Today’s data center has evolved into a distributed, highly efficient environment with multiple nodes. At the forefront of this change is the concept around cloud computing. Although not 100% adopted, almost every environment will have had at least some experience with a virtualization and/or cloud platform. In the recent year, conversations revolving around cloud computing have changed. We are no longer only discussing simple WAN-based communication – now there are application and desktop delivery methods to consider as well.
Engineers must carefully plan out their environments and have a thorough idea to what they are trying to deliver. This means understanding business drivers behind the actual solution. Since the concepts around cloud and even virtualization are relatively new young, there are some core considerations which must be made prior to deployment. These top 5 mistakes can be easily avoided – but can also bring an entire project to a halt.
- Pick the right hardware. The hardware environment must suite the needs of the platform. This means “future-proofing” your infrastructure to ensure the right amount of resources is always available. Too often organizations will work with a blade or rack-mount solution only to find out that they had miss-sized or chosen the wrong platform altogether. Take the time to understand the existing and future needs of your organization. From there, the decision to choose the right hardware structure can be made.
- Size your storage. Although some IT environments use local storage for their VMs, many are now using shared or pooled storage for their virtual infrastructure. Too often an environment will simply assume that their SAN is capable of handling the virtual platform. This isn’t the case. With new technologies such as VDI, there is a greater impact on the disks. This means increased IOPS requirements and the possible need for more shelves. To avoid this issue, size the environment prior to deploying the SAN. For example, if boot storms are a concern – consider deploying flash arrays or SSD drives which are capable of offloading this data from the spinning disks and onto solid state technology.
- Plan out your WAN & LAN. The local area infrastructure plays a big role in both design and performance of a virtual environment. This means using switches which have enough ports and bandwidth capabilities to handle a virtual platform. Will the data require 1GBE or 10GBE or more? Are there QoS considerations? Many times an organization will spend their budgets on servers, virtualization and other components but save their switches for last. Make sure bandwidth requirements are met prior to deploying any virtual environment. On the WAN side – bandwidth plays a big role in performance and data latency. Ensure that optimal WAN configurations are created and use WAN optimization technologies whenever possible.
- Design your workloads. This means controlling your VMs. A major issue within organizations (both large and small) is VM sprawl. This has now expanded into virtual desktop sprawl and even application sprawl. There must be a direct control mechanism to prevent unwarranted provisioning of unnecessary VMs. To avoid any kind of sprawl it’s important to have alerts set up within the environment and that only certain users have the ability to spin up new VMs. By taking the time to create the right VM initially, administrators won’t have to go back to create new secondary VMs which then take up resources.
- Sizing the environment. Resources are very finite and can be very expensive. Organizations looking to move to a virtual state must take the time to plan and size their infrastructure. This means understanding everything from bandwidth requirements to the number of VMs to be deployed. Too often, an infrastructure is built without a true vision on how resources will be allocated. This can and does result in runaway costs for the IT department. Resources should be treated carefully since pooled environments can share CPU, RAM, and storage and so on. Careful planning around what needs to be deployed, what resources it will require and how these resources will be distributed can save time and money during and after the deployment phases.
The reality with a cloud and virtual infrastructure is that every environment is unique. Organizations will have their own set of business drives which will dictate the course of the deployment. Still, engineers who are new to the virtual platform should proceed with caution and work to plan out their deployment as much as possible. This will mean involving various teams and the executive staff to ensure the IT vision is aligned.