Integrating NAND Flash Devices onto Servers.ppt

Integrating NANDFlash Devices onto Servers By David Roberts, Taeho Kgil, and Trevor Mudge Presented by Shuhan Xu Outline Warm-Up Architecture Modification Experiment Conclusion Is flash memory a powerful device??? Application Domain Challenge Can we use it in a bigger field? Data Center Who dominate data centers? What problem do data centers have now? Dominators DRAM Disk DRAM+Disk Budget Comparsion Calculation Power consumed by DRAM in xxx data center: 0.016 * 23 *350 *100000 *0.1 $ Calculation Power consumed by DRAM in xxx data center: 0.016 * 23 *350 *100000 *0.1 $ = 1,288,000$ Question Time Any? Architecture Physically, all devices share a common data bus. Architecture 1. Extended system memory usage model: A NAND Flash memory module is connected to the current system memory interface or to a dedicated Flash memory interface. 2. Storage accelerator usage model: A NAND Flash PCI express card is connected to the PCI express interface. 3. Alternative storage device usage model: A Solid State Drive (SSD) replaces or augments the hard disk drive. It is connected to the disk interface. An example would be a SATA SSD. Architecture Proposed Architecture Two modifications Splitting Flash-based disk caches into read and write regions. In this way, overall performance and reliability can be improved. The programmable Flash memory controller can improve Flash cell reliability and extend memory lifetime. Operating System Support Our proposed architecture requires additional data structures to manage the Flash blocks and pages. These tables are read from the hard disk drive and stored in DRAM at run-time to reduce access latency. Operating System Support FlashCache Hash Table allows the operating system to quickly look up the location of a file page and mitigate wear-out. Flash Page Status Table keeps track of the ECC strength, MLC/SLC mode and access frequency for each page. Block Status Table determines how worn out it is. Global Status Table record