Some data-intensive processing still resists the cloud, in industries like banking, insurance, travel, and the public sector. For these uses, the mainframe, IBM’s historic activity, is still widely used and it even continues to be modernized, for example by integrating into hybrid cloud environments.
IBM announces this Tuesday, April 5 its new model, the z16, integrating for the first time an artificial intelligence accelerator to help process large-scale transactions, as well as an encryption system resistant to future quantum attacks.
300 billion inferences per day
“It’s revolutionary”says Catherine Chauvois, director of the mainframe division at IBM. “Thanks to the integration of the new Telum chip, inferences, i.e. the execution of the AI engine, are carried out directly on the processor in real time, with much lower latency than before. This means customers will be able to run their fraud detection models on 100% of transactions, where previously only 10% could be scored.” The IBM z16 would be able to process 300 billion inference requests per day with a latency of only one millisecond, and the machine will require no more than 5 minutes of downtime per year according to its manufacturer.
This new system is compatible with all the main development frameworks on the market, and of course supports Red Hat OpenShift.
Real-time large-scale processing
IBM mainframes are still used today by two-thirds of Fortune 100 companies, 45 of the top 50 global banks (core banking system), 8 of the top 10 insurers, 7 of the top 10 global retailers and 8 of the top 10 telecom operators for their critical applications.
The core target of the z16 is therefore sectors that need to analyze large-scale transactions in real time, requiring very high processing power, such as healthcare and financial transactions. According to an IBM study, zSystems execute 70% of global banking transactions by value. The new capabilities could apply to new use cases, such as loan approval, bank clearing and settlement, or retail fraud modeling.
The IBM z16 is also the industry’s first system to feature quantum computer-resistant technology, according to the company. Quantum computers could in the near future be used to break current encryption technologies. “Quantum attacks are theorized. In four or five years, with a sufficiently powerful quantum machine, we could break the current encryption keys. Today, cybercriminals are stealing encrypted data in the hope of exploiting it in several years, when the computing capacities will be available”explains Pierre Jaeger, technical director of the mainframe division at IBM and responsible for quantum development.
Euclidean network encryption
To ward off these future attacks, the IBM z16 relies on Euclidean network encryption algorithms, an approach “mathematically resistant to a Lattice algorithm” developed in collaboration with IBM Zurich and ENS Lyon. “This is a first in terms of hardware implementation on an industrial system”, emphasizes Pierre Jaeger. The new mainframe also features a secure initialization process, which prevents injecting malware during “boot” to take control of the system at startup.
All this while remaining for clients in the same budgets as the previous generation, specifies Catherine Chauvois, in a context where IBM clients are “in general in logics of growth”. In Q4 2021, the company reported a 6% decline in sales of the IBM Z lineup, but over the past decade it reports a 350% growth in the hardware installed base.
We would like to give thanks to the author of this post for this incredible content
IBM’s latest mainframe bets on real-time AI and post-quantum security
Discover our social media accounts as well as other pages related to it.https://www.ai-magazine.com/related-pages/