Tuesday, April 28, 2009

Demand Planning for Cloud Computing

One of the key characteristics I learned about Cloud Computing in the SIPA Cloud computing conference was making the resources appear infinite for the users. That is when a customer subscribes for a service like storage space he should have access to additional storage space immediately on demand. This will be the flexibility expected out of IaaS (Infrastructure as a Service). Similarly other clouds such as Software aaS and Platform aaS will need flexibility.
For all of us in the supply chain field, we know that infinite capacity is not really 'infinite' but a perception. So, it is actually finite and needs a careful planning to predict the demand, thus needs an accurate demand planning, followed by careful execution.
Establishing the need for a demand planning solution for this situation needs another consideration - cost of planning / inventory. May be, for example, currently storage is not very expensive, but when the competition of IaaS providers increase, then price will drop bringing down the margin. Then there will be a rush to develop a good planning solution for forecasting, procuring, and use of the resource.
So, what is special about this industry that supply chain solution providers should target?
Some of my thoughts with examples
Resource providers for mobile application hosting, like Iphone or gphone apps will see sudden increase in the requirements as soon as the application becomes popular. This is something similar to fashion industry, where demand increase depends on fashion hits. But unlike fashion industry, the initial requirement will be very small until it becomes a hit. This is primarily due to the fact that cost of procurement is very small for clouds than fashion.
On the outset, it seems like locational demography don't play important role in a cloud environment, so demand from any location can be supplied by any other location. This is another key characteristic of clouds. But, natural disasters and communal forces do play important role in determining the availability of these virtual clouds. So, need for backup and second level procurement options are necessary for supplying the cloud. Also, demand forecasting needs seasonal models for accuracy.  
There is always a comparison between utility industry and the cloud environment, because clouds show many similarities - subscription model, virtual source, spikes in demand, etc. So, can utility demand management work for clouds then? will we see grids for clouds where multiple providers will supply to pool of demands? This will then need good supply network management with distribution centers (grids) that will help cushion demand peaks/ supply troughs. This will also create a marketplace for such providers where procurements will happen instantly, and on demand. Multiple providers will queue up to store that next byte of the customer, process the next payment, or broker the next cloud requirement.
As the clouds evolve, we will see increasing requirement for accurate planning, and this planning will direct the evolution into a direction that will create a new marketplace that look like ebay on caffeine IV with demand and supply exchanging hands within split seconds.           

No comments:

Post a Comment