IBM Brings Flexible Utility Pricing To Private Power Systems
May 6, 2020 Timothy Prickett Morgan
A lot of people, including us, focus on the technologies that go into private, on-premises cloudy infrastructure and how that is almost always distinct from compute, storage, and networking technologies based on the same raw compute – Intel Xeon, AMD Epyc, or IBM Power, pick one – available on the public cloud. But there is an equally important gap between private and public clouds, and that is the pricing methodology for the two.
IBM’s Cognitive Systems division, which controls the Power Systems platform, wants to close that pricing gap by adopting the same flexible, utility-style pricing for on-premises Power Systems setups as is available for cloudy infrastructure based on Power Systems iron that is also made by Big Blue and resold by the IBM Cloud, Microsoft Azure, Google Cloud, Skytap, Nimbix, and a few others.
As part of the Virtual Think 2020 conference that IBM is hosting this week, the company is unveiling a new packaging of Power Systems with the unwieldy name IBM Power Private Cloud With Flexible Utility Pricing, although even Big Blue admits in its presentation that “Naming is not final,” as you can see:
Might we presume to suggest the obvious and that this might be called “The Power Utility?”
The pricing on this new utility model for on-premises gear is not just restricted to Power E980s, as Enterprise Pools 2.0 has been (along with the Power E870s and Power E880 machines based on Power8 processors, of course), but will now include the Power E950 midrange box, and there is a hint from Steve Sibley, vice president and offering manager for the Cognitive Systems division, that Big Blue is thinking about how to bring Flexible Utility Pricing, or FUP if we are going to compare and contrast this with Elastic Capacity on Demand, or eCoD, or just plain vanilla Capacity on Demand, or CoD, or Capacity Upgrade on Demand, or CUoD. We have used utility pricing since the early 2000s, and before the cloud metaphor, which is not a particularly accurate one if you think about it, took off in the wake of Amazon Web Services launching in 2006. No promises, but we expect for the entry Power9 machines to somehow have their capacity metered and pooled across partitions on a single system and across multiple systems, and for IBM to do some updates to BIOS and firmware to make this happen.
This would be a true differentiator in the server market. X86 iron can’t do this. Arm iron can’t do this.
We are going to put a lot more flexibility into our consumption model,” Sibley explains to The Four Hundred. “And we were actually planning on doing this before the COVID-19 outbreak, but it has become even more important now given where the cash flow and financing is for clients out there today. We are going to put a lot more emphasis on the utility model that we introduced as Enterprise Pools 2.0 and we are going to reposition that more broadly across the line. That’s been only on the Power E980, and it will be on the Power E950. I will just say stay tuned and not give you a preview of when we might expand it beyond that.”
The flexible pricing works like this. Take a medium configuration of the midrange Power E950 as an example. The box might cost $100,000 configured up reasonably. IBM will charge $50,000 for a base machine with a certain amount of that capacity activated, and then charge per-minute, cloud-style, utility pricing – just like it does for machines on the IBM Cloud when they have Power9-based systems fired up – for the remaining capacity. IBM monitors the usage of cores, memory, and operating systems running in all logical partitions on the machine, and charges incremental for minutes used. And here’s the neat bit. The capacity can scale across a pool of machines with a set number of cores or one machine, and as workloads contract and CPU or memory capacity is not needed, then capacity is automagically cranked down and customers spend less money than they might otherwise. And if you have good credit and want to shift the whole thing to operating expenses instead of capital expenses, IBM will even do a lease for the initial $50,000 payment on the Power E950. There will be smaller and larger configurations of the E950, and of course multiple configurations of the Power E980 that are all set up and ready to go when the pricing and configurations become generally available with the offering on May 26. IBM is just giving us a taste of it all ahead of the availability.
Generally speaking, the per-minute prices for CPU, memory, and operating system capacity on the Power E950 and Power E980 under the Flexible Utility Pricing deal are costed out using a total cost of acquisition that is then spread out over three years (including hardware and software support we presume) and then divided into minute slices and then with a slight uplift – not much according to Sibley – to cover the capital expenses that IBM is assuming in giving customers the hardware ahead of time and not charging for it all upfront. The net effect of this is that IBM makes a little bit of incremental money in its charge for the CPU and memory and operating system capacity, but in early trials with beta customers, Power Systems shops saved somewhere from 20 percent to 30 percent on their overall Power Systems budget. This will, of course, have an impact on the top line revenue for Power Systems, but it also makes the Power Systems base more sticky and converts an occasional sale every couple of years into a steadier revenue stream when averaged across all customers.
At some point, IBM may as well bury the lease as a three-year reserved base instance – which is essentially a lease – and just be done with trying to do a big transaction with perpetual licenses for software. And at some point, it would be a good idea to get the pricing for the on-premises Flexible Utility Pricing more or less in line with the pricing for Power S924 and Power E880 and Power E980 systems on the IBM Cloud. Imagine if you could pool capacity not only inside the datacenter, but out into the IBM Cloud as well. This would be truly differentiated in the market, and as far as we know, much more useful than trying to build a set of on-premises infrastructure that is something like the clouds from AWS, Google, and Microsoft with their respective Outposts, Anthos, and Azure Stack variations.
“That 20 percent to 30 percent of savings of capacity, we think, is pretty conservative for the early Power E980 customers we have tested this out with,” adds Sibley. “We are going to give clients the ability to both pay less now and save overall by moving to this flexible utility pricing model.”
We will be looking into this as soon as more information is available.