DataStore article should clarify more about "Throughput Limits"

As a Roblox developer, I would like to see more documentation on the somewhat recently implemented Throughput Limits. These limits are not currently described very well and while their calculation “trivia” (‘requests always round up to the nearest KB’) is mentioned, other facts are not mentioned such as the following:

  • Is the limit applied per server or per experience?
  • Is the limit refreshed after every UNIX minute, after every server-time minute, after 60 seconds, after the first request to the key or refreshed over-time?
  • Does RemoveAsync count towards the Read limit since it returns the value of the key?
  • Does UpdateAsync and IncrementAsync count against both limits?
  • If an UpdateAsync request has to be re-run (due to data changing mid-request), am I counted for double my quota?
  • Is metadata counted against the limit?
  • When using IncrementAsync, is the total bytes stored in the key charged to my Write quota or just the amount of bytes in the number I wish to increase by?
  • Do methods such as GetVersionAsync also count towards the limit?
  • Is there a flush of this limit on BindToClose? This is a bit of a stretch of a question, yet the “Server Limits” have allegedly (at least seemingly in the past) been given the undocumented ability to increase request allowance during a server shut down phase, this is critical for unfortunate scenarios in where a player may save their data using the last bit of the server limit, updates that data and then the server shuts down. Also, if this is still the case for server limits (and I haven’t just been gaslighted this whole time or it’s been removed) this behaviour should also be documented.

As a side note, I am very against the existence of this limit, it heavily restricts many innovative projects I wish to create and has led me to actually using multiple keys as backup data storage in-case a key is throttled on server shutdown, leading to very inefficient data structures. Yet this separate feedback is more suited for a separate topic altogether.


Coming back to this because I still can’t figure out how these limits are applied (per server or per experience), I’ve had hold off multiple projects indefinitely until this can be answered because if it is an experience-wide limit, my experience will break, really fast- (It relies on a datastore GetAsync request to get up to date configuration data)

Clarity on this topic would be greatly appreciated, if anyone in staff knows the answer. The answer to the first question would be especially appreciated: “Is the limit applied per server or per experience?”

Hey there,

I’ve actually since got in contact with staff elsewhere over this and the answer to this question is unfortunately “per experience”. None of the other questions were answered.

1 Like