Sizing eventstore servers


Our development team are starting to look at using eventstore for a number of use cases within our company and I have been trying to find some information about server sizing for eventstore servers similar to what you would find with SQL Server. Our predicated load would be around 25-50 events a second an annual growth of around 50GB a year so its not a big data system but we need to ensure the performance and throughput.

We use SSD drives configured in raid array on all our servers but I just needed an idea for how much RAM we should assign to each server in the cluster. How heavily does event store utilize an in memory cache and we will run into table scan type issues like in SQL Server if we do not have enough RAM assigned? Is there a ratio for RAM vs On Disk size that should be followed?

These machines will be virtualized so we can expand the RAM / Memory / Disk but I just need to ensure we can run for a while without having to resize the cluster every couple of months.

Thanks for any help.