Flow Latency // Time to Complete Coverage

    Latency on Flow is expected to be worse than other chains because, at minimum, 4 calls are required to generate a complete block. Blocks are requested by height, which returns an array of collections by ID. These collections are containers for transactions, thus the collection response (called by individual ID) contains an array of transaction IDs. Transactions are then requested by ID, twice, for the tx body and tx response. The response contains the event data and is often an order of magnitude slower than the other 3 gRPC calls.

    Loading...
    Loading...

    12/13/23: 4x/hr experiment. Not much change in total time to complete blocks, about an hour. Seeing a steeper curve 15 and 30 mins in. i.e. there are fewer missed blocks closer to chainhead, but it still takes about an hour for 100%.

    12/12/23: Increased the frequency of node calls by 50%, to 3x per hour per gRPC call and have seen significant gains. 0% missing blocks is hit around the 1 hour mark.

    12/8/23: The chart on the left, here, shows percent of missing blocks at 15 minute intervals after chainhead (as defined by the max block_timestamp at the time the query was last run.)

    Complete coverage of blocks (100%) is currently taking 2-2.5 hours.