Cloud Computing

Cloud computing has become a hype today in the IT world. No doubt there are lot of attractive features talked about for which people will be interested. There are two  types of cloud infrastructure which are available today, i.e. private cloud and public cloud.

Private cloud is one in which the CPUs are dedicated for the user and optimization is being done through virtualisation within the limited capacity of CPUs which form a pool for virtualisation. In this case anyways the user is paying for all the CPUs so private cloud helps primarily to optimize your existing resources through virtualisation. There is apparently no difference to a situation when you host your servers in data center of a service provider, implement virtualisation and have option of deferred payment in opex model. But some organisations are there who might capitalise those investments and in those cases avoidance of capex really  does not work. However, it does not provide any substantial reduction in CPU cost other than providing the option of reduction of cost through optimization.

Public cloud on the other hand addresses some of the above concerns and can provide better cost optimization as the CPUs are shared by larger group and you pay as per use (though there are baselines which need to be committed). However the larger issue is in terms of security measures that your organisation will like to implement. Though cloud service providers provide standard security measures as mentioned in some of the standards, specific requirements of your organisations might not be fully addressed as this is being shared by larger group. Even in terms of SLAs there can be committments but to what extent those SLAs are suitable for you as well as whether the service provider is taking penalty clause for not meeting the committed SLAs is something which needs to be ensured.

Looking at both private and public cloud options probably a hybrid model is most suitable which any organisation will prefer. While organisations own private CPUs, these can be shared with others through grid so that your unused CPU at times is being used by others and when you require higher computation you can use unused CPUs of other organisations. There needs to be a monitoring mechanism where for your CPUs being used by others you get paid for it and when you use CPUs of others you pay for it. A simplified model to handle this across grid can be worked out with some central agencies who manage the grids. This can provide cost optimization with optimization of CPU usage ,and you can implement all security measures which your organisation requires to follow. You can have your own SLAs with your service provider. This helps you to operate in a secured environment with topping up of additional CPUs through grid when required and pay as per use and also get paid when you do not require some of your CPUs. This is probably a win-win situation for all through optimal utilization worldwide. However software as a service (SAS) is a different model and we will discuss over a different blog.