
2019 will be the year that new CPUs and motherboards supporting the fourth generation of the Peripheral Component Interconnect Express (PCIe) interface (PCIe 4.0) will come to market – meaning that new storage systems using the interface are also on the horizon. This is notable for storage managers because PCIe Gen 4 doubles the data transfer rate from the PCIe 3.0 specification, to 16 gigatransfers per second (GT/s). No other industry protocol can deliver this level of performance. Additionally, PCIe bandwidth has not increased this much since the transition from PCIe 1.0 to 2.0 in 2006, where GT/s doubled from 2.5 to 5.
The PCIe 4.0 is a long-awaited (over the better part of eight years) and much-needed industry advancement. The rise of big data analytics and the Internet of Things (IoT), as well as the ongoing boom in mobile device usage, are just a few of the industry trends that have set a new bar for required levels of throughput and data transfer rates, due to their performance requirements. One of the key bottlenecks that emerge when it comes to meeting these performance requirements is in fact the PCIe Bus. Especially with the shift to non-volatile memory express (NVMe) interfaces, the storage media itself – the solid-state drive (SSD) – now delivers such low levels of latency that PCIe 3.0 is struggling to keep up.
The industry need for PCIe 4.0 being acknowledged, it is also important for IT professionals to understand that by itself, the new interface is not guaranteed to provide the required levels of performance. At the end of the day, modern data center topologies are complex, and there are many additional factors that will impact an application’s performance. For instance, the networking standards and the software and services, such as inline compression, now in use, as well as various application and virtual machine communications and dependencies, will all impact performance. Performance bottlenecks may come from other – and oftentimes very unexpected – places.
The transition to a new technology like PCIe 4.0 or NVMe flash is difficult to justify and cannot be based on benchmarks alone. It is important to understand the impact that the technology will have on application performance, in the context of all other infrastructure and workload variables. To learn more about establishing a workload validation practice register for the upcoming webinar.