I’ve been investigating the performance of our “big vendor” RDBMS to hold months of sensor data. So far, the results are not really encouraging. I observe a exponential performance drop on the single-index (PK) table holding the data as more records are added. Here’s a plot of the performance of 5K records as records are continuously added to the table. Record addition is done with 5 parallel client threads, each inserting 1K records in batch mode. The client is an optimized Java app, using raw JDBC for the batch inserts. I haven’t found a faster way than that to add records to the relational DB.