toggle quoted messageShow quoted text
Yes its' an interesting and topic but I can't find any information about this except the IBM Compose.
在 2020年2月29日星期六 UTC+8上午5:17:27，Ryan Stauffer写道：
Just a quick thought experiment to put out to the fine folks in this group...
Would you use a "JanusGraph-as-a-Service" offering that works as follows:
- You're given WebSocket and HTTP endpoints, which you can use to issue Gremlin traversals from any host language or REST client of your choice.
- You can easily retrieve summary counts of the data you've stored by label and/or property
- The underlying infrastructure and storage & search backends are transparent, so you don't manage RAM, disk, compute, storage backups, etc
- You get SLAs regarding latency and throughput based on common workloads.
The idea is to minimize the hurdles to get started building and USING property graphs at scale. Right now, it's tricky up and down the stack (node sizing, cluster sizing, pluggable backend selection, etc). I'm just curious how everyone would feel if you DIDN'T have to think about these questions at all, and simply paid a price based on API calls and the total amount of data you're storing.
Thanks in advance for your thoughts!