Paying by the drink
Apr 2nd, 2008 by Rick Barnard
The parallel between computing as a true utility (like water or electricity) has often been made, but seldom realized. In the past, utility offerings have been more like financial models by offering computing capacity on demand…when you turn it on, you start paying for it. But that is just a financial model (not a true utility) – what happens if you turn off the water…what happens if you only drink 15% of the glass? What do you do with the other 85% of the glass, pour it down the drain?
Whether IT is a profit or cost center, the end goal is to deliver service levels at the lowest cost. For the customer (department, user, etc.), the ideal situation would be to only pay for the resources that are consumed, no more, no less (well, maybe less). For the IT service provider, the key is to use every available resource and not waste any capacity. If compute capacity is available, allocate it to a paying customer.
So, how do we get to this end point?
- Share! Use virtualization to share computing resources in order to drive down average costs and allocate available resources to paying customers (sharing cost of computing across more customers)
- Repurpose servers for multiple sets of applications
- Share high availability resources (e.g. N+1 HA)
- Repurpose servers for different environments (e.g. share test/dev with DR)
- Right size applications dynamically so servers are fully utilized at all time (scale-up or scale-out)
- Consolidate the maximum number of applications on the same server with built-in high availability and simplified capacity management
2. Monitor! Monitor and measure resource utilization, health, events, etc.
- Measure events and track available resources
- Monitor the health of hardware, providing automatic high availability thus eliminating the need for passive and redundant servers
- Collect information and take specific actions based on business policies
- Pass resource information to analysis and reporting tools
3. Automate! Automate and optimize the use of resources based on processes and business policies.
- Set traps, controls and triggers to collect information and/or take specific actions based on business policies
- Use as much of the available resources for an application or paying customer
- Respond in real time in order to provide the required service levels to the organization
- Set policies that allow business costs to be variable based on actual usage
4. Report! Report on the use of resources so business costs are variable based on actual usage.
- Collect resource information, which is passed to third-party reporting tools
- Report on the duration of use for physical servers that may be repurposed over time
If we now pay by the drink, what are the real cost drivers? Are costs driven by what is consumed, or how the resource is managed and made available to the user?
How about another trip to the water cooler?