Deep Recurrent Model for Server Load and Performance Prediction in Data Center

Figure 2

The network structure for two models: (a) model for workload prediction and (b) model for performance prediction. Model (a) takes request vectors as the input and output of the network, using the latest requests to predict the th request. The input of model (b) is the latest feature vectors extracted from the log of every time-window, and its output is the value of certain kind of performance in time-window.

(a) Model for workload prediction

(b) Model for performance prediction