According to the Google Cloud Datastore documentation, fast writing to a new entity kind is bad practice. It says:
Cloud Datastore prepends the namespace and the kind of the root entity group to the Bigtable row key. You can hit a hotspot if you start to write to a new namespace or kind without gradually ramping up traffic.
What about writing an initial set of records at fast rate to an event kind in a namespace, but neither the kind nor the namespace is new? The number of records will be between 5 to 10 millions.
For example, assume I have a namespace "ns1", and an entity kind "ek1". (Mentally, it is the "ns1.ek1" entity to me.) If I already have many other entity kinds populated in that namespace ("ns1.ek2", "ns1.ek3", ..., "ns1.ekX"), and I already have entities of this kind in other namespaces ("ns2.ek1", "ns3,ek1", ... "nsX.ek1"), will I still run into performance problems due to hotspot updates if I write fast into "ns1.ek1"?
I plan to prime a brand new entity kind for fast writes into the customer specific namespaces by inserting a ton of artificial records with random keys into it in a private namespace. I am wondering if that is a valid technique.
