Parallelizing search algorithms on Dgraph for an academic project

Misc
1

Hi,

My team and I have been looking to implement parallel graph search algorithms in dgraph, similar to: https://dl.acm.org/doi/10.1145/2693714.2693728

The goal is to implement various ways of parallelizing - goroutines, threads, processes and GPU (CUDA) - for the same task and compare them with different sizes of graphs.

We have been getting familiar with dgraph in the past few weeks. We have found Djisktra’s algorithm implemented in query/shortest.go.

We had a few queries regarding this:

  1. We are not yet sure how posting lists are structured, but is it feasible to run parallel operations on them without converting them to a different structure (such as an adjacency list)?
  2. Is there a documentation for code structure of the dgraph repo? There are lots of different folders and files at the root level. We were only able to find query/shortest.go because of a post here on discuss.
  3. Assuming we are able to implement this, would this be a useful addition to dgraph?
2

Postinglists are being updated. I’m also the main author of one of the more popular Go CUDA libararies. I can work with you on this. Would you like to jump on a call?

1 Like
3

This topic was automatically closed 41 days after the last reply. New replies are no longer allowed.