In January 2015 IBM introduced the z13 microprocessor, which is regarded as the world's fastest. Speed is important, but as any devotee of Moore's Law knows, it's pretty transitory. So in a world of ever-quicker chips, why does z13 matter? To cop a phrase from the radio waves, it's all about the analytics.
When I think of why the z13 matters, I'm reminded of Willie Sutton's often-cited (though probably apocryphal) response to the question "Why do you rob banks? Because that's where the money is." The z13 matters because "that's where the data is."
The z13 delivers many value propositions, including speed, but the one that I think about the most is what it delivers to organizations in terms of analytical capability. As I said above, the mainframe in the modern enterprise has the data, and that feeds analytics. With the ability to have 10TB of memory with large memory frames/pages, the practice of bringing your analytics to the data (rather than copying the data, via ETL, to the analytics) has more relevance than ever before. z13 is designed for cloud, mobile and advanced analytics capabilities, which is a function not only of speed but of design.
Think of it this way: a person buys a Lamborghini to get to work a bit faster than he or she would in a Honda Civic. It makes sense that a few moments might be saved every day thanks to the ability to zip in and out of traffic and hit 175 mph between traffic lights. But if commute time is important, a lot more time would be saved by moving a few blocks from the office rather than commuting 25 miles each way. That's sort of how z13 works.
I talk with a lot of mainframe users and I always have some fun asking them why they ETL so much data from their mainframe off into their 1990's data warehouse architecture, complete with operational data stores, data warehouses and data marts. The response is typically, "isn't that they way that everybody does it?" Upon further probing, the reality is usually, "well, that's the way we've always done it." That's not a very compelling reason to do anything. Rather than carrying buckets of water from the river, just move next to the river. That's what getting away from the ETL paradigm is like. And z13 is the enabler that lets people do that. It's not just an incremental improvement over a previous technology: it's a game changer!
One financial organization told me that they do real-time fraud detection, and that it takes them 36 hours to complete the analysis. Why 36 hours? Because they use a very elaborate ETL process that batched the transaction data and copied it repeatedly into different staging areas with various replication and brittle ETL tooling before it was finally in a form that could be analyzed for fraud. I'm happy to say that this organization is now moving to a system that does real-time analytics against the transaction data directly — on the mainframe.
Another value that the z13 brings that adds even more oomph to analytics running on the mainframe are the new SIMD (Single Instruction Multiple Data) instructions that perform vector operations against entire arrays of data in a super-linear fashion. Is anyone still questioning IBM's commitment to move the analytics to the data now?
The last value that I want to mention is simply the increased capacity of this "I/O monster." Whether it be consolidation of machines or simply growth of existing workloads, z13 delivers. This is good news for organizations using z/OS as the traditional workhorse of the mainframe or organizations using Linux for z Systems - which is VERY attractive when running Oracle on these behemoth processors from a software licensing perspective.
One thing I've noticed over the years is that mainframes are usually described in terms of their speed and power. That makes sense, but it doesn't really tell the story of why mainframes have been the gold standard for analytics (or "big data") for so long. Mainframes are built for speed, but what really makes them special is how they integrate functionality. That's where z13 is making a real difference.
It's time to stop looking at the mainframe as just a transactional data system or the "System of Record" with other systems surrounding it providing analytical capabilities. The z13 makes this point abundantly clear.
This story, "Why IBM's z13 microprocessor matters to analytics" was originally published by Computerworld.