IBM has launched an update to its flagship mainframe operating system, a release called the z/OS 3.2, that’s designed to enhance enterprise toolsets ranging from hybrid cloud management to AI to cybersecurity.
IBM touts z/OS 3.2 as ideal for companies that need to leverage AI to improve business decision making and mine more value from existing data sets. The new OS achieves this by incorporating updated APIs designed to enhance the deployment of z/OS data in hybrid cloud environments — IBM has long stressed its strategy of focusing on hybrid cloud. The new OS’s modern APIs also enable companies to use their data in AI and cloud environments without the need for Extract-Transform-Load (ETL), a data transfer technology that can be unwieldy.
Specifically, the improvements to the EzNoSQL APIs enable enterprise customers to use Python on z/OS, along with COBOL, Java, and C, which enables developers to construct scalable NoSQL databases on z/OS. Among other use cases, this allows predictive modeling and data analysis for large enterprise applications.
Supporting these AI enhancements is an extensive tech stack that features open source frameworks and IBM’s Z Deep Neural Network Library (zDNN)— geared for deep learning data processing—which provides APIs to interoperate with the mainframe’s on-chip accelerator. IBM says that all of these elements work in concert to support AI model execution directly on the mainframe, so companies have less need to transfer sensitive data between domains.
To cater to the cybersecurity concerns of large enterprise customers, z/OS 3.2 includes quantum-safe cryptographic algorithms and pervasive encryption tools. This new release offers new quarantine functionality via Resource Access Control Facility (RACF) userid containment, a technology that enables the system to manage user access to sensitive resources.
The hardware that supports this new OS, the z17 mainframe, sports a 5.5 GHz IBM Telum II processor enhanced with an AI accelerator. The system can handle 450 billion inferencing requests per day, with an ultra-fast millisecond response time. For customers with the budget, IBM’s big iron can run even faster: In Q4, IBM is releasing Spyre, a 32-core AI accelerator that’s an add-on.
The Spyre has been architected for high performance in-transaction inference, and is specially designed for tasks like real-time fraud detection and financial transaction processing. It’s available on PCIe cards and — for companies that want to massively scale AI — a cluster of eight Spyre cards can add a robust 256 accelerator cores.
The mainframe computer, a technology that gained major adoption in the 1960s, continues to be the single greatest enterprise workhorse. It’s used by almost all major banks and the majority of the Fortune 500 for high volume data processing and is well-regarded for its solid security. Given the vast volume of data that these enterprise customers need to load into large language models, the z17, and the new z/OS 3.2 OS that now boosts support for AI, will likely enjoy significant customer interest.
General availability of z/OS 3.2 is scheduled for September 30.
