Slashdot Mirror
The Economics of Federal Cloud Computing Analyzed
jg21 writes "With the federal government about to spend $20B on IT infrastructure, this highly analytical article by two Booz Allen Hamilton associates makes it clear that cloud computing has now received full executive backing and offers clear opportunities for agencies to significantly reduce their growing expenditures for data centers and IT hardware. From the article: 'A few agencies are already moving quickly to explore cloud computing solutions and are even redirecting existing funds to begin implementations... Agencies should identify the aspects of their current IT workload that can be transitioned to the cloud in the near term to yield "early wins" to help build momentum and support for the migration to cloud computing.'"
These government types always have their heads in the clouds...
Science advances one funeral at a time- Max Planck
$20B + system built by Microsoft's new 'Danger' arm + White House IT administration = dream government cloud!
All your base are most certainly permanently lost!
Cloud computing provides lower costs due to attaining economies of scale. The federal government certainly has scale to attain any efficiencies that a cloud operator might use to reduce the cost. It is scary to think the government will hand over data and processing to the cloud instead of providing a federally managed private cloud on a secure private network. This reeks of lobbying and special interests. Follow the money.
Something that isn't often mentioned when discussing cloud computing is a general problem with who has control over your data, where it resides, and what prevents others from accessing it. When you move to the cloud you need to be able to trust the service provider completely. This might not be a problem for unimportant things, but the government has privacy and secrecy obligations that it would not be able to fulfill by handing it sll over to a third party.
Cloud computing offers nothing. And by nothing I mean nothing new. Nor does it fix anything. The internet already works.
There, I said it.
For 99% of us, a web server does everything we need it to. Redundancy and fault tolerance are already very easy to buy in other forms that are perfectly reliable and non-invasive (RAID, adding servers for specific services, buying better hardware etc). These problems were solved long ago.
Yes, for the rare corporation that requires huge server clusters, cloudifying their infrastructure is the right direction to go. But that and buying a cloud are two completely different stories. If your server count is already that high, then you most likely already have the budget and the people to create your own cloud optimized for your specific needs, that works only for you.
Just like businesses love dedicated servers even when a shared server would do fine, businesses also love dedicated clouds.
Cloud providers need to think again about what and to whom they are selling. I see a market for super cheap hosting for the masses by selling competitive hosting packages by leveraging the cost efficiency and performance benefits of a cloud. I also see a market for dedicated custom cloud solutions for the high end market. However, both of these markets are extremely saturated, and if you are not selling anything new, you are primarily competing by price alone. Any such market is a lot of hard work for not so much money.
So good luck! PS. I am not buying.
Having read through this article server utilization is the most important factor driving better economics for the cloud :
"Our analysis assumes an average utilization rate of 12 percent of available CPU capacity in the SQ environment and 60 percent in the virtualized cloud scenarios."
(SQ means status quo, i.e., non-cloud.) This factor of 5 improvement in average utilization drives the overall cost savings and they are assuming a cloud overhead of about 45%. (I.e., if you look that their numbers, they assume that cloud CPU cycles cost 45% more than local cycles, but the efficiency is 5 times higher, for a overall cost reduction of a factor of 3.4 in the "public" cloud case, which has the largest savings.)
A factor of 5 in server utilization is huge; the question is, is it realistic ? Note that 60% usage corresponds to 100% usage for 14 hours per day, 7 days a week, or 20 hours of full usage for 5 days per week, and so would be quite high for a government web site. If government web servers dominate the cloud computing, the savings are likely not to be as large as this study supposes, because no amount of aggregation of government web site servers will get you much traffic in the middle of the night.
If you think about it, to be economically effective cloud computing (in the big picture) has to be about saving money by increasing average server utilization (averaged over all users). Cloud servers are not free, and require resources to service and maintain, and clouds have overhead. If some service is barely loading a single server, sure, I can see it being cheaper in the cloud. If servers are maxxed out almost all of the time, I bet that the cloud won't save much money. If the aggregate use is highly time variable, the cloud will not save as much money as a simple calculation would indicate, as the cloud will have servers sitting idle during off hours. For this particular article, its hard to say more as they don't reveal their actual data.