HPE Support in the USA After End of Life

A Decision Made Once, Permanent Thereafter

When a Smart Array logical drive is created, its strip size is chosen, and from that moment it is a fixed property of the array. It cannot be adjusted later. Changing it requires destroying the array and rebuilding it, which on a production system means a full restore from backup and a planned outage. US operators asking how to tune an array that has become slow are usually asking about the one parameter that is no longer available to them.

What the Geometry Actually Controls

Strip size determines how each write is spread across the member disks. When it is much smaller than the typical write the workload issues, every operation involves several disks and the array does far more physical work than the logical volume suggests. When it is much larger, a small write triggers a read-modify-write cycle: read the existing data, merge the change, write it back. Either mismatch appears as latency and disk utilisation out of proportion to the workload.

Why it is usually wrong by accident

The value was almost always a default accepted during commissioning, appropriate to what the machine was for at the time. Machines outlive their original purpose. A server built for file storage acquires a database; one sized for sequential backup starts serving random reads. Nobody revisits the array geometry because nothing fails, and the storage continues to deliver correct data at a cost nobody is measuring.

What this means for spending money

It changes the return on a drive upgrade. Where the geometry is working against the access pattern, faster disks recover a fraction of what an operator expects, because the extra physical operations remain. The honest sequence is to read the array geometry first, compare it against the actual write pattern, and only then decide whether the money belongs in drives or in a rebuild.

What ICD Supplies for HPE Estates in the USA

Smart Array controllers with the cache and cache-protection hardware their generation depends on, drives matched to the existing array members, ProLiant chassis components, ECC memory in documented pairs, and hot-swap supplies matched to the installed load. Where we think the geometry rather than the hardware is the constraint we will say so. The United States sits in the North America band we publish, giving a 5 to 7 day door-to-door lane from Cairo. Each controller and drive is tested as a pair before dispatch, serial-logged, with one to three years of cover.