Guest blog by Bret Gibbs, HPE Persistent Memory Category Manager
Letโs start with some basics. HPE Persistent Memory solves the problems of time that is wasted whenever your application has to transfer data from the memory space to non-volatile storage. Effectively, HPE Persistent Memory becomes a new tier of storage inside an HPE ProLiant server, vastly quicker than the fastest Flash drive. Conventional DRAM server memory is โvolatileโ, meaning that data is lost when the power is lost. Because of that volatility, as your application data state changes in memory, you periodically have to save that data to non-volatile storage media like HDDs or SSDs. When your data leaves the low-latency memory space, this creates longer latency and thus introduces potential storage bottlenecks. What if your data never had to leave the memory space because that memory was not volatile? We achieved that non-volatility, or persistence โ by augmenting the DRAM with Flash memory, right there on the same memory module. Keeping critical data in flash-backed memory eliminates the storage bottlenecks that happen when data leaves the memory space to be written to non-volatile media.
Our first product line in the HPE Persistent Memory product category, HPE NVDIMMs, are great with solving small storage bottlenecks โ like database transaction logs, index files and write caching tiers where you write to the low-latency DRAM on the NVDIMM. But how do you bring the performance levels of persistent memory to larger scale workloads like databases and analytics? We recently introduced HPE Scalable Persistent Memory which delivers terabyte-scale persistent memory and enables you to do much larger in-memory compute backed by flash.
With HPE Scalable Persistent Memory, we recently achieved the #1 TPC-H benchmark on the 2-socket HPE ProLiant DL380 Gen10 server running a 1TB Microsoft SQL Server 2017 โDiskless Databaseโ, achieving over 1 million queries per hour (QphH) and beating the previous record holder performance by 45%. The good people that referee this benchmark told us theyโd never seen the record broken by such a large margin! We also delivered 3x faster SQL Server 2017 analytics using Microsoft Storage Spaces Direct compared to using NVMe SSDs alone. HPE is clearly delivering unprecedented levels of performance with HPE Persistent Memory technology.
So how does HPE Persistent Memory help deliver on the vision of Memory-Driven Computing? Well โ you donโt realize your vision without steps along the journey to help you arrive to your vision. And one of those important steps is what I call software ecosystem enablement โ which is getting operating system and software applications ready to tap into the performance and efficiency potential of Memory-Driven Computing. The HPE Persistent Memory team has continually been working with operating system and application providers to bring real world innovation to the market. The #1 TPC-H benchmark and 3x faster SQL Server 2017 analytics are just two of many examples of that innovation. Another great example of getting the software ecosystem ready for Persistent Memory and Memory-Driven Computing is our partnership with VMware. In this video, VMware talks about how HPE and VMware are working together to bring the benefits of Memory-Driven Computing to the virtualized datacenter.
Memory-Driven Computing holds great promise in solving the complex problems of future โ from space exploration to data-driven cures for disease to running the homes and cities of tomorrow. HPE Persistent Memory is your on-ramp to Memory-Driven Computing and the future! To learn more, please visit our website and stay tuned for more exciting announcements in the future!
Featured articles
- Exascale computing: The Space Race of our time
- The Machine seeks adventurous developers
- Want to know the future of technology? Sign up for weekly insights and resources
