Bad arguments against the McKinsey report

I do not necessarily agree with all the McKinsey conclusions, but in at least two instances Golden's criticism of McKinsey is off the mark:

(1) IT administrators in fact spend very little time on maintaining hardware. Application monitoring and troubleshooting, software patches etc. (all tasks that do not disappear if we move to EC2) are much more substantial. The reduction in headcount assumed by McKinsey is plausible.

(2) Contrary to Golden's attempt to ridicule them, the McKinsey numbers evidently include facility cost, or do you seriously believe the $14k are hardware cost for a a dual CPU machine at steep volume discounts?

I hope more orgs ride the cloud. esp bots. They are much easier to ban, and I can easily ban the whole cloud. It really save time for me.

Bernard,
I cannot agree with you more on this, especially the four points depicting how wrongly 'strategy firms' understood cloud computing. Also another important aspect you said is noteworthy – that cloud is NOT just about cost saving.

I'll admit, I scanned the article... 1. it is a "good-sized" financial services company. 2. The company is primarily WinTel. 3. They "moved all systems"... What a bunch bs... any "good-sized" company that's primarily WinTel and they moved all systems to the Amazon Cloud servcies needs a sanity check... But, I'm certain they paid McKinsey a huge, worthless fee... The un-named company CEO should fire the senior IT person...

Actually the main question should _not_ be "what does a unit of compute capacity cost me?" Assuming we're only looking at data center infrastructure here, it should be "what does a unit of processing delivered to end users cost me?" That is affected _both_ by the cost of your capacity _and_ by the utilization of that capacity. Probably what you meant to say...

You need to be a bit more careful with your analysis Bernard. Notice they include $20M/MW for facilities, $.1kW-hour, and $14K per server in the footnote. Sounds like they are including facility capex costs, electricity costs, and server capex costs. Also, given how high the server costs are, I am guessing they are not assuming discounts. Wish they would publish more on their methodology.

Bernard,

I don't necessarily disagree with the report's conclusions from the following two perspectives:

a. As organizations get large, the economies of scale allow them to start operating in a manner that is similar (or analogous) to the service provider model - delivering similar efficiencies. There are many large corporations that run their own dark fiber versus buying capacity from a telco. This model makes sense for them.

b. Organizations whose business model is completely dependent upon internet access and application delivery (from brokers to banks) often can hire enough people of IT education and skills to create a long term career path for these people. When done right, these employees, can create a highly efficient core that it just as capable as a public-cloud. Note, most large companies cannot do this.

c. The new virtualization and data center technologies are making it easier for competent IT organizations to run internal clouds and squeeze more efficiency from current infrastructure.

I think the McKinsey Report is perhaps a little simplistic to assume that their results apply to all large organizations....they don't.

--Mani

Disclaimer - my company, Treova (http://www.treova.com) , is a cloud computing var

Interesting debate...for all of the arguments that support cloud computing and all of the downsides to adopting it, the approach doesn't veer too far from the legacy approaches - managed services and outsourced services by service providers. I am old enough to remember and also an ex-Telco/managed services employee so I tend to lean toward the McKinsey report as a practical and plausible discussion. The cost savings seem to be marginal by their estimation and assumptions over time. In reality, 90% of the managed services contracts for which I was engaged also showed very marginal savings - in capex/opex including headcount - over time and in most cases, caused substantial increase - especially at contract renewal or closure. Not because the technology wasn't there - it was the hidden fees, clauses and limited compensation for poor service. Always remember, the more IT services become a commodity, important metrics will degrade - reliability, dependability and security - all will decline just like all other utility models.

Having worked at McKinsey I am familiar with the risks and benefits of their analytical approaches. So, while I agree with the author about some of the technical shortcomings possibly built into their financial model, the author does not highlight one of the obvious issues that impact the financial impact of cloud computing that I strongly assume McKinsey came upon:

Cloud computing is by and large priced on some sort of volume pricing, i.e. the more bandwidth, storage, and cycles I consume, the more I pay. I.e. my hosting costs thus are mostly variable costs and tied to my volume they stand to grow over time.

On the other hand, on premise installations have a much smaller variable cost component (e.g. power consumption, direct IT personnel costs, etc.) and a larger fixed cost component that gets amortized over the life of a project (e.g. server capacity, software licenses, etc.).

This inevitably means that at some break-even usage volume, a cloud computing setup becomes more costly over time. This is analogous to owning and paying off a car and then have the free use of the vehicle in perpetuity once it's paid off. Vs paying a lease and thus incurring a monthly charge in perpetuity. At some point, the leased car becomes more expensive from a long-term ROI point of view.

If you then layer in the need for SLAs and security requirements in large, high volume applications, and the case for cloud computing isn't so clear cut anymore.

I worked at a major consumer software maker with hosted applications that serve millions of users with definite peak load periods. What my investigations into the ROI of cloud computing setups came up with was that smallish applications with not yet proven market potential where time to market is critical and/or applications with significant load spikes would be strong candidates for cloud computing implementations.

However, major, high volume applications with stringent, mission critical reliability, security and up-time requirements could indeed be more cost-effectively implemented using inhouse resources, assuming those resources are not hideously inefficient.

So, not knowing the McKinsey study in detail but being familiar with McKinsey's analytical approaches, I would conjecture that an analysis of the variable vs fixed cost characteristics of cloud computing vs on premise hosting setups was conducted and they arrived at the same, inevitable conclusions I did: For high volume implementations (which the cited largish financial institution presumably has) cloud computing may indeed be the more expensive way to go and thus should be avoided.

Regards,

You are missing the most obvious point in all of this. All things being equal cloud IS more expensive than doing it yourself. Has to be or else Cloud service providers would not have a viable business model and cloud computing would not exist in the first place. The economic benefits of cloud will vary from organization to organization and will change over time. Each organization needs to understand their own situation and reassess it over time.