bridge. 

If you have already narrowed it down or, as an extension to a storage decision workshop, we have a purpose-built facility for testing these products, a number of systems in-house, and can closely replicate your production environment, even down to mimicking WAN latency.  

We live in this broad field and can quickly assist so you can get back to innovating. In closing, it's important to have a strategy to begin with. If you don't know how to get where you're going, it's going to be a rough trip getting there.