There is no single answer for this as it really depends on your requirements.
You have to answer yourself the following questions:
* how many repositories do I expect to index? - this will affect disk space and disk IO required
* how quickly new repositories will be added? - this will affect number of initial indexing threads and amount of RAM
* how frequently I need to check for new updates in repositories, how many commits are being pushed between checks, how large are they? - this will affect number of incremental indexing threads, amount of RAM and CPU speed
* how many users will be stored in a directory? how many active (licensed) users will be? how many concurrent users I expect? - this will affect CPU speed, database connection pool size, RAM requirements
I suggest to experiment, e.g. spin up a test instance and fill with some test data (real or synthetic).
The general advice I can give is:
* if you do not expect heavy load on the instance, the m5.xlarge with SSD disk might be sufficient and will not cost you too much
* if you expect heavy load and seek for maximum performance, consider c5.xlarge or bigger with high IOOPS SSD disk (e.g. io2)
For example, a few years ago we tested indexing of a Linux kernel code repository. It took about 2 weeks to complete on M4, while it finished in just 3 days on C4.
Having said this, I strongly encourage to run your own tests, as this will strongly depend on usage patterns.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.