11 Feb Hyperconverged Infrastructure Integration & Paradigm Shifts
One of the challenges with any infrastructure change in the data center is the process of integrating and migrating workloads. This has been one of the more talked about points with hyperconverged infrastructure. We often see that the integration of hyperconverged nodes into an existing on-premises data center being touted as one of the easiest parts of adoption. There are many steps involved in bringing new infrastructure on board, so this is a good opportunity to go over some common wins and challenges with adding a hyperconverged strategy to your data center operations.
Let’s assume that we are working with existing infrastructure. Deploying any products into a brand new, often called “green field” data center, gives much more freedom during the implementation. We have to recognize that many, if not most, organizations are running a data center today. The realistic challenge in this case is that we have to work within a set of constraints which can include space and networking configuration.
The New Kid in Town – Adding New Hyperconverged Products to your Data Center
With our traditional data center design, we may have been used to having our storage separated from our compute, and our network distributed throughout in various topologies to connect everything together in the ideal design. Hyperconvergence goes counter to the traditional strategy, with a specific goal of creating a more simplified physical and logical deployment.
Generally, we will see organizations bringing hyperconverged hardware into the same racks as their compute infrastructure. Compute nodes may be connected to legacy storage by either fibre channel, or network in the case of NFS or iSCSI. Adding the new hyperconverged hosts into the racks close to existing systems has the potential to reduce latency when communicating to other hypervisor hosts.
Migrating workloads to the new hosts will often be done using the capabilities of the hypervisor. Using tools such as VMware Storage vMotion, we are able to bring live workloads over to the new hyperconverged hosts in a safe manner will full support from both the hardware and virtualization vendors.
Now that we have multiple storage platforms with the introduction of a new hyperconverged solution along side the legacy shared storage, we will have to be careful with managing the platforms as independent resources. As new hyperconverged nodes come on line, we may choose to attach to network attached storage to migrate, or to continue to host some workloads on legacy storage. The ideal situation is to fully leverage the low-latency option of bringing the entire workload into the hyperconverged nodes.
Learning to Scale Out, Not Up
The new paradigm with hyperconvergence is the thought of scaling horizontally rather than vertically, which was the traditional way to grow your infrastructure in the past. Now that we have added hyperconverged infrastructure as a part of our data center, we also have to leverage it for what it can be best at: scaling for growth.
Traditionally, we have been scaling up with data center hardware such as storage and compute. The move towards a hyperconverged strategy opens the door to a new scale-out methodology that allows for grow-as-you-go infrastructure. Legacy vendors are making attempts to catch up, but the advantage on simple scaling still seems to be in hyperconverged platforms.
Think Globally, Act Locally – Backups and BCP/DR
We have our freshly integrated hyperconverged platforms working well within our environment. Workloads are operating nicely, and we have build our operations procedures around the new platforms with relative ease. One of the key considerations with any infrastructure component is understanding failure scenarios, and implementing backup strategies, business continuity (BCP) and disaster recovery (DR) processes.
Hyperconverged platforms have many built-in options to create metro clusters (generally < 60 mile separation), plus some have asynchronous options for larger distances. Using native BCP/DR protection capabilities will be advantageous, but has to be thought out fully to be sure it integrates with existing BCP requirements. We should be revisiting our BCP/DR strategies regularly as it is, so this is an ideal opportunity to evaluate the strategy being used, and to build out the best-of-breed option using the new capabilities that our hyperconverged platforms have introduced to us.
Backing up a Hyperconverged Data Center
There should be little to no difference in how you add hyperconverged targets to your backup processes. What is interesting with hyperconverged topologies is the density we achieve using many of the available platforms. Because of the density of the workloads on our hyperconverged nodes, the potential is there to create challenges for traditional backup strategies.
As workloads become more densely populated on less nodes, the amount of I/O and network utilization during backup times may see an increase. Luckily, most solutions are built to handle this well. Even though we are able to achieve higher density, we are also building on the inherent scale-out capability. Scaling horizontally will spread the workloads out more evenly, and at the same time this also spreads out the load for backups.
As noted earlier, snapshot capability at the VM level may be discussed as a way to protect virtual infrastructure. This should not be considered to be a backup strategy, although we may be able to use snapshots rather than utilizing lower tier storage such as off-site disk or tape media to retrieve data.
Challenges with Integrating Hyperconverged Platforms
Unfortunately, integration isn’t always as simple as we may like. Integrating any new infrastructure into our data centers has the potential to highlight physical and logical design flaws in our existing infrastructure. For this reason, many customers are faced with challenges and can mistakenly attach a difficult implementation experience to the product being brought in.
Another common challenge is the fear of running multiple hardware solutions in a data center. Some organizations will choose to aggressively migrate workloads to the new platform in order to sunset the legacy hardware for removal. This can also create a challenge for operations teams who will need to do physical and logical redesign to ensure that they are continuing to operation the ideal topology.
Hyperconverged is quickly becoming the new normal, although the adoption rates vary depending on organization size and comfort with new technology. While it may not be in as many data centers today as we may have thought, the shift is definitely underway towards seeing a strong share of the data center moving to this effective, building block deployment.