Choosing vCPUs

One vCPU represents one thread of a CPU core. Search operations are compute intensive, so the number of vCPUs heavily influences the amount of concurrent searches and import/export operations that can be done on the cluster. In general, higher traffic requirements translate to more vCPUs.

So what is the right number?

The answer to this question is highly dependent on your search query patterns and (re)indexing frequencies. We recommend that you pick the lowest vCPU option available for the RAM option you pick, benchmark against your own data and search query traffic patterns and then scale up to more vCPUs as needed. 

Here are some reference benchmarks to give you an idea: https://typesense.org/docs/overview/benchmarks.html.

Burst vCPUs

If you have relatively low baseline traffic for most of the day with the occasional moderate spike during indexing, you could save costs by choosing one of our "Burst" vCPU options. These instances allow you to use 100% of CPU for the indicated amount of burst time in a 24-hour period, and once that burst time is exhausted, CPU will be throttled to around 15% of capacity. After 24 hours, the threshold gets reset. 

Please note that this option is not a good fit if you have a high baseline of traffic or high indexing volume. You will start seeing slowdowns in response times if you direct a consistently large amount of traffic at nodes in this burst offering.