Cloud Source
Showing results for 
Search instead for 
Do you mean 

Bezos' law overlooks the conservation principle, changing the equation

on ‎06-30-2014 05:09 AM

Bezos2.pngSome weeks ago in this post I discussed how Cloud services come with hidden costs, and speculated that this may be related to their continually falling prices. In the example I used at the time, public Cloud services could have been seen as the equivalent of a budget airline that undercuts more-established airlines by offering a loss-leader basic fare, but recovers revenue by imposing unexpected fees for all kinds of services their customers might expect to normally be included. These fees might include charges for carry-on luggage in addition to checked luggage, taxes, dossier costs, service charges for paying with a credit card, and etc. Americans are familiar with their telephone bills, which include nearly a page of mysterious charges, such as an administrative charge, a regulatory fee, franchise surcharges, universal service fees, equalization surcharges, sales taxes, MTA taxes, and more.


After my reference to Moore’s law gives way to Bezos’ law, my colleague George McKee from Global Cyber Security, pointed out the fact that Bezos’ Law (blue in the image) shows a slower reduction in costs than Moore’s Law (orange in the image) would predict. This means that traditional IT improves faster than Cloud IT. The article’s author doesn’t take the “Ecosystem Services Conservation Principle” into account. It makes the legacy IT case even stronger.


“Ecosystem Services,” like capacity and availability monitoring; event management and escalation; identification and authorization services; data backup and restoration; system and app updates; security configuration, and intrusion monitoring and response; and etc., (which application owners were getting “for free” from their IT departments), nearly all go away when the app moves into the Cloud. To get them back – and preserve important properties like availability and data recoverability and security – application owners will have to rebuild them on their own at great cost, and usually unplanned cost. So I asked George to argue his points in some more details. Here is what he came up with.


Moore's Law dominates Bezos' Law

There's more to the low-cost picture of the Public Cloud than simply unexpected charges for mandatory services. First, although the baseline charge may be rapidly decreasing, the charges for the additional services may not decrease as rapidly, or at all. In the long run, we may end up in a situation where – for many IT functions – computation is free, and the cost of running a Cloud-based Service is consumed by fixed auxiliary charges. Of course, this has never stopped creative developers, and it will really mean the emergence of new, compute-intensive services that were simply unfeasible in the past.


We can gain some insight into how fast this might happen by looking at Moore's Law in comparison to the drop in cost of Cloud Compute Services, which has been called Bezos' Law. Moore's Law has many variants, and we'll use the law to predict that the compute-power-per-dollar doubles every 18 months. Bezos' Law was not actually stated by Amazon CEO Jeff Bezos, but describes the observed reduction in cost of a standard compute unit in Amazon Web Services, which has been cut in half every three years. Since we're dealing with truly exponential growth, we need to use logarithms to compare these in the same terms, and a few minutes work with a calculator shows that Moore's Law promises a Compound Annual Growth Rate (CAGR) of 58 percent, while Bezos' Law promises a CAGR of 26 percent.


Now, 26 percent annual growth in cost-effectiveness is fantastic development in most industries, and it makes a compelling case for using cloud computing on its own. But you have to wonder what happened to the other 32 percent of growth promised by Moore's Law. For a Cloud provider such as Amazon and others like it, this growth is plowed back into infrastructure (such as more datacenters in anticipation of future expansion), but it also promises expansion in profit margins, which helps to keep Amazon's stock price up at extraordinary levels while current profits remain nominal.


Let’s suppose that you have an enterprise that already has its own datacenter and IT services. You might want to consider implementing your own Private Cloud, and capturing some of that 38 percent excess growth for yourself instead of giving it to a Cloud provider. You will have to support depreciation and refreshment of your own compute infrastructure, instead of letting the Cloud provider handle that overhead, but you're probably rebuilding and redesigning applications to support the new style of diverse, mobile-user interfaces, and refreshing back-end systems along with that transformation. Adding a Private Cloud infrastructure transformation is not a large addition to that slate of projects.


The Infrastructure Services Conservation Principle

Cloud Responsibilities.pngThis brings us to the second aspect of cost differences between Public Clouds and Private Clouds, which is the optional services that IT providers have learned are essential to providing consistently high levels of availability for complex applications supporting critical business functions. Through years of painful experience, IT operations teams have developed elaborate methods and organizational structures to deal with these needs at the scale of hundreds of applications involving tens or hundreds of thousands of users. A widely-used collection of best practices in this area, the IT Infrastructure Library (ITIL), currently describes 26 processes and functions in 5 volumes. HP’s own internal ITIL-based support organizations, using a plan-build-run framework, consists of 69 different teams, each performing a distinct function, as explained by Robert Crawford and Russell Jukes in their talk at the latest HP Discover conference.


The need for these processes and functions does not magically go away when a solution is hosted in the Cloud, although the way they are delivered may be significantly transformed. There are many Cloud consultants who may mumble about “DevOps,” but the outcome of their advice is that users will suffer while the Cloud developers rediscover and reinvent the wheels that legacy operations have been rolling on for decades. And I’m not even talking about managing and securing a hybrid environment where Public Cloud and Private Cloud/traditional environments cooperate.


The Security Conservation Principle

In the security world, the absence of magic has been recognized by the U.S. National Institute of Standards and Technologies, where it is called the “Security Conservation Principle.”

It simply states that the same end-to-end functional requirements apply – regardless of what kind of environment the service is delivered from. And no matter how that environment is divided among Software as a Service, Platform as a Service, Infrastructure as a Service, or legacy IT. The backend environment will need capacity and availability monitoring; capacity planning; incident tracking and escalation processes; data backup and restore capabilities; contingency planning and recovery procedures for network and site outages; and so on. “We're in the Cloud, we don't need a business continuity plan any more,” is probably not going to get a good reaction from auditors when revenue-processing systems are involved.


When Cloud migration moves applications from legacy environments to a Private Cloud, all the legacy infrastructure processes and functions that keep the organization functioning smoothly as a whole can be retained with relatively small modifications, instead of the radical reinventions or outright abandonment that is required if they are to move to a public infrastructure or to a SaaS Cloud where roles and responsibilities are drastically different.


Make things easy for yourself

As you decide for your target architecture, make sure you take into account the costs of changing your integration, management and security environments as we already pointed out in the previous blog. If you decide to go to Cloud, make sure you look at making things easy for yourself, ensuring as much consistency between the Cloud environments you decide to use.


The HP Helion architecture can ease the effort of transforming legacy infrastructure functions by providing a private, OpenStack Cloud that is compatible with public OpenStack Clouds. Whether you implement your Cloud applications first in a Public Cloud or first in a Private Cloud, if it becomes expedient later to insource or outsource those applications or portions of them, the essential infrastructure functions only need to be transformed once, and then can be redirected to other Clouds with minimal changes.

0 Kudos
About the Author


Nov 29 - Dec 1
Discover 2016 London
Learn how to thrive in a world of digital transformation at our biggest event of the year, Discover 2016 London, November 29 - December 1.
Read more
Each Month in 2016
Software Expert Days - 2016
Join us online to talk directly with our Software experts during online Expert Days. Find information here about past, current, and upcoming Expert Da...
Read more
View all