Pity the poor mainframe computer. Maligned for years as a costly relic, it has become the Rodney Dangerfield of the information technology world.

That’s despite the fact that “big iron” machines run 30 billion transactions per day, handle 90% of all credit card transactions and process more daily transactions than Google search, according to the Share user group.

But when it comes to analytics, mainframes based on the classic IBM System/360 architecture have been largely left out of the party. Israeli startup Model 9 Software Inc. aims to change all that. The company has developed and patented technology that makes it possible for mainframe data to be saved to cloud object storage for use by data lakes and data warehouses.

The disconnect between mainframes and modern analytics dates back to the earliest days of 360 platforms when disk storage barely existed. At that time, the most cost-effective way to store data was on tape drives (pictured), which are “a serial device that you can only write to one at a time,” said Model9 Chief Executive Gil Peleg. Modern mainframes still use tape emulator software to compensate for this structural limitation.

Data sets, not files

Mainframes also have no concept of a file. Rather, data is organized in collections of records called data sets that adhere to a predefined form and use a protocol called Virtual Storage Access Method, which is optimized for rapid sequential access but not for the random methods used by files and databases. Data can be saved to disk, but it’s managed serially, an approach that was appropriate at a time when computers were limited to performing the same process repetitively.

Why do so many enterprises continue to live with this burden? Because bringing applications into the modern age that were written by programmers who retired years ago in languages that few people understand anymore is difficult. “We talk to companies that are on the seventh year of a three-year modernization project,” Peleg said. “Many fail, cost more than planned and have questionable return.”

Model 9’s software extracts mainframe data and reformats it for storage in cloud or on-premises object stores. In the process, it transforms VSAM files into the flexible and lightweight JavaScript Object Notation or XML format, either of which can be queried by popular languages like SQL. “Once you can get data into JSON you can do almost anything with it,” Peleg said. The software eliminates the need for tape and virtual tape, the company claims.

All the work is offloaded to System z Integrated Information Processors, which are essentially coprocessors for mainframes. Data can be stored on any open storage system, including network-attached storage, storage area networks and object stores. “We leverage the latest data transportation technologies, can heavily parallelize the data and apply compression,” Peleg said. “It’s completely transparent to mainframe applications.”

Model9 has raised $13.8 million funding, including a $9 million round a year ago and is coming off its “best year” since its founding in 2016, Peleg said, without providing statistics. Customers include America First Credit Union and Sirius XM Holdings Inc.

Photo: Wikimedia Commons