In the Live Monitor tab, you are observing spikes in the Average Response Time per Server graph when there is only a handful of queries running, thus you would like to understand how this graph works and if there is a correlation between the number of queries and the spikes in the graph.
The Average Response Time per Server graph plots the average of the time taken for reading the first byte of response across all queries processed in that second, which approximates the time taken for the query to process on the server. This value is directly consumed by ScaleArc's dynamic load-balancing logic to determine the load on a particular server and change the traffic distribution among servers.
When there is only a small number of queries running in the database server, spikes can be caused by an infrequent slow query executed at that moment increasing the average temporarily and do not necessarily mean that the database server is slow or loaded. On a database server with a large number of queries executing simultaneously, the average will be better distributed with less probability of spiking. In the case of the latter, the metric will be a better indicator of overall server response time which is used by ScaleArc as a load level indicator.
For additional information on this and other Cluster Stats graphs, please refer to the Live Monitor - Cluster Stats article.