As we discussed in a recent entry “The Flash Performance Euphoria is OVER” the next generation of flash solutions will be compared to the performance of the first generation of flash solutions not hard disk drives. Users have learned that IOPS is not the right measurement of performance, latency is and they want next generation flash solutions to lower latency while continuing to lower cost per GB. In a poll taken during our webinar “How NVMe Will Change Flash Storage” over 60% of the respondents indicated that either latency or costs were their top concerns. IT professionals clearly understand that latency is the real bottleneck to performance and that the cost of flash capacity needs to keep decreasing.
The first step in addressing flash latency problems is to improve the software stack that the operating system and application must traverse in order to communicate with the flash device. The second is to improve the network the flash device is on, but a faster interface accentuates the need for an improved software stack. The problem is that these two steps are in conflict with each other. Improving the software stack has led to vendor created, proprietary, operating system specific drivers that only work on that vendor’s flash devices.
The solution is initiatives like Non-Volatile Memory Express (NVMe) where a documented and universal driver is developed that flash device manufacturers adhere to. As we detail in our article, “What is NVMe? And what does it mean for PCIe-SSD?”, NVMe is a specification for accessing solid-state drives (SSDs) attached through the PCIe Express interface. The combination of a fast hardware “network” and streamlined drivers designed for flash media meets the requirement for reduced latency. The openness of the standard means these solutions remain cost effective.
Click here to learn more about NVMe and how it will change flash storage and watch our webinar. It is is now available OnDemand.