In previous articles, I suggested that hyperconverged is just a step on a path to simpler IT infrastructure. I also explored how some of the simplification might work. Today, I’d like to explore some of the areas of infrastructure that are ripe for simplification. Some of these areas are already being addressed by some vendors, but no single vendor is addressing every area. I expect that over time, we will see more features become common on all HCI platforms. I also expect that many vendors will retain their own differentiating feature to avoid straight price competition.
One of the ways current HCI vendors can identify areas to simplify is to look at technologies where they currently partner. I expect to see a lot of these partnerships in-sourced and integrated into hyperconverged offerings. Some of these simplifications are most valuable to small IT organizations, while others are only essential in larger organizations.
One area of policy is hypervisor configuration, often a largely manual activity. This is the opportunity I mentioned in my last article: satisfying the prerequisites for VMware DRS. The optimal configuration of a group of hypervisor hosts should be a matter of policy. A cluster should have a configuration policy, and the cluster should ensure every host is compliant. This should cover all host configuration. It should start with hardware drivers and hypervisor versions, including both minor updates and major version upgrades. Policy should drive all the way through network configuration, time synchronization, and all other advanced settings. New hosts added to the cluster should be made compliant before workloads (VMs) are permitted. Workloads should be automatically evacuated from noncompliant hosts. Features like the automated updates in both Nutanix and SimpliVity are a start. The upgrades need to be triggered by a policy change, rather than as an ad hoc activity.
Some fault remediation can be seen in Scale Computing’s platform. There is a built-in expert system that identifies misconfigurations or incompatibilities and resolves known issues automatically. Scale Computing is continuously updating this expert system based on support calls received from customers. The updates are pushed to every Scale Computing cluster automatically, so newly discovered faults can be automatically remediated. The aim is that only one Scale customer should need to report a fault. The resolutions are automatically applied to all other affected customers. By the way, this is great for keeping your customer support team small and customer satisfaction high.
Another routine operational area is backup: providing certainty that a VM can be restored to a past state or that single files can be restored when users do something silly. This backup should be built into the platform and managed by policy, using the same UI as every other VM management task. This is definitely an area where the newer hyperconverged players are partnering. The challenge is that bundling a backup product without a unified UI does not bring much simplification. I expect to see more HCI vendors with backup integrated. The leaders designed their products from the start to include VM backup; backup policy is built into their user interface.
VM replication for DR is another common pain point for IT. Building replication into the platform is a great start. Making the replication more data efficient is a great help for customers who must pay every month for WAN capacity. Most HCI vendors are efficient in replicating only changed blocks and allowing per-VM replication. Currently, SimpliVity has the strongest story here, as its replication is deduplicated, with only unique data ever replicated between data centers. Backup and replication should both be elements of a VM’s data protection policy, and again, the policy should have the ability to be applied to multiple VMs.
There are other directions for simplification. One is to simplify the VM guest management. The same depth of patching and configuration management that we want for hypervisors could be applied to VM guests. This would suit small businesses where every server workload is a VM on an HCI platform. It is less likely to suit enterprise organizations that have a little bit of everything.
Network management is another area for simplification. Having matching physical and virtual switch configuration would remove a whole class of faults. This should again be driven by policy applied to groups of hosts, or even VMs. The same policy should update both the virtual switch configuration and the physical switch. It could even drive configuration to the edge of the data center network and automatically configure routing. This is likely to be almost complete network configuration for small businesses. Larger organizations will likely have additional networking requirements, which I’m sure would benefit from policy-based management.
The current crop of hyperconverged systems is just the beginning of the simplification of IT. We will see many more systems that simplify management and make on-premises data centers easier to manage. There is no choice here: if on-premises does not get simpler, then the business is likely to move more applications to cloud providers.