交易中的实体组（竞争）

Question

I'm reading a book on GAE. In a chapter about transactions, it says:

Updating an entity in a group can potentially cancel updates to any other entity in the group by another process. You should design your data model so that entity groups do not need to be updated by many users simultaneously.

Be especially careful if the number of simultaneous updates to a single group grows as your application gets more users. In this case, you usually want to spread the load across multiple entity groups, and increase the number of entity groups automatically as the user base grows. Scalable division of a data resource like this is known as sharding.

An often used example for an entity group is the message board, where the board is the ancestor of messages belonging to that board.

However, If updating a message (i.e. editing it) causes contention, and more often so as the userbase grows, isn't it bad design to create a huge group of messages with the board as its ancestor? The write rate of an entity group is limited to 1 per second. Does that mean any message within the board can be updated at most once per second?

Also, does merely adding an entity to a group (i.e. posting a new message) also count as "updating" and cause contention?

Answer 1

Yes, such design may be considered a bad one as it doesn't scale well with the number of users. I can't see a good reason for which messages would need the board as ancestor.

Yes, creating a new entity in a group counts as an entity group update and all updates can contribute to contention.

Side note: you might find this clarification useful: https://stackoverflow.com/a/39309022/4495081 (but for designs which have good reasons for using multi-entity groups).