LSTM forecasting on multivariate time series

I’m new to RNNs and LSTM and would like some direction with a problem I have. I have a data set containing system metrics (like CPU utilization, disk operations, memory use) of an AWS EC2 instance with a total of 7 columns and around 8000 rows. Each row also has a timestamp with a 5 min interval between each row.

I want to build a LSTM model to forecast the performance for let’s say the next hour based on my data. What would be the best approach for solving a problem like this? I know this can be done in many different ways but I would really appreciate some input how to go about this.