How hyper-converged systems change IT management

kantver - Fotolia

To reap the benefits of hyper-convergence, manage it right

Use a policy-based approach to manage hyper-converged infrastructure.

To gain the greatest value from your investment in hyper-converged infrastructure, it's critical to embrace the simplification it brings to data center operations. That simplified management of the data center, after all, ranks among the top benefits of hyper-convergence, and it would be unwise to replicate the constraints of your previous system's architecture.

If your organization is interested in hyper-converged infrastructure (HCI), it's smart to give some thought to how you'll manage it once it's implemented. You'll want to be sure this powerful new technology lives up to expectations. A well-integrated infrastructure product can make a data center simpler to operate, yes, but that doesn't mean your IT team won't have to manage the HCI to be sure it does all that it is capable of doing.

The operational model for HCI is based on VM-centered and policy-based management. Large pools of resources are controlled through these policies. Only policy exceptions need attention.

It is also important to remember that HCI is not a silver bullet for all of your virtualization management. No matter what the sales rep told you, you will still manage your hosts, VMs and networks.

Policies and pools

A policy is a set of controls you want to apply to a group of VMs. It might be a performance policy that guarantees certain resource levels, or it could be a data protection policy that mandates local backups and remote replication. You define the policy once and then apply it to a set of VMs. Each VM might have multiple policies attached to govern things such as performance, availability and disaster recovery.

Once the policy is applied, you do not need to check each VM to see that it is compliant; you'll be alerted when there is a policy breach. As an example, if you have a policy that requires a virtual machine to reside on your fastest tier of storage, you will be notified if someone migrates that VM to a slower tier.

Some HCI platforms have their own versions of the hypervisor cloning and templating functions. These tend to be more efficient than the hypervisor tools, so using them can save time.

Creating these policies requires some time and effort at the outset -- you will want to be sure you lock in the right policies for your particular needs. Once these standard policies are created, your VMs should be allocated a policy. You do not want to have a new policy created for every new VM. Policies should be standardized in the same way that cloud services are standardized, as a menu from which an option is selected.

Among the hyper-convergence benefits is how an HCI product frees you from having to do certain things. Staff simply won't need to commit the same amount of time to data center tasks, and operations will be less complicated.

With most HCI installations, you'll create fewer clusters and fewer data stores. This results in fewer choices to make about where a new VM or application should reside. Data protection and storage performance policies are applied to VMs, not inherited from a data store or cluster. There will seldom be a need for more than one data store. With fewer, and larger, pools, we can share resource headroom and run at higher total utilization.

So, what's managed?

The virtualization team manages the HCI platform. There is no requirement to open tickets for storage administrators to make new logical unit numbers (LUNs) available for added VMs. The virtualization admins are the ones responsible for making sure storage resources perform for the applications in the VMs. Most often, this is a simple matter of assigning a storage policy and configuring the virtual machine.

Your IT team will be responsible for capacity planning on the storage. Often, all of the hyper-converged infrastructure's data stores will share the same underlying physical storage. You'll need to use the HCI platform to manage the capacity, as the hypervisor seldom understands the HCI capacity efficiency.

HCI enables simplification. One important example of this is the elimination of the dedicated storage network. All of the storage traffic goes over Ethernet, usually 10 gigabit Ethernet (GbE). In many HCI deployments, the 10 GbE will also carry management traffic and VM network traffic.

Managing the configuration and load on this network can be an important part, ensuring the delivery of applications on the HCI platform. You will still work with the network team to get virtual LANs provisioned and to troubleshoot any networking issues. In many HCI products, the Ethernet cards are the only expansion cards; there is no option to add any other networking.

Most hyper-converged infrastructure products plug into your existing hypervisor management, and all of the usual VM management remains. The skills you learned about configuring and managing virtual machines remain almost entirely unchanged with HCI. You still configure VMs with the right resources and configurations for the workloads. Some HCI platforms have their own versions of the hypervisor cloning and templating functions. These tend to be more efficient than the hypervisor tools, so using them can save time.

A few HCI offerings bring their own hypervisor. That means you won't plug into your existing hypervisor, and you will use the HCI vendor's interface for all management. There will be a learning curve to build familiarity with this new interface.

With this approach, the HCI vendor can simplify and integrate more deeply. That's one of the benefits of hyper-convergence.

What to avoid

Options in the hyper-converged market are plentiful, so you'll have a lot to consider when selecting a product. Just be sure to think about how you'll actually implement that system.

When adopting HCI, it's important to leave behind the baggage from your old infrastructure. Fresh design decisions should be made to match the capabilities of your chosen HCI product.

As an example, there's no reason to duplicate the data store design from your old storage environment onto HCI. On older storage, a LUN was a boundary for performance and for replication. There were high-performing LUNs and lower-cost LUNs. Some were snapshotted and replicated for data protection. VMs would be placed on these LUNs to satisfy the availability and performance requirements of their applications.

With HCI, availability and performance are policies applied to the VM. In many cases, a single data store will fulfil the requirements for a diverse set of application needs.

Next Steps

Hyper-convergence brings scalability and software-defined capabilities

Weighing the costs and benefits of a converged infrastructure product

With so many options available, it can be difficult to select an HCI appliance

Dig Deeper on Converged infrastructure (CI)