Best Practices for Deploying Graph Databases in Edge Computing Environments?

General
1

Hi everyone,

I’m currently researching how to deploy graph databases in edge computing environments, where data processing happens close to devices generating the data. Since edge computing environments often have constraints like limited hardware resources, intermittent or unreliable network connectivity, and the need for low latency, I’m interested in learning what best practices exist for such deployments.

Specifically, I’m curious about:

  • Recommended architectures for running graph databases at the edge
  • Strategies for handling data synchronization between edge nodes and central servers
  • Performance tuning or resource optimization tips for graph databases on constrained hardware
  • Approaches to ensure fault tolerance and data consistency despite intermittent connectivity

If you have experience or examples related to deploying graph databases in edge setups, I’d love to hear what worked well or any lessons learned.

Thanks in advance for your insights!

2

IMHO - there is no single answer for all of your requirements as an architecture, it will all be a mixture of actual NFR + budget.

If you state you have "limited hardware resources" = memory/storage, that means that you can only keep a limited amount of graph data locally - if you need more data and need to satisfy "low latency", you can't have "unreliable network connectivity".

Basic NFR:

  • do you also need ACID?
  • what's your acceptable time for inconsistencies at the edge?
  • what's your acceptable data loss from the edge?

Then you come to the functional:

  • do you need a full graph at the edge? or just some subset(s)?
  • can it be decomposed?
  • what happens when you need data and the network isn't working?