In serverless environments such as AWS Lambda, the startup time of the Gremlin client becomes important. This is especially so when the serverless container is "cold". I have been getting fairly regular end user feedback on this so we started doing some analysis. There are many contributing factors but one thing that we could (and I think should) fix is the fact that the Gremlin Java client always creates a Gyro (default) serializer even if a different one such as GraphBinary is specified. To create the unused Gyro serializer adds almost a full second to the cold start time for the Gremlin client on a typical serverless container host.