Just How Cheap Can Public Cloud Computing Get?

Public cloud computing is all the rage, and cost is a big driver. It’s not the only driver, mind you, or even the top reason that developers turn to the public cloud, which is primarily about development agility. Still, cost is a serious consideration nudging enterprises to forsake building out data centers to rent space in Amazon, Rackspace, Google or others’ clouds.

Given this, just how low can public cloud pricing go?

Cheap And Getting Cheaper

Pretty darn low, if Bernard Golden’s analysis is correct. Golden looks at hardware and software infrastructure, Internet bandwidth, energy, labor and cost of capital and finds public cloud trumps data center in every category. That cost savings isn’t small, either:

[W]hile the cost savings a cloud provider achieves by designing and directly procuring servers may only be 15%, the total cost savings available to a cloud provider summed across all six categories might be as much as 75%.

That’s real money.

And given that public cloud providers will continue to wring costs from their own operations in order to lower prices for customers, it’s money that may keep multiplying. Or dividing, if you prefer. One key way, as Golden points out, is by driving ever higher utilization of fixed assets:

[W]e can expect to see public cloud providers execute sophisticated pricing schemes designed to increase infrastructure utilization rates to as high a level as is practicable in terms of service quality acceptability. We’ve seen this already with AWS’s reserved and spot instances, which are designed to drive up utilization of Amazon’s computing infrastructure. We are nowhere near the end of this sophisticated demand management.

In other words, low is set to get even lower.

Cutting Prices And Services

Not every cloud provider is chopping prices, of course. As David Linthicum, senior vice president at Cloud Technology Partners, notes:

Some public cloud providers—SaaS, IaaS, and PaaS—are … slowly raising prices on the key services even as they offer more teaser-tier discounts to get new customers.

Most enterprises will tolerate price increases, but not reduction of services. Yet many cloud providers are also delivering slower storage and compute services, while suffering more frequent outages. Raising prices while reducing service is the fastest way to get an customer interested in the competition.

Linthicum is likely not talking about AWS here, which has consistently lowered prices, even while improving services, for years. But he’s right: the “freemium” model is often abused, such that vendors feel they can get away with delivering a worse experience in return for lower prices. They can’t.

But even for AWS, there’s another side to the cost equation that many organizations don’t consider: performance.

Your Mileage May Vary

Whether comparing apples (public clouds) to oranges (data centers) or apples to apples (i.e., different public cloud providers), evaluating different vendors on a price-for-performance metric turns out to not be very useful. As Carlo Daffara highlights in a study:

[I]n some instances there may be substantial differences in the performance of virtualized instances within public clouds—both within different jobs, and even in the context of the same job—introducing a variability that must be taken into consideration when comparing execution economics. A common error, in this sense, is comparing a generic cpu/hour cost per core between an internally provisioned hardware system and an external public cloud: the comparison is in many instances biased by a factor 2 to 10.

In other words, it’s not enough to approximate costs by measuring generic performance because, as it turns out, such generic performance doesn’t really exist. CPU, storage and network performance can be highly variable.

Even between public cloud providers, raw CPU speeds prove to be a insistently unhelpful way of evaluating vendors. Eucalyptus, a hybrid cloud vendor with strong hooks to AWS, argues that “[b]uyers have to be sure that they’re getting into a cloud that has enough redundancy and services to produce real efficiency gains for dev/test teams,” and not merely clock CPU cycles.

But even a like-for-like comparison between clouds or data centers doesn’t tell the whole story. After all, as the EU report goes on to conclude, “the lack of tuned parallel file systems in public cloud environments can be a major factor both in terms of VM engineering and performance wise.” All the work you’ve put into making your infrastructure sing within the data center, in other words, may need to be redone for the cloud.

Price Is Just The Start Of The Cloud Conversation

In sum, while public clouds are going to keep getting cheaper, “cheap” depends on your enterprise’s workloads and how prepared you are to use the cloud. The cloud can offer significant benefits in terms of improved business agility and lower costs, but only if you’re prepared to leverage it properly.

Lead image by Flickr user getinet, CC 2.0