|Close Help Window|
|Overview||This model illustrates the relationship between various demand curves, capacity strategies addressing those types of demand singly or in combination, and the economics of tradeoffs. Is it better to use solely static capacity engineered to peak demand? Or to use less static capacity and accept penalties associated with inability to meet peak requirements (e.g., loss of revenue associated with unserved customers)? Or use purely utility resources to meet all demand? Or to use a hybrid approach? The answer is: it depends, and this model shows the relationships between the different factors that would lead to an optimal selection.|
|Basic Flow||First, demand is characterized using the Demand Shaping window. It may be uniform, worker, gamer, event, cyclical, or growth -- or a combination
of some or all of these. For example, worker demand occurs from 9:00 AM to 5:00 PM, Monday through Friday. In any of the hour time slots, the demand falls between
the minimum and maximum number of servers.
Second, the resulting demand curve may be viewed in either the Annual or Weekly Demand windows. The Weekly Demand window is a bar chart of every single hour of a one week period. The Annual Demand window shows the maximum, minimum, and or average hourly demand for each of the 365 days in the year.
Third, a variety of capacity strategies may be tried. By allowing additions and reductions to capacity, it may be possible to reduce the expenses associated with fixed capacity. The challenge is that too little capacity may lead to some demand not being served, which has a financial penalty associated with it. And, changing capacity may incur charges.
Fourth, cost assumptions may be modified. Whether costs for fixed capacity, costs for capacity modifications (installs or decommissioning), opportunity costs of unserved demand, or costs for utility servers.
Fifth, statistics may be viewed, e.g., minimum demand, average demand, unserved demand hours.
Lastly, the relative costs and underlying drivers of four different types of architectures / business models may be viewed as a bar chart.
|Weekly and Annual Demand||
The annual and weekly demand may be viewed. Clicking on a daily demand bar or entering a week number changes the week that is "zoomed in." Hovering over a demand bar shows the time slot and information associated with it. Hovering over the blue capacity graph shows the capacity level and the duration it is in effect.
A variety of capacity strategies may be attempted. The initial capacity and minimum capacity may be adjusted. Capacity increases may be allowed, based on recent peak, average, or minimum demand values. Delay times for placing the capacity "order" may be adjusted. Separately, capacity decreases may be similarly allowed. However, there are costs associated with having capacity (in effect, monthly lease costs or depreciation), capacity increases (installation and turn-up), or capacity decreases (e.g., transportation, removal, reconditioning). Different types of demand curves may be more or less amenable to different capacity strategies. Flat demand, with little variance (min close to max) can effectively be supported by a fixed capacity strategy. Steady growth can be amenable to relatively infrequent upgrades which may be planned in advance. Cyclical requirements can be met by alternating increases and decreases. However, trying to rapidly respond to random demand changes, like buying stocks after an up day in the market and selling them after a down day can be counterproductive. If response time is too slow, capacity may be being decreased after demand has begun to increase. And, trying to follow every twist and turn in the demand curve can cause churning of the capacity, incurring excessive charges.
|Cost Assumptions||Cost assumptions are key to deriving useful results from the model. Increasing or decreasing capacity in an enterprise data center is not likely to be free, nor instantaneous. While it is relatively rapid in a utility environment, such as a service provider cloud, it should be assumed to be more expensive than fixed capacity, on a per hour basis. This matches real world premiums typically associated with a utility. For example, a rental car costs more per day than owning a car (per day charges). A service provider can create a shared utility by statistically multiplexing multiple demand sources onto a resource, but whatever that resource costs is just the baseline. Also added in must be a capacity allocation manager, infrastructure upgrades, margin, customer acquisition costs, utilization "breakage," and so forth. Consequently, the optimization challenge is to figure out the right balance between fixed resources, which cost money whether they are used or not, and utility resources, which cost more when they are used, but cost zero when not used. Another important cost assumption is the opportunity cost of unserved demand. If it costs nothing to have angry customers for whom there was no service capacity, then it is cheapest to minimize fixed capacity. A key assumption must be that each unit-hour of fulfilled demand creates revenue or other value well beyond the cost to ensure the availability of capacity to so serve that demand.|
A variety of statistics are automatically generated each time any assumptions or inputs change. These include things like the maximum demand, the minimum demand, the average demand, and how many unit-hours of demand would be unserved due to capacity shortfalls. Also, total charges are displayed for four different models.
A pure pro forma fixed capacity model. This assumes unvarying capacity as high as peak demand.
A variable capacity model, with either capacity increases, decreases or both. Deltas may save capacity costs, but also run the danger of creating additional cost through unserved demand.
A pure utility model. The capacity model (fixed or variable) is ignored, and unbounded resources are assumed to be available in the cloud, but a premium is paid for each unit-hour of capacity.
A hybrid model. Capacity as specified is used and charges incurred, AND utility resources are leveraged to ensure that there is no unserved demand.
|Total Charge Visual Comparison||
The pure pro forma fixed capacity model, the variable capacity model, the pure utility model, and the hybrid model are displayed graphically so that they may be compared. Adjusting capacity strategies, demand curves, and/or cost assumptions alter the relative costs.