What is the difference between Converged Infrastructure (CI) and Hyper-Converged Infrastructure (HCI)? When is it best to use each one? This article helps you sort out the question of Converged vs. Hyper-Converged Infrastructure (or architecture) and what each choice might mean for your data center. It is a companion to our Top 10 Converged Infrastructure Solutions Report and the Top 10 Hyper Converged Infrastructure Solutions Report.
What Came Before Converged and Hyper-Converged Infrastructure?
Before delving into Converged vs. Hyper-Converged Infrastructure, it makes sense to ask, “Converged as opposed to what?” CI and HCI are alternatives to the traditional approach to IT infrastructure that most of us work with every day. CI and HCI are not necessarily replacements for traditional infrastructure. Rather, they represent different ways of organizing and managing the four basic components of infrastructure: compute, storage, networking and server virtualization.
Most infrastructure in use today was set up using a “best of breed” approach. Typically, architects would devise an infrastructure plan that called for the most suitable solutions for servers, compute, data storage, network and server virtualization for a specific workload. Each solution might come from a different vendor and have its own separate management tools. Even if the components came from the same vendor, they might still be controlled by different management software. An overall infrastructure management solution might be placed above everything, with some degree of integration between the management tools for storage, compute and so forth.
There are several advantages to the traditional approach to infrastructure, but also a number of drawbacks. It can provide a high degree of flexibility and customization, but at a cost. There are direct costs for all the specialized hardware and software elements. Administrative costs can also be high, given how many separate elements have to be managed all at once. As organizations start to move systems to the cloud, traditional infrastructure may not port over well. New modes of desktop deployment like Virtual Desktop Infrastructure (VDI) may also not fit well with traditional infrastructure. These are the challenges that CI and HCI try to address.
Why do storage professionals switch over to Converged and Hyper-Converged Infrastructure?
For some IT professionals it’s a matter of cost. This Senior Systems Engineer at an enterprise company reviewed Flexpod, currently ranked as the top converged infrastructure solution according to the IT Central Station community. He writes, “We needed to migrate away from our older servers. When we did the cost analysis through the FlexPod, and the cost of replacing each individual server, it just made more financial sense going with FlexPod in the long term. Previously to this solution, we were using individual Dell and HP servers. It was kind of a mishmash.”
For this CTO at a large healthcare company it was a combination of both cost and reliability. In his review of VMware vSAN he writes, “The value that vSAN brings to our organization, really there are two major areas. One is the ability to replace very expensive proprietary SANs. The other is the need to replicate and keep data available at all times across three separate data centers. Those two elements are really where vSAN plays.”
When this IT Systems Engineer moved to Nutanix, his main consideration was speed. He notes, “We can deploy new servers faster than ever. Our capacity to grow is bigger than when we had SAN storage dependency. We are now able to deploy a pool of QA virtual machines for testing purposes in minutes rather than in hours.”
Like many IT concepts, the concept of a CI platform means different things depending on whom you ask. Industry buzz can do that sometimes. For some, CI is a reference architecture that specifies the elements and configuration of a converged appliance. The owner of the reference architecture is able (or expected) to create the appliance on his or her own. For others, CI means a distinct vendor software offering that embodies CI concepts.
Ultimately, both versions end up in the same. Whether through an open source CI reference architecture or a vendor-specific software package, a Converged Infrastructure appliance is a single box system comprising networking, storage, compute and server virtualization.
Structurally, CI is a three-tier software architecture. It uses the client–server software pattern where the UI, business logic, compute, storage and data access are in separate architectural layers. Each can be maintained and updated separately. Hyper-Converged Infrastructure also follows the three-tier architecture pattern.
CI is hardware-driven, with each component able to be separated and used independently if necessary. The package of components, though, is controlled through a centralized management platform. For this reason, CI is often simpler and more cost effective to manage than traditional infrastructure. Additionally, CI makes it easier for IT departments to save money using lower-priced commodity hardware instead of proprietary or vendor-specific hardware.
Virtualization is a key enabling technology for Converged Infrastructure. Indeed, without virtualization it may be essentially impossible to have CI and HCI. The ability to set up, reconfigure and spin down VMs on demand is what makes Converged Infrastructure so efficient. Virtualization vendors like VMware are building convergence and software-defined infrastructure capabilities into their main products, applications and hypervisors, as is the case with VMware vCenter server and vSphere Enterprise. Other examples include the Microsoft Hyper-v virtual hypervisor, Windows Powershell, EMC’s EVO:RAIL and HPE’s Hyper-Converged offerings.
You can read user reviews for CI solutions from the IT Central Station community here.
Hyper-Converged Infrastructure takes convergence a step further. In this sense, “hyper” means more, as in hypersonic or hyperactive. (It also implies a smaller system, despite the general meaning of the word.) HCI is software-defined. Compute, storage and network are abstracted away from the physical hardware. An HCI system bundles virtualization software into the built-in management package and single hardware appliance. As a result, a Hyper-Converged solution resembles what users have a similar experience to what they enjoy with cloud service providers. It’s possible to add nodes, systems, virtual machines, storage and so forth without having any awareness of the underlying physical hardware. Of course, unlike the public cloud, there is a clear physical capacity limit, but the software/physical abstraction is comparable.
An HCI solution may enable functions like cloud bursting as well as disaster recovery. It can be configured to enable the management of virtual and physical infrastructure through a single interface. This is known as infrastructure federation.
You can read user reviews for HCI solutions from the IT Central Station community here.
Differences Between CI and HCI
CI and HCI overlap a great deal but there are some clear differences between the two architectures. They each deal differently with hardware, systems, compute and storage. And of course, the hyper in Hyper-Converged connotes a higher level of compactness and ease of use.
Data storage is one area where the differences between CI and HCI are pronounced. Given its software-defined approach, HCI is able to pool compute storage resources such as data storage arrays. The user does not have to be aware of whether the data storage is local, direct-attached storage, Network Attached Storage (NAS) or a Storage Area Network (SAN).
Storage innovations abound in both converged and Hyper-Converged architectures. For instance, there are a number of virtual storage area networks (vSANs or virtual SANs) available for CI and HCI solutions. A vSAN mimics the characteristics of a SAN but does not require the SAN’s usual specialized hardware or software. It’s all virtual. Many CI and HCI solutions offer data storage with inline deduplication and deduplication compression — processes that reduce the overall data footprint and leads to more efficient utilization of data storage hardware and faster data backups. HPE StoreVirtual VSA is an example of such an offering.
The two architectures also scale differently, according to most sources. CI is known as a “scale up” solution, where growth is achieved by adding CPUs, disk drives, switches and virtual machines. In contrast, Hyper-Converged Infrastructure is considered a “scale out” approach to infrastructure and storage. With HCI scaling out, one grows by adding “building blocks” of HCI capacity when its needed.
Reviewing a variety of online commentary on the scaling issue, however, reveals some ambiguity. Some vendors and architects see CI as a building block architecture that can scale out as well as scale up. Converged Infrastructure usually scales with a “Building Block” approach. With compute, storage and networking in a single chassis, it is possible to add capacity or nodes by adding more chassis. This is not always the case, however. When using a CI reference architecture instead of a vendor solution, it is possible to add needed elements, such as storage, without proportionally adding compute and networking.
CI and HCI Use Cases
When do you use CI or HCI rather than traditional infrastructure? The following are some popular use cases for CI and HCI:
To build private and hybrid clouds – The Lego-like nature of Hyper-Converged Infrastructure makes it a natural hardware basis to build a private cloud or hybrid cloud environment. The stackable blocks of compute/storage/network/VM capacity make it possible to build and expand a private cloud without excessive concern about hardware integration and infrastructure management.
This Senior Systems Administrator writes, “Having my private cloud within my Simplivity infrastructure has given me so much more than I could have ever expected. I love the ability to fire off a full, application-aware backup of a VM and have it complete in just under four seconds. Also, I can now fail my entire data center over to my DR site and have everything up an running in well under 30 minutes, with my mission-critical servers up in under 10 minutes (the servers do have to power on!). It's awesome. You do need EZ-DR by VM20/20 to accomplish this, but it is a fraction of the cost of VMware Site Recovery Manager.”
To consolidate the data center – In response to appliance and storage sprawl in cramped and costly data centers, infrastructure managers are finding CI and HCI to be an appealing consolidation solution. With a single, consolidated infrastructure, CI and HCI enable better, or even optimal utilization of resources like data center space, racks, servers and so forth. Provisioning, scaling and system changes tend to become faster with CI and HCI. Some even claim that they’ve seen drops in costs for cabling, power and cooling. There can be savings in infrastructure management software expenses and administrative overhead as well.
A Senior Systems Engineer, reviewing Flexpod writes, “It benefits the organization in that we had no downtime. In almost five years of operation, we have never had a single hour of downtime that was directly related to a storage problem. There weren't things like hard drive failures.
In any other company, it would have legitimately been an issue for us to get a hard drive out. But usually it involves some sort of extreme discussion with customer service agents about how important this is to our business operation, and there was none of that with NetApp. They adhered to the SLA.
I was willing to wait if the guy was willing to reset the hard drive. And that's more-or-less what happened. I had a failure, and within two hours of the notification of the failure, I had a new hard drive in my hands on-site. That's pretty impressive, regardless of how you put it.”
To protect data – CI and HCI make possible the centralized management of backup and restore functions. Centralization over control of data storage systems leads to consistent backup policy enforcement. It also helps with data retention and location policy compliance. For example, an IT manager can easily track whether data is stored on multiple virtual machines if that is required for data protection and integrity. Conversely, some HCI solutions facilitate de-duplication of data, which allows for better use of storage resources and an improved data lifecycle.
A virtualization system administrator in the IT Central Station community reviewed vSAN as his HCI. He writes, “It is precisely the possibility of being able to extend the capacities of the cluster of storage and calculation by the simple addition of one or more physical server that makes us lean on this solution and that in a secure way.
Moreover, with the storage policy, we were able to create different security policies depending on the virtual machines according to their needs for performance or availability.”
When reviewing HPE Simplivity HCI solution, this Head of IT writes, “We consolidated five servers and one SAN with three arrays into two OmniCubes. Our SAN was full with 20 TB at the time of migration, and we accomodate the whole amount of data into less as the half of the disk space. Since then the infrastructure grew constantly, but the data foot print on the Omnicubes barely increases. Dedup is currently around 3.3. Additionally, we implemented DR with a third OmniCube residing in a second datacenter, where we replicate our data. Recovery of VMs is done in a couple of seconds. We now almost never use our backup software to restore data anymore.”
To optimize workloads and applications - Centralized infrastructure visibility helps IT managers optimize workloads and applications. With a single management and monitoring interface, it’s possible to react quickly and easily to shifts in application load. Resources can be reallocated based on demand. It is also possible to move data from one resource pool to another, which can drive faster application performance and better resource utilization. These functions can be especially helpful with VDI where end users can be quite sensitive to relatively subtle changes in response times. In other cases, CI and HCI can be helpful for optimizing workloads and applications by enabling scale outs of resource clusters.
A DataBase Administrator at a large government organization reviewed Oracle Exalogic on IT Central Station. He writes, “It has improved the way my organization functions by migrating apps to one consolidated platform that is dedicated to WebLogic and Oracle apps.”
To enable VDI – organizations where information workers perform similar, clerical tasks are good candidates for desktop virtualization. With VDI, workers use what is essentially a terminal that replicates the functioning of a desktop PC. The PC is actually running on a virtual machine somewhere else. There are a range of benefits to this approach, including reduce maintenance, better protection against malware, lower hardware costs and so forth. HCI is well suited to the task of provisioning virtual desktops.
The Pros and Cons of Each Approach
Which one is better? That depends on many factors, of course. Neither is superior in all use cases. And, what might be considered a good feature in one scenario can be a negative in another. For instance, some CI solutions are from a single vendor, while others offer a multi-vendor capability. A single vendor CI stack could be an excellent choice for an organization that wants simplicity. If there are specific requirements best met by a multi-vendor CI stack, then that is preferable. Software licensing costs may add up, however, in a multi-vendor solution.
Industry research suggests that organizations choose CI for mission critical workloads. They like the efficiencies of CI compared to traditional infrastructure but they want to maintain a highly granular level of control over systems and data. They want the customization inherent in CI. In contrast, Hyper-Converged solutions trade off customization with simplicity.
HCI is seen as being better for infrastructure consolidation. It also tends to get favored for ease of use. The software-defined approach is also considered more flexible than either CI or traditional infrastructure. Having everything in a single appliance and a completely centralized “pane of glass” for management leads to more agility. HCI is viewed as better for agility as a result.
The ease of use in HCI, driven by its single, software-defined management toolset, is seen as being easier for IT generalists. HCI doesn’t require as much infrastructure specialization as traditional infrastructure or even CI. It’s designed for ease of use. You don’t have to be a storage or network expert to configure and manage an HCI platform. This plays well for smaller IT departments or those with hiring constraints.
IT managers who are devising private clouds also like HCI better than CI in many cases. Some HCI solutions function effectively the same as a cloud platform. The ability to add appliances quickly and easily makes it advantageous for private cloud environments. With cloud bursting and hybrid cloud capabilities, the use case is even more compelling.
It’s also important to remember that in many circumstances the best choice is “neither.” Traditional infrastructure is not going away, nor should it. It would be a mistake to think that CI and HCI are the “new things” and can therefore be your only choices for new infrastructure projects. For example, some CI and HCI solutions may be rigid when compared to their traditional counterparts. Assuming you have the expertise to configure the storage, compute and network elements the way you want them, you might find their pre-packaged natures to be restrictive.
Approaches to Realizing CI and HCI
What is the best way to implement converged or hyper Converged Infrastructure? As is the case with the pros and cons, there is no one right way to do it. That said, a number of best practices are emerging as the technologies receive a wider embrace.
Focus on the big picture – Moving forward with a converged approach to architecture is part of a bigger conversation inside IT. It’s about how to best realize the vision of virtualized infrastructure and the cloud.
Consider an incremental approach – The nature of Converged and Hyper-Converged Infrastructure lends itself to starting with small projects. A single workload, a single department can test drive the CI concept for your organization. Some applications, servers and systems are more suited to a converged approach than others. Then, based on what you learn in that experience, you can plan a bigger rollout if that is what’s needed. New CI and HCI instances can be introduced to build a “system of systems” over time.
Look at where you are currently experiencing stress – Converged solutions can be great stress relievers if they’re applied in the right areas. Where are you stressed? Where are your people having trouble keeping? Where are service level agreements falling apart? For example, if storage is a pain point for your IT organization, that might be a good place to start with a converged approach. Alternatively, if you’re short of expertise in a particular area, that might be the place to look at introducing HCI and turning the administration over to IT generalists.
Keep the business case and value in perspective – This is good advice not just for Converged Infrastructure. It’s a general principle when considering any new technology. An assessment of CI or HCI needs to answer fundamental business and value questions: Will it be worth the investment? Will it make the business operate better or more profitability? Will it save money? These are the fundamental that must be addressed.
Try to automate as much as possible – Converged and Hyper-Converged Infrastructure solutions lend themselves to infrastructure automation due to their centralized management. Automating virtual machine provisioning and data protection processes, for example, can pay off in terms of faster time to market for new systems as well as reduced administration costs.
Thinking about Converged Infrastructure versus Hyper-Converged Infrastructure takes you quickly into some pretty deep IT topics. Their very converged nature pulls in discussions of storage, compute, virtualization, cloud and more. It’s a lot to consider. Each has a distinct advantage for a given set of workloads. Neither is a cure-all or a singular replacement for other infrastructures that may be working well. Best practices are emerging to ensure a positive, cost-effective experience of deployment.