Event banner
Event details
How do you manage the cost of using Azure Search as vector store for RAG use cases? I had an experience trying out an azure search example that involves some 90k arvix and covid journals and it costs £500 just to load them. Would there is a way to estimate cost better and how should one approach it?
fsunavala-msft
Microsoft
MicrosoftCost ultimately comes down to your SKU and the number of Search Units (SUs) which many factors play a role in the decision making such as document count, availability and reliability, index schema design, count of vector fields, vector field dimensions, etc. The most accurate way to estimate cost is to perform a proof-of-concept index with the desired schema you want and load a sample of documents that will reflect your production workload. Then, you can extrapolate your sample document count and index size with your production document count and get your production index size. You can visit our documentation on service limits by SKU and decide which SKU and Search Unit Count (partitions/replicas) fit your needs best. https://learn.microsoft.com/azure/search/search-limits-quotas-capacity
Once you have an estimated SKU and number of search units, you can visit our pricing calculator here, select your region, and see a cost estimate of your Cognitive Search service: https://azure.microsoft.com/pricing/details/search/