Memory to Storage Ratio
Tue, 09/20/2011 - 07:33
In one of the webinars Paul mentioned there was a calculator on the Couchbase site to help one determine the type of node to use in your cluster. I can't seem to find it.
If each node in my cluster (dedicated Couchbase boxes) had about 1-2TB of storage, what would be a good rule of thumb for how much memory each node should have? The cluster is used almost exclusively for writing docs (once) and fetching results of reduce queries (often). Individual docs are almost never pulled.
You're probably looking for this page:http://www.couchbase.org/wiki/display/membase/Sizing+Guidelines
The amount of RAM needed is less dependant on the total amount of data you have, and more related to how much of your dataset you need very low latency access to. We will totally support the pulling of data from disk, but it's a different performance level than you get from RAM.
In your particular use case, it sounds like you'll really just need enough RAM to hold the metadata for the items and cache the results as they come in from your writes. All of the indexes will continue to be served from disk, so RAM really doesn't come into play there (other than the OS caching disk access...which can be quite useful as well).
Hope that helps answer your question, let me know if you need anything else.
Perry
Forum support is great for free but sometimes you need a guaranteed response time and dedicated resources for your questions or issues.
Consider purchasing enterprise-level support from Couchbase: http://www.couchbase.com/products-and-services/overview
Call or email "sales -at- couchbase-dot- com" today!