HLCID is an OTP application which runs on each of your nodes. You can ask it for an ID like so:

id = HLCID.generate()
url_safe_id = HLCID.to_base64()
=> "AWy00jCrAABKaDWL0Z6heg=="

Each ID is a slightly modified hybrid logical clock. Each timestamp is composed of 48 bits of physical time, 16 bits of logical time, and 64 bits of random bytes. When encoded as binaries each id maintains lexical ordering.

HLCs automatically account for clock drift by utilizing both physical and logical time. But if the physical clock drifts beyond a given range

• 300 seconds by default - you won't be able to generate IDs. If you run HLCID in production, you'll want to monitor NTP for clock drift. Because HLCs use logical time to avoid conflicts and provide one-way causality tracking, there is a limited number of ids available in a millisecond window. This limitation isn't a problem for our use case, but if you need to generate more then 65000 ids per node per millisecond, than HLCID won't work for you.