We have sensors that take measures with different frequencies. This frequencies can go from 2 measures per second(172000/day) to 1 measure per hour(24/day).
Our tipical query is: Give me the measures for sensor X from timeA to timeB.
Therefore one approach would be to use the sensor as the partition key and the timestamp as the clustering key. This would lead to exeding the recomended size for partitions for the sensors with high frequency measures.
One solution would be to add the date to our partition key. However this leads to lots of small partitions with just 24 rows (for the sensors with low measurement frequency). We have enough of both types of sensors so the data distribution in the cluster should not be a problem, however, we are concerned about having so many small partitions and about having both small and big partitions coexisting. Could this lead to a problem?
Is there any better way to deal with this issue? Thank you in advance.
与恶龙缠斗过久,自身亦成为恶龙；凝视深渊过久,深渊将回以凝视…
