The Long Road Back To Utility Pricing For Systems

One of the big problems of servers, storage, and networking, oddly enough, is that in the individual elements in which they are sold, there is a perception that they are not all that expensive. But the money does add up for large enterprises that are supporting hundreds to thousands of distinct applications on thousands to tens of thousands of pieces of equipment. And that capital is a sunk cost that is spent all at once and the value of that gear to be depreciated over time.

This is about as far from utility-style pricing as you can get. Back in the day five decades ago, when computers had many frames – storage, peripheral controllers, and so forth – and the one with the compute complex was called the main frame and it lived in a glass house so it could be appreciated for its largesse but have its noise muffled, well, no one actually bought computers.

First, they were too expensive and exotic, and second, vendors preferred the annuity-like revenue stream pouring out of these systems from the Global 2000, month and month out, year in and year out. Renting was the common practice, and vendors like IBM had great control over what companies paid for systems and when they would be encouraged to upgrade through a mix of new technology and rental rates that went up or down on new or old gear. It took the settling of antitrust lawsuits between the US government and Big Blue to compel the latter to make machines available for sale, to provide list pricing for both rental and sale (including component and peripheral prices), to open up peripheral specifications and support third party devices, and to eventually allow for IBM operating systems to run on clone systems. These changes form the basis of the modern computer industry, and started the practice of buying rather than renting machines because it was more economical, in the long run, to squeeze all the work out of a machine for four or five years compared to the three year run IBM more or less expected for a mainframe system.

One side effect of the ability to acquire big iron was that a lot of banks and then independent leasing companies figured out that they could make a fortune buying machines at a relatively discount and then leasing them back to end users for lower rates than the companies making the gear charged. So a rental was turned into an acquisition and then sort of back into a rental again. Leasing was common, but not universal by any means. And if you look at the finances of the public hyperscalers like Amazon Web Services, the company uses a mix of acquisitions and leases on its infrastructure as it turns the gear around and charges hourly rentals on it. Some of the premium that AWS charges is explicitly to cover the enormous capital outlay it has to create a utility computing system that spans the globe.

Several of the remaining big iron machines still sold in the world have the ability to turn CPU cores and memory capacity on and off on the fly, with hourly and daily pricing for that temporarily activated capacity – both IBM System z mainframes and high-end Power Systems iron offer this capability. But these capacity-on-demand features are not available in smaller machines from Big Blue, and generally are not available from other vendors. (Hewlett Packard Enterprise and Sun Microsystems, back in the dot-com boom, were doing similar enterprise-grade virtualization and on-demand pricing, but those RISC/Itanium Unix systems businesses are essentially noise in the server data these days.)

In a way, the public cloud has spoiled CIOs and CFOs and has brought home the joys of renting capacity rather than owning it, even if the cost can be very high over the long term compared to buying it and managing the assets internally. But that is just the thing. Sometimes, companies don’t want to own or manage that asset, and they are willing to pay a premium – a slight premium, mind you – to make that someone else’s problem while at the same time getting the utility pricing for their on premises systems that they have craved for a long, long time.

This is why Lenovo, which is the third largest systems maker in the world, has launched a hardware as a service (HaaS) effort, called TruScale, to offer subscription pricing on hardware, related systems software, and rolled up technical support services across the Lenovo datacenter product line for machines installed on-premises or in co-location hosting environments.

“We are launching this as a true consumption-based model, where they will pay, scaling up or scaling down, based on how much they consume of the datacenter equipment rather than having a traditional lease,” Laura Laltrello, vice president for services at Lenovo Data Center Group.

True consumption pricing, allowing customers to scale back as well as up, has been elusive excepting the instances we outlined above. “We have heard that complaint from our customers,” says Laltrello. “So we have set up the capability to meter and bill monthly based on usage. You use 30 percent, we bill 30 percent.”

This is distinct from capital leases, where the lessor owns the gear and the user usually has the right to buy the equipment outright at the end of the lease at fair market value, and operating leases, where the lessor owns the gear throughout the process and the user pays a monthly fee to use the gear. There is no concept of usage in the lease, and the assumption is that companies potentially have access to 100 percent of the capacity, so they pay 100 percent of the value of that equipment. The TruScale offering from Lenovo goes the extra step of monitoring usage and metering charges for actual usage, and then skips the whole lease provision and just does a monthly rental contract.

There are some stipulations with TruScale, as you might imagine, because no system maker can afford to create a complex and expensive piece of hardware or software that it has developed or acquired and put it out in the field without locking up a tremendous amount of capital. Just like customers want to only pay for what they use, Lenovo wants to have a pretty good idea of what customers are going to want to use before it puts the equipment on the truck for delivery. The goal is to match the usage as closely to the planned usage as possible, and Lenovo is taking an interesting tack in its approach to pricing.

The Lenovo machinery has monitoring chips built in that gather up metrics on usage for CPU, memory, I/O and sort forth so it can, if customers desire it, gather up per-minute usage by component at a very high granularity and, because the monitoring is done out of band using this chip and a separate management console, the gear can be monitored without the act of monitoring actually putting a load on the equipment. But after playing around with different equipment in the labs, Lenovo figured out that there is a direct correlation between idle power and idle machines and wall power relative to peak power that is a good proxy for the amount of capacity that is being used in gear. (This works for servers and switches a lot better than it works for storage arrays, obviously.) So the initial price metric that Lenovo is using in the TruScale offering is measured power by minute minus the idle power added up by a month as gauged against the total potential peak power the gear could draw minus its idle power. But if customers want to pay on actual usage, Lenovo has the data to do a contract that way, too.

Here is the other twist. After going through a capacity planning session with Lenovo, companies will take their best shot at guesstimating what capacity they will consume per month. The closer that customers come in hitting those estimates, the lower per unit of capacity Lenovo will charge, which is another way of saying that customers will share the risk – and the costs of the hardware – if their capacity planning is not good. If customers aim low on their capacity usage but think they might have big demand spikes that require more iron to be installed, they are going to pay more per unit of capacity but they will have gear on site to handle that spike – which has an economic value – and as they learn more about how their workloads behave they will have the right to renegotiate the contract with Lenovo so usage and charges better match what is happening. “We don’t intend for these two to be misaligned,” says Laltrello.

Customers will probably try TruScale before committing wholeheartedly to it across the datacenter, says Matthew Horne, general manager of TruScale infrastructure services at Lenovo. “Despite the education in the market with these types of pricing and even with the uniqueness of what we are doing with TruScale, customers still want to prove it for themselves. This gives customers the option to start small and scale smart with it, perhaps containing it within specific projects or lines of business or applications.”

The TruScale program is also designed to provide a much more expansive portfolio of products that come under utility pricing, and it does not have a 50 percent capacity minimum as some other offerings do, according to Horne.

Lenovo has two big initial customers that are kicking the tires on TruScale. The first is a managed service provider in Latin America that clearly wants to have costs scale with usage because that is how this company charges for capacity on the other end of the wire. The other big customer is a large system integrator who is selling through capacity to one of its end user customers.

The obvious question is why should a program like TruScale succeed. On-demand pricing for on-premises equipment is not a new thing, but it is obviously tough to build in the kind of flexibility that the public cloud offers because the gear is ultimately sitting in one datacenter dedicated to one customer if you are talking about enterprises. (The MSPs and SIs are more like clouds or hosters in the scenarios outlined above.) About 23 percent of customers lease at least some of their IT gear, and there is probably not one of the Global 2000 or even the largest 20,000 customers in the world that hasn’t bought some cloud capacity. The big difference now, says Laltrello, is that customers are used to cloud pricing. And another difference is that Lenovo is sharing the risk with the customer without either party assuming all of the risk. “We believe customers have good intent,” says Laltrello.

And unlike many leasing giants in the past, who benefitted financially and mightily when customers had to do a system upgrade in the middle of the lease. That usually stung pretty bad.

Sign up to our Newsletter

Featuring highlights, analysis, and stories from the week directly from us to your inbox with nothing in between.

Subscribe now

Be the first to comment

Leave a Reply

Your email address will not be published.


*


This site uses Akismet to reduce spam. Learn how your comment data is processed.