|
Clouds, Grids, shared infrastructures or any other of such technologies aims at lowering the cost of ownership for the end-users. There are fundamental questions that still need to be answered from a service provider and service user stand point:
What is it that I am paying for? Or charging for?
Think about it – are you paying for CPU? Then should you – the CPU-intensive app user – pay the same amount as the user who caps the CPU at 10%? Should you pay for bandwidth even if you are utilizing it at less than 5%? What is fair? Can we be fair? If you are a service provider, you aim to recover the cost – and maybe even make some profit in the process. If you are a service user, you want to save money as it affects your bottom line. So what is the right answer? Is there a right answer?
I know I just jumped into the fire here with this topic, but these are some of the main questions that need to be answered by you the CIO, by you the user, and by you, the person who is signing the check. So what’s it going to be? What will make you happy?
I am afraid that not only we do not have the answers to these questions, we are not even sure if we are asking the right questions. The fact of the matter is that with the increase of CPU power, number of cores, etc, some of these questions are becoming more relevant – some already are. Most organizations used to charge for CPU usage without prorating for utilization. The same was true for network and memory usage (I will cover memory usage in a later post as I think memory usage is a very interesting topic). With the increase in the number of cores and applications that can take advantage of these extra cores, most organizations are rethinking this strategy and are interested in a more fine-grained usage reports. If an application started on one core of a multi-core machine, and it started to use the other cores, it needs to be charged for them as well and not just one the core.
The other school of thought – from the use’s perspective I bet – argues that if a core is under powered, they should not be charged for the full amount of usage. If an application requires 8GB of RAM to process data , but only 4GB is available, then the trashing and disk access must be taken into consideration and the charge back must be reflect that.
These are just some of the topics that we will discuss – I wanted to touch on them and start laying out questions that we can refer to at a later point. W have a lot to cover, and I will cover these points and many others ; feel free to drop me a note and suggest a topic or question.
Only registered users can write comments.
Please login or register. |
