Everything you need to know about converged infrastructure, hyperconverged infrastructure, and reference architecture

 In Data Center, Hardware

infrastructureGone are the days when IT professionals had to rack, stack, connect, and configure every piece of data center infrastructure. Now, you can plug in a single box, thanks to converged infrastructure.

Converged infrastructure, hyperconverged infrastructure, and reference architectures are in essence the same thing. They all combine compute, network, and storage resources into one cohesive unit, but the main differentiator is the way they are procured and installed. Before you try to decide whether these options are right for your organization, and which to choose, it’s critical to understand the subtle but significant differences among the three types of converged infrastructure.

Generally speaking, there are three approaches companies can take to settle on a new data center installation:

  1. The hardware-focused, building-block approach of VCE (a joint venture of EMC, Cisco, and VMware), often known as converged infrastructure.
  2. The software-defined approach of Nutanix, VMware, and others called hyperconverged infrastructure.
  3. A reference architecture, such as EMC’s VSPEX or NetApps’ FlexPod, is a documented blueprint that provides recommendations for servers, storage, networking, and virtualization that addresses a particular workload, and has been tested by manufacturers to ensure compatibility.

Here’s the key difference between the first two options: In a converged infrastructure, each of the hardware components (for example, the server or storage) is a discrete element that organizations can use for its intended purpose. Meanwhile, in a hyperconverged infrastructure the technology is software defined, resulting in an integrated approach that can’t be broken into separate components.

Both options have benefits. Integrating your systems under one roof tends to reduce IT costs and usher in uncomplicated maintenance and troubleshooting. But IT departments must understand the differences among the technologies, so here’s a breakdown of the choices.

The different options for converged infrastructure

A number of manufacturers — VCE, Hitachi, PureStorage, SimpliVity, and others — currently sit in the pre-built converged infrastructure space. Manufacturers configure these systems in the factory and send them to the customer as a complete unit that organizations can plug in at their facility. They’ve been pre-tested, so they’re fully functional once they’re turned on. Customers have a handful of choices for a converged infrastructure, a few of which I’ve detailed below.

VCE (recently acquired by EMC) systems utilize VMware as the hypervisor, Cisco for the compute and networking components, and EMC VNX/VNXe as the storage unit. These systems are sold as vBlocks.

Hitachi has taken a similar approach, combining Hitachi Blade Servers with HDS Storage systems and Brocade network switching to comprise the Hitachi Unified Computing Platform (UCP). Hitachi UCP solutions are tested with specific applications, such as Microsoft Exchange, SQL Server, SharePoint and Lync Server, Oracle Databases, and the SAP HANA platform.

PureStorage offers FlashStack CI for VMware and VMware Horizon. These systems incorporate PureStorage Flash Arrays with CISCO UCS B-Series Blade Servers, CISCO Nexus switches, and VMware software. The FlashStack is built and configured based on the customer’s requirements and shipped as a turnkey system.

What you need to know about hyperconverged solutions

Hyperconverged systems scale out performance and capacity to end users and applications by virtualizing compute, storage, and the hypervisor across multiple nodes. Data protection and failover are managed between the nodes, and customers typically must start with a minimum of three to account for availability. After the initial implementation, organizations can individually add nodes to increase storage and computing resources. The main disadvantage to hyperconverged systems is that computing power and storage cannot be scaled independently, so a customer that requires more storage capacity must add additional nodes, not just additional disk.

Nutanix systems are probably the most familiar product line at this time. Nutanix utilizes an Intel-based x86 server platform with either VMware or Hyper-V as the hypervisor. Storage is internal to the Nutanix appliance and employs Nutanix Distributed Filesystem to virtualize the storage across nodes. Nutanix requires at least three nodes/appliances, which can be scaled up based on the customer’s needs.

EVO:RAIL is a VMware-developed system announced at VMWorld 2014 that will be delivered by a variety of different manufacturers. The base configuration will support approximately 100 general purpose virtual machines. In the coming months, Dell, EMC, NetApp, HP, SimpliVity, Supermicro, HDS, and Fujitsu will announce their version of the EVO:RAIL hyperconverged infrastructure.

SimpliVity offers the OmniCube, a single software stack that combines multiple traditional IT infrastructure functions in a single shared x86 resource pool. Using its Data Virtualization Platform, the core technology that performs inline data deduplication, compression, and optimization on all data at inception, SimpliVity enables data granularity of just 4 KB to 8 KB blocks. The Data Virtualization Platform is powered by the OmniCube Accelerator, SimpliVity’s PCIe card that offloads compute-intensive tasks to deliver enterprise-class performance, enabling data movement and sharing in multi-node and multi-site environments, as well as global VM-centric management. A single administrator manages the entire global infrastructure through a single pane of glass.

Build your own with reference architectures

Reference architectures are not pre-built systems, but blueprints on how to combine the compute, network, storage, and hypervisor in a manner that has been tested and certified for various workloads. Manufacturers provide specifications on the compute power and storage required to support specific workloads, and customers can select which products they prefer to use.

For example, VSPEX is EMC’s reference architecture, and it details the paths customers can follow by selecting a listed server manufacturer, hypervisor, and network devices to combine with EMC storage. Customers can configure their system to more effectively support their workloads. HDS also offers organizations a Hitachi UCP Select solution that supports third-party Cisco Unified Computing System (UCS) servers with Cisco network and Hitachi storage options. VSPEX, FlexPod (NetApp), SmartStack (Nimble), and PureStorage have reference architecture documentation as well.

Circle back to the SHI Blog for my next post, which will help you determine whether you should adopt one of these converged infrastructures and which to choose. Contact your SHI representative to learn more today.

Related Posts: You may also be interested in...

Showing 2 comments
  • Ranjit Nayak

    Any reason Oracle Engineered systems are not in this list of converged solutions? (https://www.oracle.com/engineered-systems/index.html)

  • Ed McNamara - Director of Communications & Mktg - SHI

    Hello Ranjit – hope all is well. Thank you for pointing out this oversight on our part!

    SHI customers seeking the very solutions your link points to was a huge reason Oracle was SHI’s fastest growing “Top 10” partner in 2014.

    I will ensure we double-check our recommendations in the future to ensure ALL our top technology partners are duly represented.

    Please let me know if you need anything else or have any other questions.

Leave a Comment

1 × two =

Pin It on Pinterest

storage3D printers