【NEC】 ACOS-4/XVP PX

The ACOS-4/XVP PX (which stood for “extended multi virtual processor, parallel extended”) was an operating system intended for NEC’s mid-range and large mainframes. NEC announced the operating system in July 1994 and started shipments in October 1994.

ACOS-4/XVP PX offered connectivity with open systems, realized large cost savings by adopting cutting-edge hardware technologies, and lengthened its advantage over other operating systems through various mainframe architecture revisions, such as parallel processing and cluster technologies. ACOS-4/XVP PX ran on Parallel ACOS mainframe series models, starting with the PX7500, which was announced in July 1994, and followed by the PX7800, PX7600SV, and PX7800SV.

The main features of the operating system and its new architecture are given below.

1. Parallel processing
ACOS-4/XVP PX provided a new information processing platform by adding parallel processes for online transaction processing (OLTP) — important in the service processing area — batch jobs, and database processing.

(1) Division of transaction loads
ACOS-4/XVP PX provided parallel OLTP for cluster systems consisting of multiple hosts. Parallel OLTP raised the system’s overall transaction throughput by allocating and parallel processing transactions among multiple hosts while keeping the load balanced.
To split transaction loads, parallel OLTP used a load balancing controller, to allocate transactions equally among multiple hosts, and a global transaction queue, a transaction queue that is shared between hosts. Parallel OLTP was also fault tolerant; if a host failed while processing a transaction, the system would recover and re-execute the transaction on another host.

Figure 1: Parallel OLTP

Figure 1: Parallel OLTP

(2) Load division and parallel execution of jobs
To efficiently execute large batch jobs in a cluster system, ACOS-4/XVP PX provided a parallel batch function and a parallel job step function, which ran job steps in parallel.
The parallel batch function improved turnaround times in cluster systems that shared service databases and files using a global batch scheduler. The global batch scheduler monitored the loads on each host and scheduled the next batch job to run on the host with the lowest load.
The parallel job step function improved turnaround times by running job steps in parallel using synchronous temporary files. When data is passed from one step to a subsequent step within a job, the synchronous temporary file allows the subsequent step to start inputting data records before the previous job step has completed writing all the data records.

Figure 2: Parallel job step

Figure 2: Parallel job step

(3) Parallel SQL
ACOS-4/XVP PX provided a parallel SQL function that processed large databases in parallel. The parallel SQL function improved the system’s turnaround and response times by dividing multiple queries to a relational database and executing each query in parallel.

2. Cluster systems (file-sharing parallel systems)
Cluster systems made up of Parallel ACOS mainframes were fault tolerant, distributed loads between hosts, and were manageable as a single system from the administrator’s point of view. A multi-system control facility (MSCF) was used to connect multiple hosts in a high-speed network configuration. Parallel ACOS cluster systems had the following characteristics.
  • Provided fault tolerance with a graceful degeneration function that would continue to execute online or batch operations on the remaining hosts if one host failed for some reason.
  • Automatic control of CPU load balancing at all hosts ensured the system operated stably and efficiently.
  • It was possible to upgrade the operating system or add or service hardware components at one host without interrupting the entire cluster system.
  • The entire cluster system could be controlled and managed as if it were a single system.

Figure 3: Cluster system

Figure 3: Cluster system

The system further increased its fault tolerance with a function that provided for a failure in the multi-system control processor (MSCP), the key component of a cluster system because it has exclusive control over resources shared between hosts. In the event of a MSCP failure, the function automatically isolated the failed processor and switched in a spare processor.
3. Open system connectivity
To support the move to open terminals, networks, and server machines, ACOS-4/XVP PX supplied connectivity functions with open systems in seven processing areas. This enabled the construction of systems that combined open systems with Parallel ACOS mainframes.

Figure 4: Open system connectivity in seven processing areas

Figure 4: Open system connectivity in seven processing areas

(1) Transaction connectivity
Transaction connectivity with personal computers, server machines, and other open devices was accomplished with the provision of an original protocol (the open OLF/TP protocol) that ran on the TCP/IP communication protocol.
(2) Job connectivity and delivery connectivity
ACOS-4/XVP PX supplied the functionality to import, monitor, and manipulate batch jobs from open servers and to output deliveries to printers connected to open servers. The NQS protocol was used for job connectivity. ACOS-4/XVP PX also provided the functionality to easily build job networks between open systems and Parallel ACOS machines. Delivery connectivity was implemented not by sending delivery files by FTP but by sending Parallel ACOS spool data to the open server. This configuration meant no modifications had to be made to service applications.
(3) Database connectivity
ACOS-4/XVP PX supplied functions to deliver databases quickly to open system and functions to access remote databases.
To affect the delivery of a database, the ACOS-4/XVP PX functions automatically ran and coordinated the extraction of data from a database on a Parallel ACOS machine, the transfer of the data, and the storage of the data on an open system.
For remote database access, ACOS-4/XVP PX provided a server function so that Windows applications that supported ODBC could directly access RIQS databases on Parallel ACOS machines. It also provided a client function so that Parallel ACOS applications could directly access databases on open servers.
(4) File connectivity
ACOS-4/XVP PX addressed the need to use Parallel ACOS machines as file servers by providing Parallel ACOS machines with a UNIX-compatible file system, an NFS-compatible file server function, and a network operating system server function.
The UNIX-compatible file system had a C-language application interface that enabled file access from Parallel ACOS applications. Files were also easily accessible from UNIX applications due to the inclusion of an NFS-compatible file server function.
Parallel ACOS machines could be used from personal computers thanks to network operating system support. ACOS-4/XVP PX supplied server functions for LAN Manager and NetWare, two of the most common network operating systems.
(5) Operation connectivity
ACOS-4/XVP PX supplied a function so that Parallel ACOS machines could be monitored and controlled from a UNIX GUI. This function improved the operability and visibility of Parallel ACOS machines because operators did not need to know specific Parallel ACOS commands and because operators could view the machines’ states graphically.
(6) Program connectivity
ACOS-4/XVP PX provided connectivity functions between programs, enabling bidirectional communications with other hosts using the international OSI-TP standard.

4. Base architecture revisions
NEC made sweeping revisions to the basic architecture to provide a parallel processing platform using advanced hardware that lowered costs while ensuring that existing user resources could still be used.

Figure 5: Parallel ACOS using advanced hardware

Figure 5: Parallel ACOS using advanced hardware

(1) Tightly coupled multiprocessors using CMOS processors
To implement parallel processing, Parallel ACOS machines used parallel processors with shared memory and a cluster system that joined these parallel processors with shared files as its base architecture.
A highly parallel TCMP configuration, which used up to 32 inexpensive single-chip NOAH CMOS processors, provided a parallel processing foundation that efficiently ran the parallel batch, parallel SQL, and parallel OLTP functions.
(2) FBA disks
The FBA disk architecture was employed everywhere. This not only allowed the use of cheaper and smaller disk units, it also enabled high-speed access throughout the system, including the operating system. To simplify the adoption of FBA disks, the operating system maintained the previous JCL and application interfaces and provided a function to migrate data in bulk from CKD disks.
(3) New front-end network processors (FNPs)
Inexpensive, high-performance, highly functional front-end network processors, which made use of the latest hardware and software technologies, supported open network environments, such as TCP/IP and OSI, and high-speed communication networks.
And to support cluster systems, the FNPs came with a load distribution communication function for balancing transaction loads and running fault tolerant operations and a pass-through communication function that enabled ATSS, VIS, and other subsystems and service applications to run without being aware of the cluster configuration.

 
Cover of the NEC ACOS-4/XVP PX manualPage from the NEC ACOS-4/XVP PX manual