Tips for a pain-free journey to software-defined infrastructure

By some estimates, 70% of the servers in enterprise data centers are now virtualized, meaning that nearly every company is enjoying the benefits of flexibility, high utilization rates and automation that virtualization provides.

If you’re one of them, you might be tempted to move your network, storage and desktops to software-defined infrastructure (SDI) as quickly as possible. That’s a great long-term strategy. In fact, Gartner predicts that programmatic infrastructure will be a necessity for most enterprises by 2020. But you should move at your own pace and for the right reasons. Don’t rush the journey, and be aware of these common pitfalls.

Have a strategy and a plan. Think through what you want to virtualize and why you want to do it. Common reasons include improving the efficiency of equipment you already have, improving application performance or building the foundation for hybrid cloud. Knowing your objectives will give you, and your technology partner, a better fix on what to migrate and when.

Be aware that many areas of SDI are still in early-stage development and standards are incomplete or nonexistent. This makes mission-critical applications poor candidates for early migration. Start with low-risk applications and implement in phases, being aware that a full migration may take years and that some legacy assets may not be worth virtualizing all. If you’re new to SDI, consider virtualizing a small part of your infrastructure, such as firewalls or a handful of desktops, to become familiar with the process.

For all the flexibility SDI provides, it also introduces complexity. You’ll now have a virtual layer to monitor in addition to your existing physical layers. That’s not a reason to stay put, but be aware that management and troubleshooting tasks may become a bit more complex.

Map dependencies. In a perfect world, all interfaces between software and hardware would be defined logically, but we know this isn’t a perfect world. In the rush to launch or repair an application, developers may create shortcuts by specifying physical dependencies between, say, a database and storage device. These connections may fail if storage is virtualized. Understand where any such dependencies may exist and fix them before introducing a software-defined layer.

SDI requires a new approach to systems management as well. Since new devices can be introduced to the network with little or no manual intervention, it can be difficult to forecast their performance impact in advance. Be sure to factor analytics and performance management metrics into your planning so that you have a way of modeling the impact of changes before making them.

Use standards. Many SDI standards are still a work-in-progress. While most vendors do a good job of adhering to a base set of standards, they may also include proprietary extensions that could affect compatibility with third-party products. To ensure you have the greatest degree of flexibility, look for solutions that conform to standards like the Open Networking Foundation’s OpenFlow and OpenSDS for storage.

SDI relies heavily on application program interfaces for communication. Since there are no universal standards for infrastructure APIs, they are potential source of lock-in if your SDI solution requires APIs specific to a particular vendor. Look for solutions that adhere to APIs defined by industry standards instead.

Double down on security. Virtual connections create certain security vulnerabilities that don’t exist in a world where everything is physically attached. For example, the heart of a software-defined network is an SDN controller, which manages all communications between applications and network devices. If the controller is breached, the entire network is at risk, so it’s essential to choose a trusted platform with the ability to validate any new applications or components. Make sure the platforms that manage your virtual processes are locked down tight.

Don’t forget the human factor. One of the great benefits of SDI is that it enables many once-manual processes to be automated. This will impact the skill sets you need in your data center. Deep hardware knowledge will become less important than the ability to manage applications and infrastructure at a high level. Prepare your staff for this shift and be ready to retrain the people whom you believe can make the transition.

These relatively modest pitfalls shouldn’t stop you from getting your organization ready to take advantage of the many benefits of SDI. Working with an experienced partner is the best way to ensure a smooth and successful journey.

Leave a Reply

Your email address will not be published. Required fields are marked *