4 IT Pitfalls to Avoid in 2018
scale last edited by scale
Technology can be a great investment if you invest wisely. As technology changes, it is always a good idea to check if the ideas you had a year ago are still valid in the coming year. Here are a few ideas to think about in 2018 so you can avoid pitfalls like Pitfall Harry from the classic Activision game Pitfall pictured above.
Don’t buy a SAN. I repeat. Do not buy a SAN. Whether you’ve bought a SAN in the past or not, it is now a dying technology. SANs have been a staple of datacenter infrastructure for the last couple decades but technology is moving on from the SAN. A big part of this reason is the rise of flash storage and storage speeds overtaking the speeds that controller-based SAN architectures can provide.
NVMe is the new generation of flash storage and is designed to allow storage to interact directly with the CPU, bypassing controllers and storage protocols. We are entering into computing resource territory where storage is no longer the slowest resource in the compute stack and architectures will need to clear the compute path of controllers and protocols for optimal speeds.
Whether the SAN is physical or virtual, it still has controllers and protocols weighing it down. Even many new virtual or software-defined storage architectures still follow the SAN model and have virtualized the controller as a virtual storage appliance (VSA) which is a VM acting as a storage controller. You may not be ready for NVMe right now, or in 2018, but don’t let a 2018 investment in dying SAN technology keep you from moving to NVMe when you may need it in 2-3 years.
Instead, look for controller-less storage architectures like SCRIBE from Scale Computing. In testing, Scale Computing was able to achieve storage latency as low as 20 microseconds (not milliseconds) with NVMe storage. Controller-based SAN technologies could never come close to these speeds.
Going All-In on Cloud
One of the recurring themes I heard in 2017 was, “Everyone should have a cloud strategy.” Still true in 2018, but from what I saw in 2017, many interpreted this as abandoning on-prem and migrating entirely to cloud computing. There are clearly many cloud providers who could be pushing this notion of an all-in cloud strategy but the reality is that those that have already been executing cloud strategies are largely landing on some kind of hybrid cloud architecture.
The cloud is a beautiful resource and most organizations are probably already using it in one way or another, even if it may be Salesforce, Office 365, web scaling, a few VMs in AWS or Azure for dev and test, or IoT. The benefit vs. cost varies not only by service but also because how these services are used is different from business to business. It can be easy to jump into a cloud-based service without fully understanding the cost or performance characteristics and in many cases it may not be easy to escape once you’ve committed.
If you are considering cloud, it is important to evaluate the solution thoroughly for each aspect of your IT needs. Understand not only the cost but the performance capabilities vs. on-prem solutions. There are many systems, like manufacturing, that don’t easily tolerate some of the latency and even outages that can come with cloud computing. On-prem solutions for these systems that we refer to as edge computing may be a requirement.
It is very likely that a combination of on-prem solutions (like hyperconverged infrastructure) and cloud-based solutions may be the best overall strategy for your IT department. Cloud is just one more tool in the IT toolbox to provide the services your business needs.
Over-Centralizing the Datacenter
The pendulum always seems to swing back and forth between centralized datacenters and distributed datacenters. When cloud computing was becoming more mainstream, the pendulum seemed to swing toward the centralized approach. As I just discussed about cloud computing, the pendulum now seems to be swinging back away from centralization with the rise of edge computing and micro-datacenters. These on-prem solutions can provide greater availability and performance than cloud for a number of use cases.
The benefits of centralizing are attractive because it could lower operational costs by consolidating systems under one roof. However, there are far better remote management systems available these days that can also lower the operational costs of remote site infrastructure. In addition, simplified infrastructure solutions like hyperconverged infrastructure to create micro-datacenters are much easier to deploy and manage than traditional infrastructures.
As the pendulum continues to swing, we will likely see most organizations landing closer to the middle with a combination of solutions. The IT department of the near future will likely include IoT devices, micro-datacenters, cloud-based computing, and more traditional datacenter components all combined in an overall IT infrastructure strategy.
Premium (and Legacy) Hypervisors
As virtualization continues to evolve with technologies like cloud, containers, IoT, hyperconvergence, and beyond, the need for the hypervisor as a premium solution is diminishing. Hypervisor licensing became a big business with high licensing costs and those initial hypervisors did make virtualization mainstream by pulling together the traditional infrastructure components like servers and SANs. That traditional approach has now reached a plateau.
For cloud, hyperconverged infrastructure, and containers, hypervisors have become a commodity and big premium hypervisors with features you may never need are often not the best fit. Hypervisors that have been designed specifically to be lightweight and more efficient for technologies like hyperconverged infrastructure or cloud are part of a growing market trend. Traditional or legacy hypervisors that were designed to work with servers and SANs over a decade ago are not necessarily the best investment for the future.
Unlike Pitfall Harry, a misstep will most likely not get you eaten by an alligator but it may end up costing your organization in the long run. Only you know what is best for your organization but it is important to consider your strategies carefully before blowing your IT budget. The experts at Scale Computing will be happy to help you understand the benefits of hyperconverged infrastructure and datacenter modernization into 2018 and beyond. For more information contact us at [email protected].