Navigating Server Refresh Possibilities—Rackmount, Blade Server, or Hyperconverged?

Jim Fields
Jim Fields

It’s time for another round of server refreshes in your enterprise, so will you continue to deploy the same server style (rackmount, blades, or hyperconverged), or consider something new? Here are some thoughts to be considered when selecting one server style over another. As every major server vendor has one (or more) offerings in each of these categories, a review of the pros and cons can be helpful.

What data center today doesn’t have some rackmount servers? They seem to proliferate like rabbits.

When you need to scale out an application, or add a new one, it’s a simple decision to go with another rackmount server. But has their familiarity possibly blinded you to a better option?

Rackmount Servers—Pros

  • Tried and true, very well-known capabilities
  • Processor performance and memory capacities continue to grow
  • Sizes for most every scale-out or scale-up situation
  • Offer the most flexibility when it comes to I/O options

Rackmount Servers—Cons

  • Each rackmount server requires a set rack unit (1U, 2U, 4U), further extending server sprawl
  • Each rackmount server requires additional Ethernet and Fibre Channel SAN core switch ports
  • Power requirements may be taxing your rack PDUs (and UPSes); at some point more power could be required, at significant expense
  • Management tools are okay, but lack some capabilities in other server solutions
  • While storage capacities have continued to grow, this storage can be stranded for one server, forcing a consideration for new or expanded SAN or NAS storage systems

In the over 15 years that blade servers have been around, they’ve gone mainstream for many mid-size and large enterprises. Initially, they were somewhat limited in what they could do, but today, there are very few applications that a blade server can’t do just as well as a rackmount server.

Blade Servers—Pros

  • More servers can be consolidated into less rack space
  • Same processor and memory options are available as rackmount servers
  • Integrated Ethernet (and Fibre Channel) switches require far fewer core switch ports
  • Wire once means adding additional servers to an enclosure in the future is a snap
  • Redundant, shared power and cooling saves power in today’s expanding data centers
  • When adding four or more servers at once, blades can be less expensive than rackmount servers
  • Once an enclosure is in place, add-on servers are less expensive than rackmount servers
  • Storage blades make 12 or more drives available to blade servers
  • Management tools make it just as easy to manage a group of servers as to manage one

Blade Servers—Cons

  • Price of entry is higher due to integrated switches, power, and cooling for multiple servers
  • Enclosures require greater space and can cause hot spots (due to greater numbers of servers)
  • Enclosure power requirements (to support 8 or more servers) are higher than simple rackmount servers, requiring different power (C19s rather than C13s)
  • Only a few blade enclosures offer storage equal to or greater than rackmount servers

Hyperconverged servers, a building block approach to provide server, storage, and storage networking, started as an ideal solution for VDI, and has now been adapted for many diverse virtualized workloads. Based on industry-standard rackmount servers and local storage (spinning and/or flash), hyperconverged servers offer a way to power your applications without concern for separate expensive storage (which requires separate management and maintenance).

Hyperconverged—Pros

  • Can provide the compute and storage needed for one or more virtualized applications
  • Simplicity of a hyperconverged solution
  • Shorter time to deployment and relatively high reliability due to vendor testing
  • Hyperconverged can allow an enterprise to scale as they need to
  • Management tools make managing multiple nodes no harder than managing a single server
  • Pay-as-you-scale model, but costs are more predictable, and you’ll know exactly how much power and capacity each additional node will add
  • Since a hyperconverged solution is provided by a single vendor, you have one point of contact for everything

Hyperconverged—Cons

  • Price of entry can be higher due to required number of nodes, software (and sometimes hardware), but when compared to all the components they replace, can still be cost effective
  • Require applications to run virtualized; bare-metal applications that don’t support virtualization can be a problem
  • Some hyperconverged offerings require three nodes minimum
  • While based on industry-standard rackmount server platforms, there may be fewer processor and/or memory options than a rackmount server
  • Nodes within a hyperconverged cluster often must be configured exactly alike—processor, memory, and storage
  • While some hyperconverged server nodes support GbE networking, they really need 10GbE to provide the proper performance, especially for the storage sharing across multiple nodes
  • As these are building blocks, if you need only more compute or only more storage, this solution often requires the addition of another block of compute and storage (some vendors have solutions that can add either additional compute or additional storage to the cluster)
  • Some hyperconverged platforms support only one or two of the standard virtualization platforms (not a problem if they support the one you already use)
  • For a specific hyperconverged cluster, you’re locked into one vendor

With such a variety of solutions, the options are there to match your specific workload to the best server style. There are a number of variables that might impact your decision: your enterprise has already invested in a FC SAN infrastructure, making hyperconverged less attractive; your data center is running into power and cooling issues that require a change to something more efficient, so blades might be the right direction. With the pros and cons above, you’ll be better able to make the correct selection. If you’d like to discuss the future of your data center with a Connection engineer in the Data Center Practice, please let your Account Manager know.

Jim Fields

Jim Fields is a Senior Systems Engineer at Connection with more than 38 years of experience in servers and server management. He holds several technical certifications, including HPE Master Accredited Solutions Expert (MASE)—Advanced Server Solutions Architect V3, HPE Accredited Solutions Expert (ASE)—Composable Infrastructure Integrator V1, and HPE ASE—Server Solutions Architect V4. In his free time, Jim enjoys playing the trombone and singing at his church, as well as computer programming and volunteering with the Boy Scouts.

Leave a Reply

Your email address will not be published. Required fields are marked *

© 2019 CONNECTION, INC. ALL RIGHTS RESERVED.