Serverless RPU estimator
RPU calculations can be difficult to understand and generally involves how much "processing" is being done by MongoDB to complete a particular read operation. There are instances where RPU's can skyrocket such as when COLLSCANs are performed due to lack of indexes. This can result in large spikes in the bill for serverless instances. It would be great to have some RPU estimator to calculate the approximate RPU's a particular operation would generate.
Jason
shared this idea
-
Tyler Termini
commented
Yes, 100%. I got an unexpected bill for over $1000 after converting an M5 cluster to a Serverless, because I have so little traffic that it didn't even warrant an M5, but I still needed >= 5GB disk space. I had a basic understanding of what RPUs were but I still could have benefited from an estimator based on current M5 activity—I might have a runaway process, or have a vast misunderstanding of how RPUs work.
In any case, though my app receives almost zero traffic (other than a couple devs using Compass), I was unable to anticipate a bill of over $1000.
So much for saving money on my $25/mo M5 cluster!
I still love MongoDB <3, but this would be so helpful!
-
Oleksiy
commented
RPU is actually about 50% of cluster costs, so that it would be great to have estimator for WPU, storage (incl. backup!), etc.
Thank you!
-
David Sanchez
commented
Also it would be interesting to be able to estimate how much the cost will be if we migrate to Serverless based on historical usage of an existing shared or dedicated instance