Deploying a shiny new HCI platform can transform a data center, but there are a few preparatory steps IT administrators...
should take to ensure a smooth transformation.
Some of the preparations for installing hyper-converged infrastructure (HCI) are fairly standard; administrators who have switched virtualization platforms before will be familiar with them. Other steps are more specific to HCI. Every HCI deployment needs physical data center preparation, and most involve changes in organizational responsibilities.
For many admins, the largest task is preparing their existing virtual machine (VM) fleet for migration. Getting existing VMs in order helps smooth the migration process, particularly because installing HCI involves moving VMs from one platform to another.
First, administrators should look over existing VMs and identify any that have snapshots in place. Production VMs should only have snapshots during backups. Admins should also take a look at every powered off VM and determine which ones to keep. Finally, admins should scan every data store for VMs and disk files that are not registered in management tools. Sometimes, VMs are removed from inventory, or disks are removed from VMs, but they are not deleted from the underlying storage. These clean ups should be a part of regular monthly health checks on the environment, but are easy to forget.
Deploying HCI involves migrating VMs from older storage and servers to the HCI platform. Each VM needs two migrations, storage and compute, which admins can accomplish using tools such as VMware vMotion. Ideally, the storage and compute migrations are two separate steps. VMware vMotion requires both platforms see the same storage. Storage migration requires one platform sees both the old and HCI storage.
Having VMs on storage that is accessible to both the outgoing servers and the new HCI stack makes migrations far easier. If the existing storage is Ethernet based (iSCSI or NFS) then it should be directly accessible from the HCI, but few HCI platforms have any option for Fibre Channel (FC) networks. Bringing the FC data store into HCI is seldom an option. Some HCI systems also allow the HCI data store to be accessed by existing servers, which lets admins migrate FC-based VMs to the HCI. The least preferred option is to use a concurrent storage and compute migration. With so much in flight at once, the consequence of failure is a bigger mess.
Administrators must also prepare the physical infrastructure for the HCI platform. There should be space in racks for the new nodes. Make sure there is sufficient power, and that air flows to provide adequate cooling.
Most HCI products have higher physical density than the platforms they replace. The power supply that used to feed half a rack of 2U servers may only be sufficient for five HCI blocks. If the HCI nodes consume the same amount of power, then they will generate the same amount of heat in the smaller space. Make sure there aren't huge airflow gaps in the racks that would allow warm air to circulate back to the front of the HCI.
Also, make sure there are adequate network ports. Most HCI platforms need two 10 GbE ports to each node, but many put four nodes in 2U of rack space. Most configurations will need even more 1 GbE ports. It's usually three per node: two for VM traffic, and a third for hardware management. The new hyper-converged infrastructure must be in place and operational before the old platform can be removed.
The reason many companies deploy HCI is to simplify VM storage management. HCI does not involve managing RAID and LUNs. Usually, HCI has large pools of storage and VM storage policies. This is a significant shift in storage management, and may remove the need for the storage team to be involved in managing VMs.
There are likely to be some unintended consequences. For example, say there's a software testing process that involves the storage team cloning a LUN full of production VMs and presenting it to the QA team servers. Without LUNs and the storage array, this process will probably change to cloning groups of VMs.
Most HCI products also have backup and replication technologies, so companies might need to review their backup and DR protection. It may be that existing backup processes can be reduced or retired when the HCI tools are in place.
The backup team will need to be involved in planning and operating these protections. Just as with storage, VM-based policies for backup will replace legacy concepts of jobs and schedules. But operational and organizational changes take time; the existing backup may remain in place permanently, or take months to be replaced with the HCI backup.
HCI market guide for IT buyers
Get the HCI management job you want
How to pick the right hyper-converged infrastructure appliance