Yes. You can start with an instance type and seamlessly switch without any downtime. Reach out to support@rockset.com to migrate.

Data retained and data ingested are measured as the total number of uncompressed bytes required to represent your documents in Rockset.

The size is measured after data is parsed and transformed, but before it is indexed. If the data source uses compression, size is measured after uncompressing the data. Dropped fields using field mappings do not count towards size. Rockset automatically builds 3 indexes to speed up queries, but space amplification due to indexing does not count towards size.

Data for a replicated collection counts once as data ingested and once for each replica. For example, adding 1GB of data into a collection that has 2 replicas counts as 1GB data ingested and 2GB data retained.

An individual document's size is approximately 1 byte for each field path key plus the size of each field path's value.

The units for data size follow base 1024, e.g. 1 GB = 1,024 MB = 1,048,576 KB.

Rockset simultaneously runs multiple copies of a collection for availability and increased throughput capacity. Each copy of the data, including the primary copy, is called a "replica." All collections have 2 replicas by default. Each replica counts separately towards the total GBs stored.

A millicpu is roughly one thousandth of a core. In Rockset, compute is a multi-tenant resource and is distributed across all users. The portion allocated to a particular collection depends on the collection's size and instance type, growing roughly linearly with collection size. For a user with dedicated instances, compute is isolated and can only be allocated to that user.