Estimating Cost Of Running The Web Application On Windows Azure

In this section we use December 2011 prices. These will serve as the fundamentals for the following computation.

Important
The prices used for calculations in this article may not correspond to the most recent prices. For the latest prices refer to Windows Azure Pricing Overview.

Note

The key concept here is the pattern that can be overlaid on top of the costs for usage of each service: there is a meter that defines the measured usage, and a cost per meter unit captured by the meter. Taking Windows Azure Compute as an example, the extended cost for running a Medium instance for 24 hours is calculated as: Meter x Cost = 24 (Medium Instance) Hours x $0.24 per meter unit = $5.76. We will use this notion of a meter unit and a cost per meter unit throughout.

The tables below show the meter costs per service. The services and costs are broken into three groups. The first group shows basic services and costs. These are the basic Windows Azure costs that you will most likely use for any Windows Azure application, and will probably make up the greatest share of your expense. Use the costs in the first group to get a first approximation of the major part of your costs.

The second group comes into play as you actively plan your application. You need some understanding of how the Service Bus relays and brokered messages work before you can leverage their features. That understanding may result in a much more efficient application, which will save you money.

The third group consists of bandwidth costs. As the application progresses, you will have a better understanding of how much bandwidth you will use. Since bandwidth only affects data egress (outbound data), you will want to have that fact in mind as you develop your application. For example, paging data logically enables you to measure it more easily (do not present query results in one, uncategorized page).

Basic Services and Costs

Secondary Services and Costs

Secondary Services and Costs

In order to provide accurate estimates that are more in line with usage you need to pro-rate the aforementioned/above Windows Azure prices to the smallest/applicable time slice possible. For most services with a time-based meter in Windows Azure this can be computed hourly. The key to pro-rating generally across all Windows Azure services, is to take special measures to handle minimum charges, such as in the case when the smallest billed meter increment is daily and billing for just one or two hours of use results in being billed for the full day.

The following tables are all adjusted from the published costs to reflect hourly pro-rated charges according to the billing meter.

There are various offers available that effectively discount the use of various Windows Azure services. The following lists the offers currently available to MSDN subscribers, Microsoft Partners and Start-up Community (BizSpark), as well as those discounts made available through trial or commitment offers.

MSDN Subscribers

The current offer to MSDN subscribers provides an annual benefit valued at $3,700 which includes the quantities described the following table.

Microsoft Partner Network Members

Bizspark Members

3 Month Trial

Six Month Commitment Plans

Enterprise Agreements

With the basics of computing costs based on usage, the next step is to help the customer to estimate the amount of consumed resources, such as:

• Fixed or variable (elastic) resource consumption.
  
  
• Consumption composition (what is fixed, what is variable).
  
  

In general, it is very difficult for the customer to answer these questions without having a live site collecting the desired metrics:

• How many storage transactions will you perform in a month?
  
  
• How many messages will you send per month across the Service Bus?
  
  
• What size of compute instance and how many instances of each will you need?
  
  
• How large of a cache will you need?
  
  

Defining Usage Scenarios

Supporting Elasticity in the Usage Estimation Model

Extensibility in the Model

You can estimate Windows Azure costs using the following algorithm.

1. Estimate the usage for a single service.
   
   
2. Apply the estimated usage from a scenario to allocate quantities of the meter or directly specify the quantity of the meter used. Optionally, vary the usage according to elastic needs (for example, on an hourly scale).
   
   
3. Apply discounts, if any apply, from enterprise agreements, special offers, or commitment plans.
   
   
4. Repeat for each service involved in the solution.
   
   
5. Sum the Meter x Cost for each service involved in the solution:
   
   
   1. Roll up hours into days, taking care to apply daily minimums where required.
      
      
   2. Roll up days into a reference 30 day or 720 hour month.
      
      

In the coming months, we hope to take this general algorithm in combination with reference scenario data and produce a web based solution that provides interactive cost modeling for Windows Azure customers, supporting models with high degrees of flexibility and moderate complexity.

In this article we provided some basic understanding of the Windows Azure pricing, specifically by including details that are either not described on public price lists, are only available by calling Microsoft sales reps, or have relevant details that simply are not spelled out. In addition, we provided insights into an algorithm you can implement today to start modeling your costs. The same methodology will likely be employed by a tool in the near future.