This week I’m with IBM, and as most ex-IBM employees will tell you, once you have Big Blue in your blood you never really get it out.
The company has changed a great deal since I worked there and struggled to overcome a series of challenges. These days IBM is all about transparency choice, and they are a poster child for “open” and have become the leading Linux advocate, having recently bought Red Hat.
But it was their High-Performance Computing effort that really caught my attention. IBM is the only firm approaching this segment with a very different strategic plan that, if they execute, should allow them to dominate the segment again if all the planned elements come together. They have a solid foundation with two legs of a three-legged technology stool – the last leg is an exceedingly difficult – but they are the only firm in their class committed to this unique strategy
The three legs to this strategic stool are bits, Qubits, and Neurons. The bits are their traditional platforms of OpenPower and System Z, the Neurons revolve around Watson and AI, and the Qubits and the final leg of the stool is Quantum computing.https://o1.qnsr.com/log/p.gif?;n=203;c=204663295;s=11915;x=7936;f=201904081034270;u=j;z=TIMESTAMP;a=20410779;e=i
Let’s chat about the blending of these elements and why it will revolutionize the future of HPC.
OpenPower – System Z
IBM’s approach to the server market is very different in that they don’t use the industry-standard X86 platform anymore, and they are the only vendor providing a Cloud In A Box solution in their System Z.
Now, in terms of pure performance, it is hard to compete with X86, but by controlling the ecosystem more completely IBM can better assure both the security and reliability of their platforms.
With penalties for breaches moving from the millions to the billions in some cases, the pure performance advantage of X86 is offset by the related risks associated with the more common platform. And, interestingly, we can now argue AMD leads in this segment and that company is helmed by an ex-IBM employee Dr. Lisa Su.
Their System Z is unique in the market optimizing I/O and secure partitioning and balancing it with unparalleled security to own a unique segment in the industry. The fascinating thing about System Z is that in many ways it is more similar to an Apple product than a typical IT product regarding customer loyalty and satisfaction.
So, they largely lead on security and reliability but trade-off performance. Yet IBM realized there were multiple paths to performance. One is just to make sure your platform doesn’t fail, but the other is to apply intelligence so that it works smarter not just faster.
Smarter performance is where AI comes in and firms using the blended intelligent solutions from IBM report that they can cut the workload on their systems by up to 2/3rds.
This smarter performance not only saves energy, which is obvious, but it also increases throughput because if you can cut the loading by 2/3rds you can increase the amount of work a system does by 3x. In a race car this would be the difference between just focusing on the engine and balancing the focus by making the driver more capable.
I watch a lot of car races and often the fastest car in qualifying won’t finish the race and the fastest car, with a poor driver, won’t place. But if you can combine a great driver with a good car, they will have a winning season. It is about balance and by combining AI with the platform means that, in terms of productivity, with the right jobs, you should get the best of both worlds, market-leading performance and the security and reliability that makes IBM’s hardware unique.
Now this is the game-changer, and IBM has not really delivered on this yet, but they are making impressive progress both with development of the technology and in bringing up developers around this technology.
Quantum computing isn’t good for everything. But what it is good at, it should be able to do near instantly. This performance booster is more than a blower for an engine providing a significant boost to the performance of the car and similar to the rocket engines Tesla (they are crazy) is planning on putting on their coming sports car. (I figure this car is Elon Musk’s way to getting rid of a lot of old guys who can afford, but can’t control, the resulting car and am probably on the shortlist of soon to be dead crazy Tesla drivers).
Once the Quantum technology is blended with the overall solution, you’ll get things like Quantum level encryption, a level of security over the data stream we can only imagine today. Also, the ability to offload some complex mathematical problems that currently can bring a system to its knees and have them completed in microseconds.
The resulting combination will be a system that automatically configures itself on a job-by-job basis, can handle massive amounts of concurrent loading, and can massively outperform the systems of today much like the systems of today outperform the mainframes of the 1970s.
One of the bad things that have happened over the last few decades is that tech companies are mostly owned by Hedge Funds, forcing them to focus on tactical projects and putting them at long term strategic risk. This inability to function strategically is one of the reasons Michael Dell took Dell private.
IBM was created at a time when strategy was far more pronounced, and it was built into the firm’s DNA. Nowhere can you see this more pronounced than with their HPC efforts. If they can fully execute, they will not only leave their competitors in the dust, those competitors will be years out of position. Both AI and Quantum elements suggesting IBM’s only real threat may have to come out of China, where State-level support is resulting in similar advancements that have yet to come to market.
In short, IBM, which carried the US to technology dominance mid last century is likely the best bet for us to hold on to that dominance through this century, at least regarding HPC.