Performance Analysis of 3D XPoint SSDs in Virtualized and Non-Virtualized Environments

Intel's Optane SSD recently came to the market as the pioneer of 3D XPoint based commercial devices. They have much lower latency (about 14 μs) and better parallelism properties than traditional SSDs, and as such are set to replace NAND flash SSD in commercial settings. To best serve cloud and enterprise data centers' higher performing storage demands, it is necessary to know the performance characteristics of the new devices in both virtualized cloud environments and traditional non-virtualized environments. In this paper, we present an analysis of Optane SSDs based on a large number of experiments. We use several micro-benchmarks to gain knowledge of Optane's basic performance metrics. We also discuss the impact of state-of-the-art storage stacks on the performance of Optane SSDs. By analyzing the test results, we provide configuration suggestions for storage I/O applications using Optane SSDs. Lastly, we evaluate the real-world performance of Optane SDDs by running MySQL database based experiments. All the experiments are performed in non-virtualized and virtualized environments (Linux and QEMU) with a comparison study between the Optane SSD and a SATA NAND flash-based SSD.

[1]  Jayanta Basak,et al.  Dynamic Provisioning of Storage Workloads , 2015, LISA.

[2]  Mrinmoy Ghosh,et al.  Performance analysis of NVMe SSDs and their implication on real world databases , 2015, SYSTOR.

[3]  Heon Young Yeom,et al.  An empirical evaluation and analysis of the performance of NVM express solid state drive , 2016, Cluster Computing.

[4]  Heon Young Yeom,et al.  An Empirical Evaluation of Enterprise and SATA-Based Transactional Solid-State Drives , 2016, 2016 IEEE 24th International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems (MASCOTS).

[5]  Cheng Huang,et al.  Rethinking erasure codes for cloud file systems: minimizing I/O for recovery and degraded reads , 2012, FAST.

[6]  Philippe Bonnet What's Up with the Storage Hierarchy? , 2017, CIDR.

[7]  Dongwoo Lee,et al.  Improving performance by bridging the semantic gap between multi-queue SSD and I/O virtualization framework , 2015, 2015 31st Symposium on Mass Storage Systems and Technologies (MSST).

[8]  Dong Li,et al.  Early Evaluation of Intel Optane Non-Volatile Memory with HPC I/O Workloads , 2017, ArXiv.

[9]  Subramanya R. Dulloor Systems and applications for persistent memory , 2015 .

[10]  Angelos Bilas,et al.  Using transparent compression to improve SSD-based I/O caches , 2010, EuroSys '10.

[11]  Philippe Bonnet,et al.  Linux block IO: introducing multi-queue SSD access on multi-core systems , 2013, SYSTOR '13.

[12]  Krishna Kant,et al.  Data center evolution: A tutorial on state of the art, issues, and challenges , 2009, Comput. Networks.

[13]  Dan Williams,et al.  Platform Storage Performance With 3D XPoint Technology , 2017, Proceedings of the IEEE.

[14]  Irfan Ahmad,et al.  Pesto: online storage performance management in virtualized datacenters , 2011, SoCC.

[15]  Hong Jiang,et al.  A Comprehensive Study of the Past, Present, and Future of Data Deduplication , 2016, Proceedings of the IEEE.

[16]  Hao Wu,et al.  High Performance and Scalable Virtual Machine Storage I/O Stack for Multicore Systems , 2017, 2017 IEEE 23rd International Conference on Parallel and Distributed Systems (ICPADS).

[17]  Gang Wang,et al.  Leveraging Page-Level Compression in MySQL - A Practice at Baidu , 2016, 2016 IEEE Trustcom/BigDataSE/ISPA.

[18]  Rina Panigrahy,et al.  Design Tradeoffs for SSD Performance , 2008, USENIX ATC.

[19]  S. Aghav Database compression techniques for performance optimization , 2010, 2010 2nd International Conference on Computer Engineering and Technology.