Parallel Query Server makes its debut
The new IBM S/390 Parallel Query Server is a specialised system for processing queries against large relational databases those storing tens of Gigabytes of data and upwards. It accepts dynamic read-only Structured Query Language queries from client programs, including existing decision support tools and applications – using technology licensed from Information Builders Inc, although IBM seems loath to tell anybody about that. Once again, parallel processing is used to reduce query response times and cost – making it practical to extract data from tables with millions of rows of data, says the company. The system is essentially an open database server supporting an open application programming interface via a dynamic SQL interface. This provides open access to a wide range of client query tools.
Users can access the server from any machine that supports Systems Network Architecture and Transmission Control Protocol/Internet Protocol – TCP/IP – protocols and interfaces to the front-end DB2 database on the host mainframe. The S/390 Parallel Query Server is attached to a S/390 front-end system via Escon channel-to-channel connection. S/390 Parallel Query Server gives the performance improvements of parallel processing at a lower price, the company claims. Large table scans that use lengthy sequential processing are split into smaller queries, which are then executed in parallel on the new S/390 Parallel Query Server. Faster response times and the use of lower-cost microprocessor-based hardware are claimed to make this a particularly cost-effective system – but as it will be an incremental purchase rather than an upgrade for most users, the market for it is likely to be limited in these straitened times – but we’ll know soon enough: the machine should be available in the third quarter.
New version of MVS/ESA Sysplex at heart of parallel systems
As you’d expect, IBM Corp has launched a new version of the MVS/ESA Sysplex, first announced in September 1990, which takes advantage of the parallel processing of the S/390 Parallel Sysplex. The parallel Sysplex of MVS/ESA SP 5.1 supports up to 32 systems and is designed to improve communication and data sharing among those systems. Both MVS/ESA SP-JES2 5.1 and MVS/ESA SP-JES3 5.1.1 exploit the Sysplex capabilities, and a Sysplex configuration is required for a JES2 Multi-access Spool or a JES3 complex. An important element of the system is the coupling facility, which enables high performance data sharing between authorised applications such as MVS subsystems. These subsystems and components then extend the benefits of data sharing and recovery to their applications. Authorised applications, such as subsystems and MVS components can use the coupling facility services to cache data, share queues and status, and access Sysplex lock structures in order to implement high performance data sharing and rapid recovery from failures. The subsystems and components transparently provide the data sharing and recovery benefits to their applications. There’s no single system creating a bottleneck; the data doesn’t need to be partitioned or re-partitioned when adding another system. Use of a coupling facility by software subsystems, such as Information Management System Database Manager – IMS DB, ensures integrity and consistency of data. This makes the Sysplex ideal, says IBM, for transaction processing applications. The technology that makes this data sharing possible is a combination of hardware and software services available in MVS/ESA SP 5.1. High bandwidth fibre optic links called coupling links provide high speed links between the coupling facility and systems connected to it. The initial IBM data management system using the coupling facility is IMS/ESA Version 5 DB. IMS/ESA 5 uses the MVS XCF Cross System Coupling Facility to provide cross MVS image OTMA, Open Transaction Manager Access, in the Sysplex. IMS OTMA together with MVS XCF provides an interface for host-based communications servers to access IMS applications.
access to the checkpoint
MQ, TCP/IP, RPC and IMS systems can be used across their appropriate systems in conjunction with each other, the idea being to take advantage of existing application and data skills. CICS uses the MVS Cross System Coupling Facility to provide cross MVS image multi-region operation for communication with other CICS systems in both the same and different MVS images, in the same Sysplex. JES2 enables installations to define checkpoint data sets on both coupling facility structures and disk. Placing the checkpoint on a coupling facility provides for more equitable access to the checkpoint by all multi-access spool members, enabling a mixture of S/390 microprocessors, 9121 511-based or 9021 711-based processors in the same multi-access spool. Replacing the JES3 CTC support with XCF communications enables an installation to reduce their hardware investment to run a JES3 complex by sharing hardware resources with other components that also use XCF signalling. The use of Escon CTCs allows for improved monitoring of the hardware and greater distance between processors in a Sysplex. This support enables the installation to run multiple communication paths between processors concurrently and thus reduce the potential for single failures disrupting operations. Communication paths between processors can be added dynamically and communication hardware can be changed without having to change the JES3 initialisation stream. In addition, the requirement to define CPU IDs to JES3 has been removed. This means that a JES3 complex-wide warm start is no longer necessary when you want to change the CPU ID on a processor. To help achieve balanced workloads in a Sysplex, Workload Manager, a component of MVS/ESA, provides goal-oriented dynamic resource balancing. With workload management, an installation defines performance goals for workloads and assigns a business importance to the workload. Workload management focuses on attaining those performance goals through dynamic resource distribution. This type of workload management is a change from the way workloads were previously managed. Prior to MVS Workload Manager, an installation defined how resources were to be distributed in anticipation of how workloads would use them. With workload management, the installation defines performance goals – in the same sort of business terminology used for defining service level agreements – and the system adapts to meet the goals. New functions have been added to the MVS base code to support additional open interfaces. OpenEdition integrates open interfaces into MVS/ESA SP 5.1 by building on existing MVS functions. Thus DCE and Posix support ship as components of MVS. Integrated sockets support, along with TCP/IP modifications coupled with DCE and Posix support, make it easier to convert Unix applications for MVS/ESA SP 5.1, which ships in June, with some features held for September.