Technology innovations in the computer storage industry have enabled storage system manufacturers to dramatically improve storage system performance by utilizing NAND flash storage media. However, many IT organizations have been reluctant to move their most important applications onto untrusted flash storage systems, due to a lack of familiarity, and for some storage systems due to the lack of reliable enterprise class protection features required by their mission critical business applications.
Performance is important, but is a secondary consideration behind the requirement for reliable, enterprise class data protection features available with data-center class storage systems. New virtual infrastructure applications require sustained, high random I/O performance in order to deliver their promise, while still demanding highly reliable, enterprise class storage features. These seemingly contradictory demands have left many organizations unsure of their options; continue utilizing traditional enterprise storage that is unable to meet the performance requirements, or deploy unproven all-flash systems that lack the enterprise availability features necessary.
Enterprise and cloud infrastructure continue to migrate towards using virtualized applications, to increase server utilization rates. While this results in higher number of applications per server, it also drives a corresponding increase in the demands for random I/O operations to service these workloads. Database applications, along with virtual server applications and virtual desktops continue to be the top deployments for flash storage systems, due the demand these workloads place on storage systems.
In this validation, the IBM FlashSystem® V9000 was tested running two different application workloads: A virtual desktop infrastructure (VDI) workload of both linked clone deployments and fully provisioned VDI desktops, and a mixture of common virtual server infrastructure applications running a standard workload mixture. Performance testing consisted of application workloads using the IOmark tool, running both the IOmark-VDI and IOmark-VM benchmarks, with details provided in the Appendices.