Cloud Datastore performance for highly parallel single-key access pattern

[Rune K. Svendsen]

I've been thinking a lot about what kind of storage to use for my application, and I've ended up at the Cloud Datastore, but I'm wondering whether it is suitable, since I don't see my use case covered in the overview.

In my app, each incoming request begins a transaction on a single, distinct Entity in the store, which is expected to contain ~millions of Entities. The transaction is committed before the request response is delivered. So, during high load, there could be thousands of concurrent, pending transactions, all acting only on their own Entity.

In other words, each incoming request references a single Entity only, which it "locks" (by starting a transaction), and it is then "unlocked" (committed) before the response is delivered. Essentially it's just: 

   1) fetch Entity + mark as busy

   2) check request data against fetched Entity

   3) commit transaction with updated Entity on good request data / abort transaction on bad request data

Will I be able to get ~1k transactions per seconds with this usage pattern for Cloud Datastore?

[Faizan (Google Cloud Support)]

Hello Rune,

Your question is related to Cloud Datastore, to get assistance I would recommend posting it on Google Cloud Datastore discussion group[1]. This group is meant for discussions related to Google Compute Engine. As such, this is not the best place to get answer to your question.

I hope that helps.

Faizan

[1] https://groups.google.com/forum/#!forum/gcd-discuss

[Rune K. Svendsen]

Thank you, Faizan. I will repost there.


/Rune