Hyper-converged systems have gained a reputation as a plug-and-play virtualization platform. They are designed...
to make the deployment process almost effortless, but every vendor does things a little bit differently. While there is no such thing as a planning checklist that applies to every hyper-converged architecture system, there are certain aspects of the process that are applicable to most systems.
Most hyper-converged architecture vendors have requirements (or recommendations) for the number of free switch ports needed. For example, VMware requires eight 10-gigabit Ethernet ports on the top-of-rack switch for each EVO:RAIL appliance. VMware also states that link aggregation must be disabled for these switch ports. Keep in mind that as important as top-of-rack switch ports are, you need to ensure there is sufficient bandwidth to connect the hyper-converged system to the rest of your network.
One of the most critical aspects of the planning process involves establishing a plan for IP address provisioning. While your DHCP server might provide each node with one or more network addresses, there is typically more planning that must be done.
Once again, each vendor has different requirements for their hyper-converged architecture, but at the very least you will need to provision blocks of IP addresses for management traffic, monitoring, out-of-band management, clustering (virtual machine-level migration) and perhaps, storage. VMware, for example, has established the following guidelines for its EVO:RAIL systems:
- Reserve one IP address on the management virtual LAN (VLAN) for EVO:RAIL/vCenter Server
- Reserve one IP address on the management VLAN for vRealize Log Insight
- Reserve four contiguous IP addresses on the management VLAN for ESXi hosts for each appliance
- Reserve four contiguous IP addresses for virtual SAN for each appliance
- Reserve four contiguous IP addresses for vSphere vMotion for each appliance
- Reserve four IP addresses for out-of-band management for each appliance
While these requirements are VMware-specific, hyper-converged systems based on competing hypervisors, such as Microsoft's Hyper-V, tend to have similar requirements.
Not every hyper-converged architecture makes use of a VLAN, but many of them use VLANs to isolate various types of network traffic. It is important to check your vendor's deployment guide to see if VLANs are required.
If your hyper-converged architecture supports out-of-band management, and you wish to take advantage of this feature, you will either need to have a separate switch to handle out-of-band management traffic, or enough extra ports on your top-of-rack switch to provide network connectivity to each node's out-of-band management adapter.
Planning a hyper-converged architecture deployment tends to focus on network-related issues, such as cabling, VLAN usage and IP address provisioning. But don't forget to check your system's deployment guide to ensure you've met the power and cooling requirements.
Hyper-converged use cases continue to expand
Hyper-converged architecture flexibility reduces complexity
Hyper-converged products may not be for every storage environment
Hyper-converged architecture criteria you need to know
A guide to hyper-converged architecture for data storage professionals