Next we investigate the impact that error resulting from relaxed measurement intervals would have on the decisions made by a service placement infrastructure. Figures 14 and 15 already contain the answer to this question, as there is an analogy between acceptable migration interval and acceptable data staleness. For example, consider collecting measurements every 30 minutes. If a migration decision is made at the same time as the data is collected, then on average the quality of the sliver-to-node mapping will be the value of the curve at t=0. If a migration decision is made 5 minutes later, but using the same measurements, then the quality of the sliver-to-node matching will be the value of t=5. And so on, up to t=29. This analogy leads us to a similar conclusion regarding stale data as we made regarding migration interval, namely that data staleness up to 30 minutes is acceptable for making migration decisions for this workload.