I am planning to design a system for a small-to-medium-scale social medium platform and was unsure which cluster tier would best suit my use case.
We are expecting to have around 25K users with roughly 10-15 million reads (~100k writes) on a daily basis. The database must reside within Canada, so serverless instances are not an option for me. When exploring the dedicated cluster options, I was confused about the max number of connections.
For example, the M20 tier is stated to have 3K max connections. Is that referring to the connections formed for the length of a read/write request? Assuming there were 10 000 active users, would this max connections limit be met?
Additionally, I have experience with auto-scaled serverless services that charge according to the number of read/write operations. However, I am unfamiliar with whether an M20 tier cluster can handle such a quantity of reads.
I apologize for my lack of knowledge and would greatly appreciate any help!
Firstly, can you help me understand if the users here specify the number of people using the application and are not the Atlas users.
As mentioned in the Connection Limits and the Atlas Tier documentations, the limit is for the concurrent connections. The connections here represents the connection pooling and not the users using the database on the Atlas.
You may want to experiment with setting readPreference to something other than Primary (the default) to other settings such as PrimaryPreferred, or Secondary, to see if any of them suit your use case best.
Lastly, to plan a deployment of this magnitude, I’d suggest seeking Professional Advice . There are numerous considerations, and an experienced consultant can provide better guidance based on a more comprehensive knowledge of your needs. Some scalability decisions (such as shard key selection) become more difficult to course correct after a large quantity of production data has been collected.