Basically, no! In the financial world we are used to seeing advertisements for investments with a cautionary note at the bottom:
Warning: Past performance is not necessarily a guide as to future performance
Would your IT Director be happy to take your hardware investment advice if it had the same proviso attached? Unlikely. Too much rests on the decision to get it wrong: budget, user satisfaction, business performance, his job.
Trends require a history of past performance that will be valid for future scenarios.
Trends are based on a single metric.
Trends are based on utilization not service level.
If you are consolidating workloads across servers you are entering a new world so there is no valid history to point to likely future performance. Estimating which workloads can co-exist based on utilization misses the vital fact of how these workloads will interact when they compete for finite resources (processor, memory, disc...) on a consolidated configuration. This interaction might mean that two workloads whose combined utilization adds up to a nice low figure might both suffer a disastrous drop in service level provided because of contention for these resources.
Trending has its place for general performance analysis, prediction for relatively stable workloads, a general view of the effect of growth in the use of existing applications and when correlating business and observed IT performance statistics.
If the cost of failure of a consolidation exercise is significant to your business however, trending is too risky a means of assessing what consolidated resources you will need.
More sophisticated planning techniques such as benchmarking and modeling are the only ways of accounting for the competition for resources and identifying accurately what resources you require.
Next Management Performance Tip |
