Hyperconverged Infrastructure (HCI) brings together compute, storage, and networking into a single system, reducing operational costs and the number of physical machines needed in the data center. But the main benefit of HCI is that it solves a problem related to storage scalability. As a matter of fact, HCI is often referred to as software-defined storage.
Problems With Storage
Traditional servers consist of compute (processor and memory), networking (the speed at which data can be transferred between two nodes – software is constantly talking to databases and other systems) and a tiny bit of storage. In this traditional, non-converged infrastructure, most of the storage is separate from the server. A storage array, a dedicated system containing numerous storage disks that connect to devices over a shared network, can be contacted by multiple servers at one time. Servers constantly contact the storage array to save data to and retrieve data from it. However, the storage array can only handle a limited number of conversations per second. Suppose an array has a limit of 100 conversations per second and has 50 conversations co-occurring with one server and 50 conversations with another server. In that case, the storage array has reached its communication limit. So, if one of those two servers, or another server, tries to start an additional conversation with the storage array, there will be a lag time before the array has the capacity to respond.
The performance measurement of these storage conversations is referred to as input/output operations per second (IOPS). Storage systems are built to handle a specific number of IOPS.
As companies grow to handle workloads, they will add more and more servers, which will need to contact the storage array to write and retrieve data. With the traditional storage model, when you add a server, it doesn't come with additional storage IOPS. Therefore, you further reduce the number of IOPS (conversations) available to each server. Ideally, companies would increase the IOPS as more servers are added. Still, often the array is either incapable of adding IOPS or the company needs a budget to increase IOPS. At some point, as the lag time becomes longer and longer, organizations must increase IOPS. They must either add another storage processor to an existing array or purchase a whole new one. On the low end, a new storage array typically costs between $100,000 and $150,000. A new storage processor typically costs one-third to half of that.
Converged Infrastructure vs. Hyperconverged Infrastructure Management
Since so much of a data center is now virtualized, compute, networking, and storage are all very interdependent. Each of those three resources has its own needs and each resource is typically managed by three teams, each with its own highly technical skills. Having three interdependent components managed by three completely different teams makes day-to-day operations more complex. HCI is much simpler to manage. The skills needed to manage the HCI platform are more broad than highly specialized, so it only takes one person or team to manage it.
The HCI Storage Solution
When companies use an HCI architectural infrastructure, the storage is packaged together with the networking and compute capabilities into one physical box called a node, so companies don't need a storage array. When more storage is required, organizations either buy another HCI node or add more storage to an existing node. Two or more HCI nodes that function as a single group providing storage and computer are referred to as a cluster. Built into each node is a storage controller, which communicates with the storage controller on all the other nodes in the cluster, creating a storage pool or datastore that will be seen and shared across the cluster as new HCI nodes are added. It takes only one person or team with skills that broad, rather than highly specialized to manage the entire HCI cluster. Having only one team rather than three to manage compute, networking, and storage saves much money on operational costs.
What is HCI?
HCI is a software-centric approach to storage management, which contains computer, networking, and storage in a single node, reducing the complexity of managing three different things. These three resources are packaged together in one box from a single vendor, just as resources are packaged together in a mobile phone or laptop.
In a traditional non-converged infrastructure, organizations have licensing fees for the servers, networking devices, and storage devices, creating complexities around licensing issues like Service Level Agreements and repairs. With so many different types of devices from various vendors, there may be compatibility issues. And it can be challenging to determine where the problem lies when a resource stops working correctly. For example, when IT engineers call the server vendor about a problem, they're told it's not an issue with the server but seems to be with the network. When the engineers call the network vendor about the problem, they're told the problem lies with the server. With HCI, an organization works with just one HCI vendor because the networking, compute, and storage are all bundled together.
Earlier, we spoke about the traditional infrastructure storage array, which contains purpose-built storage drives. With HCI, the storage portion of the nodes is less powerful than purpose-built storage devices, which may have twice the IOPS. However, although HCI uses generic commodity storage devices, they are configured in a way that delivers the same or better performance than the purpose-built storage drives in a storage array.
Increasing IOPS for HCI
Some HCI vendors may provide the ability to scale resources independently, meaning that you can add compute, networking, or storage as needed while other vendors may require scaling all resources in lockstep.
There are all-inclusive HCI nodes that consist of compute, networking, and storage, and there are specialty nodes of compute only or storage only. For example, you may be able to add additional storage nodes to increase storage capacity without adding more compute nodes. If you add storage or compute nodes, networking always comes with it. Similarly, some HCI solutions allow you to add additional network bandwidth by adding more network adapters or switches. Some vendors may require you to add compute, networking, and storage simultaneously, while others won't allow you to add any resources to an HCI node, which means you must add an entirely new HCI node.
From an operational risk perspective, many organizations prefer to buy entirely new HCI nodes to ensure their HCI environment functions optimally. Let me explain: If a company had only two HCI nodes and consistently added more resources to those nodes rather than buying a new all-encompassing HCI note, and if one of those two HCI nodes were to go down, the HCI environment would only be operating at 50%. By the same token, if four nodes were working together in an HCI environment and one went down, the environment would only be functioning at 75%. That's not good, but operating at 75% is better than working at 50%. It's often worth it to add another HCI node to disperse the workload rather than add specific resources to an existing node to reduce the operational risk. Generally speaking, medium-sized organizations have between 15 and 20 nodes, while enterprises have 100 or more.
Other requirements for HCI
Software must be installed on all the HCI nodes. This HCI software is responsible for virtualizing and aggregating the servers' compute, storage, and network resources, providing a unified, software-defined infrastructure. Some HCI vendors make both the hardware and the software, but some, like VMware, make only the HCI software, which needs to be installed on a server. Most HCI vendors' software can be installed on industry-standard servers from various vendors, such as Dell, HPE, and Lenovo. Once HCI software has been installed on hardware, that system becomes an HCI node. Nutanix provides the HCI hardware and software. You can buy Nutanix HCI nodes with the accompanying software built into it, or you can buy just the Nutanix HCI software and run it on validated hardware manufactured by Dell, HP, or Lenovo. Before buying HCI software, check with the vendor to find out what hardware platforms it's compatible with.
HCI solutions are more flexible than traditional data center infrastructure, allowing IT organizations to respond quickly to changing business requirements. We've already covered several reasons for flexibility in this article, but below is a recap.
- Simplified deployment and management: With HCI, all resources can be deployed and managed from a single interface. This simplified approach can make it easier to manage and deploy new workloads and move them between nodes.
- Increased agility: HCI enables organizations to rapidly provision and scale resources in response to changing business needs. Since organizations don't have to have so many separate devices because the compute, networking, and storage are all bundled together, it's easier to scale and deploy new resources as needed.
- Reduced footprint: HCI can significantly reduce the data center footprint, saving space, power, and cooling costs.
- Hybrid cloud integration: Many HCI solutions integrate with public cloud platforms, allowing organizations to move workloads between their on-premises HCI infrastructure and the cloud. Organizations can leverage the benefits of both on-premises and cloud resources.
- Simplified Management: Organizations can manage all infrastructure components, including data protection and disaster recovery features, from a single user interface, simplifying management, improving visibility, and reducing the risk of errors in data protection and disaster recovery workflows.
- Highly Available and Fault-Tolerant: HCI solutions provide built-in resiliency to infrastructure failures, helping to ensure that applications and data are available even in the event of a hardware or software failure.
Companies that deploy HCI
Several companies, including Netflix, Volkswagen, and the University of Arkansas, have successfully adopted HCI for their IT needs.
Netflix, the leading entertainment streaming service, chose to adopt HCI to improve its storage and data management capabilities. By implementing HCI, Netflix could simplify its infrastructure, reduce management overhead, and achieve greater flexibility and agility in responding to changing business needs.
The global automotive manufacturer Volkswagen implemented HCI to support its high-performance computing needs. By leveraging the flexibility and scalability of HCI, Volkswagen was able to deploy new applications and services quickly while also improving data protection and disaster recovery capabilities.
The University of Arkansas adopted HCI to modernize its IT infrastructure through VMware and improve its data management capabilities. With HCI, the university consolidated its storage and computing resources into a single platform, simplifying management and reducing operational costs.
One risk of HCI is the security architecture because it's more complex to architect your IT security. Since all components are combined into one system, a security breach in one area can compromise the entire system.
Another risk to consider is that you'll have vendor lock-in for all the infrastructure types: compute, network, and storage because they will all be bundled.
HCI provides all the elements of a traditional data center and cost savings. In the next ten years, HCI is expected to become even more widespread as it continues to evolve to meet the changing needs of businesses.
Furthermore, the increasing adoption of cloud-based solutions and the need for cost optimization in the IT environment will bolster the appeal of HCI. HCI will provide enterprises with resources that are far more agile and scalable than traditional solutions can offer, with the flexibility to adjust quickly to changing needs and conditions.
To learn how to work with HCI, start off with a course like Nutanix Enterprise Cloud Administration (ECA) or VMware vSAN: Install Configure Manage [V8], and if you’re more advanced, check out Advanced Administration & Performance Management (AAPM) or VMware vSAN: Troubleshooting [V7].