I have a client with about 20 TB in an on-prem Aerospike instance who needs to migrate this to an AWS-managed instance. Is there a way to get some relatively precise notion of pricing? They’ll be doing around 15,000 operations per second on average (both reads and writes). I’ve been hoping to find a tool that could take that basic data and give back an estimated price (a number of other databases do this). A tool would allow us to play with different usage patterns to get an idea of the cost landscape. Does anyone have an idea where to find a tool or even to do the calculation?
The SA team has some handy formulas to punch things into. If you have enterprise they can hook you up. The answer to your question is either “use in-memory and just make sure you have enough network” (probably) or “use ACT to figure out your cap planning”. https://docs.aerospike.com/docs/operations/plan/ssd/ssd_certification.html i3en/*5d type instances usually score around ~30x per drive. If you translate your workload into ACT scores then you should be ale to figure out the IO needs, then the rest is simple capacity/storage calculations.
oh… AWS managed. Not really certain, but I am certain it definitely depends on object-size. 15K QPS on 8 byte objects is easy. 15K QPS on 8MiB objects is a different story
Thanks for the reply, Albot! I forgot to mention these are 6KB objects. I’m very interested, in particular, if AWS managed Aerospike is significantly cheaper than, for example, DynamoDB (I suspect it is).
I think you’re probably going to get better answers contacting enterprise sales through Aerospike or contact AWS directly for details. Forums are more DIY 
1 Like