'It's the operating system, stupid'
Mills says that the real secret sauce in IBM's high-end system is in the operating systems. Its battleship AIX Unix variant, he said, has taken some of the job-control and workload-scheduling techniques used in the MVS-OS/490-z/OS operating system for mainframes and grafted them onto AIX so they can take work and spread it out across the many threads in a Power Systems machine – and that it can do so despite the fact that commercial workloads are difficult to parallelize because of the serial dependencies in commercial applications.
The AIX software is also making use of memory compression, plus dynamic thread and memory optimizations such as memory affinity that can tie cores and threads to specific memory chunks to speed up performance, to push more transactions through the box. Each of these techniques, which are at the nexus between the Power7 processor and the AIX kernel, can boost performance on certain workloads by 10 to 15 percent, depending on the situation.
"We benchmark all the time," Mills said, and he pulled out some real tests to support his point. "We have a favorite competitor who likes the color red. We like the color blue. This is real workload benchmarking, not some phony baloney made-up thing that goes in an ad. We deliver a system that is fast for what customers run."
IBM's Smart Analytics System 5600 spanking an Oracle Exadata
You'll notice that in the above comparison, IBM used its x64-based version, not a Power Systems variant.
Mills had nice things to say about IBM's $1.7bn Netezza acquisition as well, saying that it produces true appliances. "It literally installs in hours, requires very little tuning, and has tremendous scaling characteristics," he said.
The Netezza machines are based on the combination of IBM blade servers and special field programmable gate array (FPGA) coprocessors that prechew and presort data stored in a heavily modified PostgreSQL database created by the formerly independent company.
In the fourth quarter of the last fiscal year, IBM did 22 deals of more than $1m for Netezza products, and signed up 29 new customers. Revenues were up 70 per cent over the year-ago period. IBM now has over 500 Netezza customers, and the growth rate is accelerating. Mills said that Netezza had an 84 per cent win-rate once it could convince a customer to do a proof-of-concept installation.
"We solved the problem of doing more than just a bundle, more than just a bucket of bolts dropped on the floor at the customer shop, which is what we see happening at a lot of our competitors that are talking about packages," Mills said. "What they are really doing is shipping in a lot of stuff together, dropping it on the floor, and then have five engineers over a two-week period get it installed and up and running."
Taking another shot at Oracle's Exadata and Exalogic appliances – and perhaps at HP and Microsoft, which began delivering SQL Server appliances in January – Mills said other vendors exaggerate what they are really doing with workload optimized systems.
"At the end of the day," he said, "you get all of the vendor high-testosterone claims in the tech industry. The reality is: 'What do you bring to bear?' Don't write checks with your mouth that your technology can't cash. We've got the technology, we've got the capability. We will show up with our people, our code, and our capability, and prove it."
Some of those people who are going to show up in IBM versus Oracle engagements are the researchers who put together the Watson question-answer machine, the hundreds of mathematicians who Big Blue has on the payroll at IBM Research, and the thousands of experts the company has hired in its Global Services behemoth, who know the ins and outs of a particular industry.
How does Watson compare to Exadata? Let John Kelly, director of IBM Research, count the ways, as he did in this table:
IBM's comparison shows no love for Ellison's Exadata
Watson was created by IBM Research to play – and win – the Jeopardy! game show, and it has yet to be productized. But this statistically driven QA machine is every bit as shiny and glamorous to a CEO and CFO as an Apple iPad is to a teenager.
"Watson is aimed at an entirely different space," Kelly explained at the Investor Day briefings. "If you look at an Exadata system, it's mainly doing work in relational databases, consolidating workloads in a very tight form factor. That's about it. How many industry-standard parts can you pack into a sheet metal box? That's fundamentally what it does."
Watson is not interested in the packing density of commodity parts, explained Kelly, but is instead interested in answering questions using advanced analytics. "It's leaps ahead of what anyone else can do, and I don't think anyone else has even started work on this on such a scale." ®