In most organization, the IT function is almost always a key enabler for an effective and efficient running of the business. With the recent credit crunch and economic downturn, many a times, IT departments find themselves in a dilemma. On one hand, they are faced with the need to refresh EOL technologies, expand capacity to support the business or add new assets for new application services required by the business. But on the other hand, they are often forced to cut budgets, headcount freeze, reduction in force, faced with restraint in funding and capital investment for any refresh or expansion. This dilemma pushes IT leaders to begin taking a closer look at the technology services that they are providing to their business units (i.e. internal customers) and evaluating where it makes sense to convert these as Shared Services with chargeback of the costs to the business units.
IT Shared Services refers to the provision of IT services by an IT organization where that service had previously been found in more than one part of the organization or group. Thus the funding and resourcing of the service is shared and the providing department effectively becomes an internal service provider. The key is the idea of ’sharing’ within an organization or group. Hence, the concept of IT Shared Services is similar to collaboration. IT Shared Services is different from the diametrically opposite model of Outsourcing which is where an external third party is paid to provide a service that was previously internal to the buying organization, typically leading to redundancies and re-organization. One purpose of Shared Services is the convergence and streamlining of an organization’s functions to ensure that they deliver to the organization the services required of them as effectively and efficiently as possible. This concept is applicable not just for the IT services, but can also involve the centralizing of back office functions such as HR, Finance, and middle or front offices.
IT Shared Services are more than just centralization or consolidation of similar activities in one location. Shared Services can mean running these service activities like a business and delivering services to internal customers at a cost, quality and timeliness that is competitive with alternatives. It also means being able to recover the costs of operating and supporting the IT services to the business. In the traditional IT landscape before virtualization, it is normal for each application to have its dedicated servers. In such cases, it is easy to do cost recovery. But with the introduction of virtualization, cost recovery gets a bit complicated. Furthermore, with virtualization, if we do not implement a chargeback model, it is very easy for the business units to take IT services for granted and inadvertently causing a “VM sprawl”. 
So, the drivers for introducing a chargeback model are:
- To manage and control VM sprawl – what is available for free, will be taken for granted and may never be released. If you charge the business units for use of VMs, they will learn to use it sparingly and release it when it is not required anymore.
- Helps to transform the budgeting process for IT departments into a business process instead of just “playing around with numbers”. It helps to derive and build the business justification for investing in a new datacenter, technology refresh or investment.
There are various methods of implementing chargeback model for IT services. For example, businesses can be charged a fixed cost for a basic VM (Virtual Machine) with a predefined set of compute resources, or at a more granular level, it is also possible to do a utility-based charging model where businesses can be charged on how much resources they consume over time (for example, CPU usage by hour, storage usage by GB, etc.). Or it is possible to develop a hybrid model where there is a fixed cost for a basic VM but any usage beyond the predefined set of compute resources will incur additional costs to the user (similar to burstable network bandwidth).
As a first step, it is important to plan and define a Service Model, which will subsequently lead towards the design and development an IT service catalog that will provide a list of offered IT services to the business. A Service Model should provide a framework of classification for IT services. For example, such a Service Model can be formalized with the following category of IT services:
- Business Application Services
- Consumer Services
- Enabling Services
Business Application Services category would be to classify and contain all major business and mission critical applications supporting the businesses. Consumer services category can be used to classify all customer-facing services and business applications used directly by employees, or the end-user community. Examples of these includes email, file & print, etc. Then, finally is the category of Enabling Services which provide support to some or all of the consumer services or business application services. The WAN, LAN, tape backup services, etc. are examples of enabling services.
Once the Service Model is formulated, then the next step is to define the IT service cost components and cost visibility under each of the category of the Service Model. One way of doing this is to breakdown each of the services’ cost components into consumables and non-consumables group:
- Consumables include:
- Server hardware or systems compute platform based on Standard Operating Environment (SOE), e.g. LDOMs, Solaris Containers, or VM units based on SOEs (e.g. Windows 2008 Server) with predefined sets of compute resources (CPU, Memory, Internal Storage, etc).
- Storage usage based on tiered storage and utilization (e.g. Amount of capacity in GB).
- SAN fabric usage based on number of FC port(s).
- Network usage.
- Applications
- Non-consumables include:
- System administration, database administration, network administration and other support fees.
- Floor space (facility).
- License costs, etc.
This result will then constitute into a classification of the IT service cost components for each category of the Service Model. Next, in order to develop an effective Service Model and capacity planning, it is advisable to determine the demand and consumption behavior or drivers. Understanding this behavior, such as elasticity of the demand, pattern, duration of usage, required service levels, etc., will help to formulate the chargeback model’s method and calculation of unit of measurement for charging.
The following step is to formulate the cost pools for each service category in the Service Model and allocating the costs to its appropriate ‘buckets’. When the cost pools are established, populated and finalized, then we will have cost visibility of each service category, i.e. how much it costs to deliver that particular IT service. It will be rather easy to formulate the cost pools for Business Application Services and Consumer Services. However, for Enabling Services (because of its shared nature), a decision needs to be taken on how to recover the costs for Enabling Services. One way is to divide equally between the number of users, e.g. WAN costs for a point-to-point connectivity, will be divided equally between the two sites and apportioned to the business units at each site depending on the number of employees in their respective departments. Another approach is to allocate the costs of Enabling Services to the Consumer and Business Application Services. For example, by analyzing WAN traffic, it is possible to allocate WAN costs to the top ten services that account for most of the usage of that bandwidth. However, in this approach, pricing for each IT services may have to be revised and re-calculated on a periodic basis, as usage patterns can be dynamic.
The IT services catalog items above can also be structured into multiple tiers according to the required service level. For instance, a Solaris Container for 98% service availability will be of a different cost comparing to a Solaris Container for 99.9% service availability.
Finally, in terms of how to derive a unit price for chargeback, there are many ways to price it. It depends on the demand and consumption behavior. Some of the common ways are:
- Fixed Pricing: This method charges a fixed pricing for each unit of service used. The unit may be a fixed and predetermined configuration, for example, a VM instance with a fixed CPU, memory and storage allocation. The advantage of this method is that it is easy to implement, easy to measure and carry out the chargeback but it does not take into consideration of unutilized resources within that predetermined configuration or if there are additional ad-hoc demand.
- Variable / Pay-per-used Pricing: This method is a true utility based pricing where users pay for exactly the resource that they use. It defines rates for consumable resources such as CPU, memory, storage, network, etc., and non-consumables, and usage is charged accordingly. This method is fair, however, it is very complex and will be complicated to implement.
- Hybrid (Fixed with Variable): This method is a hybrid of fixed and variable based pricing. Each service can be predefined with a fixed base configuration at a fixed price, and should the usage requirement expand beyond this base configuration, then any additional delta would be levied according to the variable rates.
- User-based Pricing: For an environment where user management is more costlier than hardware usage, then a method that will meter usage by person might be more appropriate. Situation such as insurance agents connecting to a portal for most part of each day, and for fairly similar periods of time, for almost the same usage patterns, then it may be fair and easy way to charge by per user rather than putting in sophisticated method to measure usage. This method is easier and simpler to implement. However, this method assumes each user will have almost the same usage demands and patterns.
- Peak-level Pricing: Some services may have a lot of volatility in terms of demands and usage pattern over each hour or even within the hour. Peaks and troughs may have a huge gap even in a short period of time. The peak-level method takes the variable model and adds a mechanism to monitor and record peak consumption. Consumers are then billed according to their peak use, and not according to their average use. As it is only required to measure peak-level usage, so metering is simpler. However, the disadvantage is that it assumes the peaks are generally predictable peaks over a given period of time and not just irrational or sporadic occasional peaks. Policies which are easy to understand and communicated to the users must be a pre-requisite prior to establishing this method of pricing.
- Ticket-based Pricing: For IT environments where quality of service is critical, IT usage can metered and controlled tightly using electronic “tickets” that use a short validity period (say four hours). This method of using tickets allows IT to control system load to a very fine degree and regulate the consumption. Such regulation helps to eliminate usage peaks and ensure business continuity. For the ticket-based method to operate effectively, it is often necessary to implement “use-by” dates on tickets to avoid stockpiling. This means that each ticket should have a validity date / expiry date to prevent ticket hoarding.
Chargeback is a way to put IT services in the same terms that business people understand and value. When IT services is bought and consumed like other services, IT can become a business within the business. And that is the path to true IT value. That is the path that IT can justify its existence in the organization.
Tags: Business, Chargeback, Consumers, Cost, Data Center, Service Management, services, Virtualization