Monthly Archives: January 2009

There are many debates about the economics of cloud. The reality is that you at some point you are going to have to make your own mind up, but it is useful to have some information to use internally if you wish to make a case to your business to be a cloud consumer.


Firstly, Cloud moves costs from a CAPEX model to an OPEX model. A capital expenditure is incurred when a business spends money either to buy fixed assets or to add to the value of an existing fixed asset with a useful life that extends beyond the taxable year. Operating Expenditure is is an on-going cost for running a product, business, or system. The simple difference is rather than accruing hardware and software costs immediately and having ongoing operating costs for rent, power, heat, light etc the hardware, software, rent, power, heat, light etc all become ongoing operating costs.

This can have several advantages:

Ok, all that is great but what about the real economics of cloud. We often read / hear that using the cloud is “cheaper”. This is always going to only have applicability on a case by case basis but we can generally investigate the cheaper claim.A Microsoft employee blogged using some assumptive information about the costs to run a 50,000 node data centre . The summary of these costs are outlined below:

Having a 50,000 node datacentre may not be unusual for an enterprise organisation, but the interesting statistic fro this is the monthly cost of running a single server which we can assimilate from these statistics as being $112,42.


It’s not clear from the data what types of server this represents or whether the server has an OS pre-loaded, but it gives a reasonable, if imprecise, indicator of the monthly net cost of running a server . We could expect this net cost to be multiplied by a factor of 4 to 5 times if the server were to be offered to a consumer.


If we now look at Amazon EC2 and the retail costs of running a cloud server:


EC2 pricing starts for what Amazon call a “small instance”. A small instance is equivalent to a single AMD 1.2 Ghz and pricing is:

10c/h. 0.1 * 24 * 30 = $72/month + bandwidth + storage


A small instance is probably not equivalent to our data centre server. This is likely to be an extra large EC2 instance i.e. a 1 CPU 4 core machine. Pricing for this is:

80c/h. 0.8 * 24 * 30 = $576/month + bandwidth + storage


This is 5 times the estimated cost of running a single server from a 50,000 node data centre, so I’d say given margins for error in estimation and Retail Cost V Net Cost, this about works out i.e. Cloud users pay a Retail cost for being able to use Cloud server. It is not cheaper necessarily than data centre costs. It’s not a like for like comparison but it shows that using Amazon probably stacks up well for infrastructure pricing compared to the alternative of using in-house servers.


Infrastructure costs are only one cost however. What about software ? Using Open Source is one thing, but what about if we wanted to use enterprise software ? Lets take a look at one of the leading middleware cloud providers, GigaSpaces.

One of the reason to looks at GigaSpaces is that they are one of the only Middleware companies to offer subscription pricing as well as perpetual pricing and cloud pricing for their software, and this allows us to make a much better “like for like” comparative analysis.


Lets take a scenario where we iare nterested in using GigaSpaces on 4*4 core server instances. If we were to take a subscription of GigaSpaces to be used in-house then this would be this would be:


GigaSpace subscription cost: $7500 per cpu per year for 4 * 4 core servers this would be equivalent of 8 cpu’s or $60,000 per year. You can validate this at gigaspaces.com/licensing (actually GigaSpaces offer a ’starter’ subscription package, but this is not applicable in this case).


Lets compare this to using GigaSpaces on EC2. One EC2 instance is equal to 1.2 GHz AMD CPU. Using 4 extra large EC2 instances would gives 4 EC2 compute units (2 cores and 2 instances).


The charge for each large GigaSpaces instance is .80 cents per hour. For a year this works out as:

- $1.6 *7 *24 = $268.8 for 1 week
- or $13977.60 per year per EC2 instance
- for 4 large instances this is 13977.6 *4 = $55910.4


The in-house subscription cost has support built in whereas there is a $5000 charge for the equivalent Gold support for Cloud. This makes deploying to the cloud using subscription and deploying internally using subscription costs approximately the same. You can validate the pricing at http://www.gigaspaces.com/ec2-pricing


Again it seems that using the cloud is not necessarily cheaper. So where is this notion of cloud being cheaper come from ? It comes from the utility compute or compute on demand model. You only pay for what you use and when on the Cloud so in reality using the Cloud can work out much cheaper because:

Ultimately it is a trade off. You need to look at security, latency, costs implications etc (see the next section on cloud best practice) and make a choice as to whether the cloud makes sense for you and your organisation.

This is particular blog post is a personal opinion and no way associated with my role at GigaSpaces and is taken from my forthcoming book: “TheSavvyGuideTo Grid, HPC, DataGrid, Virtualisation and Cloud Computing” which is due to be released in February.

Continue reading →