The Pivotal Year Ahead For Big Blue And IBM i Shops
January 10, 2022 Timothy Prickett Morgan
When we look back on it many years from now, this year, 2022, will be a pivotal year in the history of the Power Systems platform from Big Blue, which was commercially significant starting in February 1990 when the RS/6000 was launched but which obviously traces its roots through the IBM i and OS/400 branches of the Power Systems family tree all the way back to the System/3 launched in July 1969 just 10 days after humans first landed on the moon.
There are different kinds of pivot points in the history of the IBM midrange platforms, including the launch of the System/38 with its single-level storage and integrated relational database in 1978 and the launch of its sibling, the System/36, in 1983, and the converged AS/400 platform launched in 1988 that brought these two platforms together to create a united line. Moving the OS/400 platform to the PowerPC processor in 1995 set the stage for the initial quasi-convergence of the RS/6000 and AS/400 lines in 2000 and then the full convergence to a single Power Systems platform that could run IBM i, AIX, and Linux side by side or individually a few years later.
Since then, IBM has put a new processor into the field every three to four years, tweaked the operating systems and hypervisors to support it, expanded memory, storage, and networking options, and a few years ago tried to position the low-end Power9 platform as a GPU-accelerated node perfect for either HPC simulation and modeling or AI training workloads, but as the entry and midrange Power10 machines are set to debut in maybe May or June of this year, Big Blue is unsure of what to do in the HPC and AI markets, and indeed, in the mainstream server market where X86 machinery still rules and Arm machinery is on the rise.
For more than a decade, the Power Systems platform has been the undisputed leader for big iron machines that support big relational databases and huge ERP applications. And the “Denali” Power E1080 system launched last September just continues this dominance. IBM has won the battle here, but for many datacenters the war has shifted from big NUMA machinery running ERP applications and systems of record to massively parallel clusters of more generic machinery supporting data analytics and other kinds of systems of engagement.
The challenge for IBM i shops is to figure out who they want to be in 2022 and beyond, and that is a lot harder when IBM is not sure who it is, and what Power Systems should be and can be from this point forward. Yes, IBM can keep dropping new chips into essentially the same architectures for the Power10 and Power11 generations and keep existing IBM i and AIX shops running databases and ERP-style applications to 2030 or even further. Who else is going to build machines that run these applications? No one. Most IBM i customers get by on a few cores of Power compute, and they have hundreds of cores of scalability, and they could get by for another three or four or maybe even five decades on just Power10 and Power11 iron. So the scale of the Power Systems platform relative to the vast majority of the IBM i and AIX bases is not the problem.
What is pivot point for the Power platform here in 2022? Well, there are two of them, actually. The first pivot point is the customers, not the machines. This time, the installed based has to realize and appreciate all of the features and functions that IBM will be bringing to bear with the Power10 machines and invest in the future that IBM is offering, essentially high performance computing, ultra high security, and superior performance to run legacy and new applications in the modern era. They have to invest in this technology now so IBM can invest in and expand the future of the Power Systems platform.
And that is the second pivot point. IBM itself. IBM has to believe in its own Power Systems platform, use it to the hilt on its own cloud, and prove that it is better than X86 machines running Linux or Windows Server for modern applications while at the same time allowing for legacy applications like those coded in RPG and Java to be modernized and brought forward. IBM has to believe strongly enough in the platform to help – both technically and financially – IBM i and AIX shops modernize their hardware and systems software and then modernize their applications and databases and extend them with AI and where applicable HPC extensions.
There is never going to be a better time to do this than now, and frankly, this may be the last time that IBM will be able to afford to create a dramatic impact like the one that we are pondering. Again, IBM can walk to a future, dragging legacy applications to 2030 and beyond. This is more like IBM putting on an Iron Man exoskeleton and Tony Stark’s glasses and recreating a new Power Systems business that shows Big Blue is still one of the smartest systems builders on the planet and that it, more than the hyperscalers and cloud builders, knows what enterprises need and how to best take them there.
Notice I didn’t say hybrid cloud once there. Cloud is a deployment model, not a systems strategy. Who the hell cares where the machine sits in 2022? That is not the point. There will be a mix of on premises, co-location, cloud, and edge machinery. So what? What we are talking about is making sure that machinery has a Power Systems label, and that it is powerful enough and affordable enough that the 150,000 or so customers using Power iron as the heart of their mission critical systems can finally invest – and fully invest – in a Power Systems platform, top to bottom.
That’s a tall order. We know. But as we pointed out a decade and a half ago, IBM could have a $5 billion Power business where it had maybe 200,000 customers and all of them were grumpy about the high cost of the platform and were stingy in their investments or it could have a $5 billion Power business with 1 million customers who were all happy because the machinery was affordable and the software was excellent. Now, we are down to a $1 billion or so Power business with maybe 150,000 customers, and while these are the diehards, they are all still very stingy and spend grudgingly. And for good reason. IBM owns Red Hat, which helps quite a bit, of course. It can do all sorts of interesting things here to promote HPC, AI, data analytics, database acceleration, visualization, and other advanced workloads.
It is up to IBM to get all of its customers on Power10 and to make it affordable to do so, and it is up to all Power Systems customers to get current and use the innovation Big Blue has created to make more relevant, responsive, and revenue-generating application systems to drive their companies ahead of their competition. This is the only path back to a $5 billion and profitable Power Systems business. It will take a partnership between IBM and all of its customers to do this, and IBM needs to get a strategy together well ahead of the Power10 entry and midrange server launch in May or June. This cannot be more of the same from IBM, and it also cannot be more of the same from customers. It is time for everyone – and we mean everyone – to move on into 2022 and beyond.
All of our respective businesses depend on it.
RELATED STORIES
Expanding The Operating System Matrix For Power10
The Big Iron Customers That The Power E1080 Is Aimed At
IBM Drops Power10 Into Big, Bad Iron First
Balancing Supply And Demand For Impending Big Power10 Iron
Awaiting The Power10 Rollout And The New Sales Cycle
Time To Design – And Deliver – The Application System/360
In my opinion, it doesn’t matter who IBM is or whether customers run traditional ERP suits on Power10. For a thriving and relevant hardware platform one needs thriving and relevant software. Given how important open source has become for systems and applications software, the best way to have a thriving software environment is for actual hardware to readily be available to independent software engineers. Since developer access to Power 10 hardware is essentially zero at this point, it would appear free cloud services along with a comprehensive selection of entry-level hardware is needed as soon as possible to create and support interest in software development.