The Pillar Data AX600 is not unusual physically, with three components, a management unit (called a Pilot), a controller head (the Slammer), and drive shelves (Bricks), but Pillar Data's approach to storage provisioning and management is definitely unusual. The company calls the approach "application aware storage," which means that rather than specifying a volume that is RAID5 with a 64KB stripe and scheduled snapshots, you specify an application type, Oracle Database or SQL Server or VMware ESX hypervisor, for example, and the system sets up the volume to optimise the performance for that app. A variety of application profiles are included with the system, and more specialised ones are available from the Pillar Data support site.

Rather than specifying a storage tier (SSD, 15K, 10K, SATA), you specify a performance class (premium, high, medium, low, or archive), and the system picks the best storage on which to place the volume, based on what's available. These storage options not only include the speed of the disks the volume is on, but also the place on disk (outer, middle, or inner tracks) and CPU utilisation, amount of cache, and allocated bandwidth. The system even includes a predictive analysis module that can analyse the impact on existing storage volumes of a new volume, depending on the performance requirements and application type it is given.

The Pillar Data system was the most expensive of the ones I tested, but it also had the highest capacity, multiple paths throughout for high availability, a very complete feature set, and great usability. Coalescing multiple snapshots into a new image went very quickly, as did mounting the new volume. The system would easily support multiple VMware servers without any storage bottlenecks.

OUR VERDICT

The Pillar Data system was the most expensive of the ones I tested, but it also had the highest capacity, multiple paths throughout for high availability, a very complete feature set, and great usability. Coalescing multiple snapshots into a new image went very quickly, as did mounting the new volume. The system would easily support multiple VMware servers without any storage bottlenecks.