IFR in the Clouds: True Heading & Cloud Computing

If there is a part one, there is a part two.

in the first Long Tail blog entry I  described the influence of Cloud Computing to the revenue streams of a ISP or an Internal-IT / data center.

Today we will discuss the usage of Cloud data store, like Amazon S3 services, using the same Pareto model.

The typical pricing model for a data store is a combination of

3 different pricing factors:

1. Price per GB uses. This is somehow in a range from 0.180 $ (Dollar ?? for Europe ??)
2. Price per GB transferred out. Range from $0.170 per GB – first 10 TB / month data transfer.
3. Price per request. Range from $0.012 per 1,000 PUT, COPY, POST, or LIST requests

Here is my second Long Tail recommendation:

x-axis : Describes the amount of computing = CPU power a given service/application needs per storage request. To the left we find applications with almost zero CPU requirements like file services. To the right applications get more and more CPU bound, like ERP, Mail etc. 

y-axis : This is the combined costing out of Storage cost, data transaction cost and service request cost (PUT,GET etc.).

The higher, the more total access cost will occur when the application is in use. It is the cost function out of the  3 different pricing factors.

The Head = The Red Head

These are data that are frequently used and that require high up/download bandwidth.

Typical scenarios are:

• videos
  
• high quality picture / photo storage
  

The Long Tail = The Long Green Tail

Here you find data that is probably not used frequently or that has limited storage requirements. These applications are in general a good fit for the Cloud Space.

My recommendation

• Keep in mind that there are 2-3 price components
• Keep data with very high access rates in your own data center
  
• Focus on the Long Tail. The Cloud is a perfect place for this kind of data
• Think about legal or Compliance rules. This is important for Europe, probably less in the US.
• Keep an eye on the workload distribution
  

Other influence factors

Workload distribution is the next big decision factor. If you have a quite good balanced access to the data during a month, the above recommendations will

work fine. The infrastructure cost are somehow balanced and you only need to keep out of the red = Head area.

Things can get different, if you have  workload hot spots. Or if Hot spots are your main business. On CloudCamp I had the chance to talk to companies that have only

a few hot spots per year. Some web-pages are only created for special events like the Olympics or other sports events.

This is a clear indication for high flexibility requirements: Use Cloud Computing. No question.

Thanks to Jan Brosowski, who helped me get the x/y coordinates right, as I was lost in Pareto space.........