4 ms·

Mainframe stack is very simple COBOL-Db2, online will be CICS-COBOL-Db2, batch is JCL, COBOL with sorts dfsort or syncsort, SAS, much modern will be COBOL/JAVA

Mainframe stack is very simple COBOL-Db2, online will be CICS-COBOL-Db2, batch is JCL, COBOL with sorts dfsort or syncsort, SAS, much modern will be COBOL/JAVA with Db2 Queue replication. Scripting language mostly REXX. Last 3 years, lots of things are happening, more offloads to zIIP than General CP. In terms of applications more and more Db2 Queue replication for the time monitoring. CDC target being Kafka, IDAA for analytics and after acquisition of RHEL, now I see admins, sysprogs installing Spark, Anaconda, Jupyter in mainframe. Usually in mainframe for detecting frauds, it used to be rule based now some companies are planning for machine learning in mainframe and be close to source and detect frauds. Only thing mainframe cannot support is Deep learning as that requires GPU, no GPU here. Anyones companies already started ML and spark in mainframe and what's your experience till now.

rbanffy 6y ago
> Only thing mainframe cannot support is Deep learning as that requires GPU, no GPU here. You can have racks full of PCIe slots if you want. I don't think there's anything that prevents you from adding a dozen GPUs on every IO drawer if that's what you want. And, unlike PC-based servers, there's a ton of silicon dedicated to keeping all IO channels fed, so, you'd probably have very low CPU usage.