Flash storage is integrated on the zEC12 as PCI Express (PCIe) attached RAID 10 cards which plug as cable connected pairs in the I/O expansion drawer.
Initially the main application of Flash Express is as an extension to main memory in z/OS where it is integrated within the memory hierarchy to provide increased system availability and resiliency. There are several examples of how Flash Express on System z as exploited in z/OS provides for additional system availability and resiliency that is transparent to applications and requires no application changes. With Flash Express the system can reduce workload and response time delays due increased memory pressure during periods of peak demand, transitions between different workloads, or during collection of diagnostics. For example when the workload on a customer’s system shifts from a transactional workload during a prime shift to a batch workload and then again back to a transactional workload during the next prime shift, response time delays can occur. These delays can be drastically reduced when data for the next prime shift needs to be transferred from disk backing store into main memory. The same occurs when data is needed to be transferred into main memory as part of a diagnostic dump. Flash Express’ fast IO rates and low IO latency provide for decreased first failure data capture time, faster page-ins, of critical work, and allow the system to return to normal workload operations a lot faster. In z/OS flash memory has been integrated within the memory hierarchy to provide higher levels of system availability and speed. As Databases and JVMs are consuming larger amounts of memory in order to provide better transactional response times, lowering the OS memory management cost can only help improve overall response times. The very fast random access and higher IOPs for reads in flash memory relative to disk drives, have enabled z/OS to provide the support for Pageable Large (1MB) Pages. Managing memory in 1MB granularity vs. 4KB provides application performance benefits by reducing memory management costs. In today’s data intensive applications with large access to buffers, exploiting Large Pages allows for transferring larger chunks of data at faster speeds between main memory and flash. That can translate to better performance for DB2, Java or other analytic workloads. Elpida Tzortzatos elp...@us.ibm.com ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
