Organizations face more options for where to place their workloads than ever before. I'm constantly getting questions about where the next workload should be hosted. IT decision makers now face the following platform choices:
- Mainframes - powerful, complex, extremely reliable, extremely scalable.
- Midrange systems - powerful, complex, highly reliable, highly scalable.
- Industry standard systems - powerful, getting more complex by the minute, reliable, scalable.
- Distributed solutions - support multitier applications, as scalable as the budget allows, complexity growing by leaps and bounds.
- Virtual systems - mainframes and midrange systems have supported virtual systems for decades. Industry standard systems joined the party about 10 years ago. This can reduce hardware complexity in exchange for increaed software complexity. Scalability and reliability can be enhanced though the use of the proper technology.
- Cloud-based systems - combine the joys of virtual systems, industry standard systems, outsourced data centers and a pay-as-you-go usage model
I'm often asked to help IT decision makers make sense of all of the possible combinations. Since these are highly complex, company dependent considerations, there isn't a single right answer to the question "where should I host this workload?".
Here are some rules of thumb:
- Consider staff expertise before selecting a platform. Bringing something new into the data center can cause a cascade of staff-related costs.
- Consider the requirements for agility, scalability, reliability when selecting a platform.
- Consider software and hardware lock-ins. It may be entirely worthwhile to accept a locked-in solution if it is the best choice for the task at hand. Just know when the organization is donning shackles that might pinch and chafe in coming years.
- Is the organization buying or building the application? The choice of application software, development tools or database software often drives the selection of the hardware platform.
- It is better to have an overall architecture in mind rather than
allowing ad hoc selections of platforms, operating systems, development
tools, application frameworks, database software and the like. Building
your own in-house Tower of Babel might be fun during the building
process, but is going to be painful later.