How did the development of hyperconverged infrastructure (HCI) come about? Did someone decide that storage networking and storage arrays were too complex? Did a server vendor look at SANs and decide to do things differently? As far as I’m aware, it didn’t happen either way. Future HCI vendors looked at the challenges of running a virtualized data center and decided it could be done better. Those vendors tended to focus on making it easy to run VMs. Building HCI was the result of this focus on simplicity; it was not the vendors’ initial objective. Seeing the first hyperconverged vendors succeed, a number of followers then built the same thing. I’m not sure that simply building HCI is the same as trying to make running VMs easier. HCI is not the primary goal of the leading vendors, and it’s probably not the only thing the followers need in the long term.
There is no doubt that VMs are a major feature of most enterprise data centers. There is also no doubt that running a virtualization platform can be a very complex activity. It relies on multiple specializations. In many organizations, the complexity of operating the virtualization platform reduces the benefits of having virtualization. VMs that should take five minutes to create can take days to deploy. Complex processes and coordination between multiple specialist teams reduces agility. Much of this complexity can be traced to the reuse of technology and processes that predate virtualization. The way to simplify these operations is to reimagine the infrastructure required to deploy and operate VMs.
By starting with VMs as the only workload and discarding a lot of legacy architecture, we can build a data center in which VMs are easy to deploy and operate. This design process starts with VMs and builds an infrastructure around them. Existing technologies can be reused if they do not add end user complexity. Simpler new technologies are built where the existing technologies are too complex. HCI usually involves eliminating the legacy storage silo and replacing it with a scale-out host-based storage solution. Standard hypervisor and network stacks are still used. Storage management is vastly simplified by eliminating LUN and storage network configuration. Storage management is then converged into the hypervisor management console. Network management is harder to incorporate, as the network extends far beyond the data center where the HCI resides.
However, HCI is not the end of making VMs the core of the data center and simplifying their management. There are many operational tasks that must be undertaken in the data center. Simplification needs to be applied to as many of these tasks as possible. Updates to firmware and hypervisor software are a frequent pain point. Particular hardware drivers have caused hypervisor stability and security issues in the past. Operations teams would benefit from an automated process to update the firmware and hypervisor on each node, as well as any HCI-specific code and firmware. Maintaining consistent and validated firmware, driver, and hypervisor versions is one of the key operational values of converged infrastructure like Vblock. I’ve not seen this in the follower HCI products. It is a feature of the leaders in data center simplification. Another frequent operational challenge is VM backups. This month, we have seen another VMware Changed Block Tracking (CBT) issue identified. Backups cause a great deal of pain and need to be reliable and regular. Again, the leaders in simplification usually have built-in backup methods. Most are not affected by CBT issues. In addition, backup management can be simplified by policy-based management. With policy management, there is no need to manually validate successful backups. Any VM in compliance with its policy is properly backed up. Only noncompliant VMs require examination.
The next operational task to simplify is data replication (DR) setup and ongoing maintenance. Storage array–based replication is often difficult to set up. VM-to-LUN mapping must be carefully configured and managed to keep VMs protected. The leaders in simplification have built replication into their hyperconverged platforms. They made it a part of the protection policy for VMs. The same policy that keeps local backups will also ship backups to a remote data center for DR. All of these protection policies are configured in a single user interface. Again, only noncompliance needs to be investigated.
Hyperconverged infrastructure is a step along the way to a simplified data center; it is the not the endpoint. For most organizations, the need to spend time and people on managing data center infrastructure is a necessary evil. Reducing the time and effort required will produce a direct business benefit by reducing cost. If the business can get a more agile IT infrastructure at the same time, it will be happy. HCI delivers significant benefits, but it is not the end of simplification.