John S. Trotter

Multiple-logical-channel subsystems: increasing zSeries I/O scalability and connectivity

With the advent of the z990 multi-book multiprocessor family of server offerings, significant increases in total system capacity and scalability can be realized. Essential to an increased processing capacity is the corresponding need for significant increases in total I/O scalability and connectivity. With the z990, increased I/O capacity is provided by increasing the number of physical I/O channels that can be configured to the system and by restructuring the physical channel subsystem (CSS) into logically distinct channel subsystems. This restructuring is commonly called the multiple-channel subsystem (MCSS) facility. Each logical CSS is then assigned to one or more logical partitions as necessary in order to provide the total I/O connectivity necessary to accommodate the increased processing capacity of the system. An overview of the z990 MCSS architecture is presented with respect to how it is structured, the channel-subsystem constraints that have been removed, and how MCSS functions are provided to the operating systems executing in each of the systems logical partitions (LPARs) in a predominantly transparent manner. Also discussed is the channel-subsystem ardware and firmware (embedded software) design necessary to accommodate the MCSS architecture, as well as overviews of the MCSS I/O configuration process and the z/OS® programming support necessary to accommodate the MCSS facility. Finally, enhancements to the MCSS I/O measurement facility necessary to facilitate autonomic computing are discussed.

Enhanced I/O subsystem recovery and availability on the IBM System z9

Although part of the IBM System zTM strategy is to improve design and development processes to prevent errors from escaping to the field, improving recovery is another element in the strategy to keep a machine up and running should an error occur. The z9TM continues on an evolutionary path of enhancing I/O subsystem (IOSS) recovery to further advance the reliability, availability, and serviceability (RAS) of System z platforms. This paper presents an overview of recovery and how it interacts with other RAS functions--such as error-detection mechanisms in hardware, including automatic identification and recovery of failing elements--up to the point in time prior to the advent of the z9. It then presents the innovations to IOSS recovery and error detection in the z9 that further improve machine availability. The recovery infrastructure, which significantly reduces recovery time and makes recovery much less dependent on machine scaling for this and future generations of System z servers, is described. Also described are such innovative uses of this new infrastructure as improvements in error detection related to elusive firmware problems seen in prior machines, the ability to detect and recover from firmware hangs or lockups related to inadvertently leaving control blocks locked, and the capability to perform recovery in parallel by multiple system-assist processors.