It’s that time of year again. As we enter 2019, let’s take a moment to slow down, look ahead, and predict some of the IT- and data storage-related trends that will emerge the next 12 months.
If we think about how the role of data has evolved in recent years, how data storage infrastructure has responded, and examine its trajectory, there are a few trends that we should expect to continue almost with certainty:
- Public cloud adoption will continue, even accelerate.
- All-flash adoption will continue with increased deployments of both NVMe and NVMe over Fabrics technologies.
- Business success will become even more dependent upon IT services and capabilities, with analytics, AI, and IoT all expected to grow substantially.
Those three trends are not my predictions for 2019; rather, they are my expectations. But these serve as the backdrop to my three 2019 predictions.
- I expect more companies to bring a workload back from the public cloud. Last year, we identified that 41% of businesses brought at least one workload back from public cloud infrastructure to be run on premises. Those of you who have already read some of my analysis of our follow-up study into these businesses will know that these moves are not an indictment of public cloud services; the cloud offers a wealth of benefits. Rather, these moves are the result of a lack of necessary upfront due diligence prior to cloud adoption, such as not modifying the workload or not understanding the specific performance or sensitivity requirements. With cloud adoption poised to accelerate even faster, more cloud excitement unfortunately often translates into less firms doing the upfront due diligence, which means even more costly missteps. While overtime organizations should become more adept at workloads analysis and modification for the cloud, reducing the number of workloads that have to come back, I expect it to get worse before it gets better.
- 2019 will be the year of NVMe over Fabrics, and I expect NVMe over Fibre Channel adoption growth to outpace NVME over Ethernet starting this year. So, why not Ethernet? Some in the industry have been predicting a shift to Ethernet for storage networking for some time it seems. Additionally, some of the RDMA-based protocols, such as RoCE, look to have a head start. The answer is ease. Many of the leading Fibre Channel vendors have already integrated NVMe over Fabrics support into their existing platforms. I also expect growing interest in NVMe over TCP to complicate NVMe over Ethernet decisions. For example, do I go with an RDMA-based protocol, such as RoCE, and invest in new interconnects? Or do I go with NVMe over TCP? What are the performance impacts? Can we even assess the performance impacts today? I expect questions such as these to slow NVMe over Ethernet adoption relative to NVMe over Fibre Channel. Of course, I also expect many of the ecosystem support components, such as OS support, hypervisor support, and multi-pathing, to be resolved for NVMe over FC shortly. If that doesn’t happen, I will have to reassess.
- A significant percentage of analytics, AI, and IoT projects will miss their goals (maybe even fail) due to lack of an adequate metadata handling. Artificial intelligence is hot right now. I expect many of the technology predictions for 2019 to involve AI in one way or another. Here is the thing about AI: It requires data, and the greater the volume of data, the more successful the project. File data has been growing for decades, during all those years of growth and storage, and few if anyone thought about AI. We just stored it where it was safe and cheap. As a result, modern storage infrastructures have become increasingly disaggregated, especially with the rise of the public cloud and the edge. Finding all this data now takes too long and this lost time diminishes the value of analytics and AI initiatives. The solution to this problem has long been thought to be hidden in metadata handling, whether with consolidation, acceleration, or the introduction of custom metadata, but the ideal solution has been elusive so far. For the smaller environments, this not may be a huge issue. For companies betting big on AI, IoT, or even analytics, though, solving how to quickly identify and isolate the right data, whether with metadata of some other solution, will become a top priority.
2019 should be an exciting year for IT. And I am looking forward to what the new year brings. Let me know what you think! Do you agree? Disagree? Start a discussion.