Is fullness metric calculated based on vectors or vectors + metadata? Lets say S1 pod takes 5M vectors. For 50M vectors I would need 10 S1 pods. If per vector approximate 30% bytes are for metadata in my operation. In this case would I need 13 S1 pods instead of 10 for 50M vectors ?
The number of vectors per pod are dependent on a number of factors. The most significant factors are number of vectors, vector dimensions, and metadata size per vector. To estimate the number of S1 pods for 50M vectors, we also need vector dimensions and metadata size per vector. The pod usage estimator at Pricing | Pinecone accounts can help you. 5M vectors for S1 pod is based on vectors being 768 D and each vector having few hundred bytes of metadata.
Fullness metric does account for metadata. Please let me know if you have any question.