Abstract: A system and method for providing increased availability of input/output (I/O) drawers during concurrent I/O hub repair are provided. The illustrative embodiments provide an alternative cabling scheme between an I/O planar and a plurality of I/O hubs that facilitates the I/O drawers being available even when an I/O hub fails. With this cabling scheme, a hypervisor or other virtualization mechanism configures routing tables in the firmware of the host system to cause data to be sent/received through the I/O hubs. In the event of a failure of an I/O hub, the routing tables may be updated by the hypervisor to utilize a different I/O hub coupled to the I/O drawer. By virtue of the modified cabling scheme of the illustrative embodiments, such updating of the routing tables allows I/O operations to continue at a single loop, or single barrel, throughput via the other I/O hub.

Abstract: A system and method are provided for performing capacity upgrade on-demand operations for input/output (I/O) adapters of a data processing device. The system and method involve providing a data processing device with additional I/O adapters in excess of current I/O capacity requirements of an owner/user of the data processing device. These additional I/O adapters remain in an inactive state after booting of the data processing device until they are specifically activated by the owner/user. Although inactive, memory resources are reserved for the address translation data structures for these inactive I/O adapters. The owner/user may, at a later time, obtain an activation code from a provider of the data processing device and enter the activation code into a hardware management console to thereby activate the additional I/O adapters. The reserved memory resources are then utilized to initialize the address translation data structures for these now active I/O adapters.