The new indexing strategy uses an inverted index to associate every tag in a timeseries with the identifiers of timeseries that contain the tag. This allows for efficient query execution, as queries can be executed by making a single key-value lookup for each queried tag and retrieving a set of relevant timeseries identifiers. The previous strategy required scanning the data for the entire metric when only a small subset was requested, which could lead to full table scans and poor user experiences. The new strategy also introduces write and space amplification, as every unique timeseries identifier has to be stored multiple times, but this is not expected to become a problem due to disk space utilization being minimal. Intranode sharding allows for parallel execution of queries by fetching data from each RocksDB shard in parallel and then merging the results. The service was rewritten from Go to Rust, which resulted in significant performance improvements, with CPU-demanding operations being up to 6x faster. This change allowed Datadog to query higher cardinality metrics on the same hardware, reducing query timeouts by 99% and making the timeseries index nearly 50% cheaper to run.