In my application (mostly read oriented), i didn't set any partition key, so by default it will use the id as partition key. According to graph, we have a hot partition.
Storage for all the collections (collection that hitting Max RU is the second collection, 149.7 MiB one)
Why the use of the ID alone is insufficient as a partition key? i thought id is unique and it will scale evenly.
It's not a hot partition. A hot partition means that one or a few partitions use a high amount of RU compared to other partitions. In your case you have so little data that all your logical partitions reside into a single physical partitions. Thus it's not a hot partition; It's just a single partition that makes most of its RU throughput.
Having the id
as partition key is also no issue. Preferably you have a property that can act as partition key that you often query over. If you have no property that's suitable using the id
as partition key is no issue at all. By definition the id
partition key should be unlikely to cause hot partitions as your items are 'randomly' split between physical partitions once your collections get big enough to cause the splits.