breezykermo / oak

1 stars 0 forks source link

Re-run ACORN experiments in OAK on CloudLab #29

Open csirianni opened 3 weeks ago

csirianni commented 3 weeks ago

This might require some additional infra setup. Run the Docker image on CloudLab and load the datasets into the load.

Run one experiment per dataset and produce a latency/throughput graph per #28. The general idea is explained here, but Akshay has some research code that can help.

Figure out how to set recall at a fixed rate. Table 3 in the ACORN paper provides parameters for 80% recall.

We do not plan on creating a 3D graph. Instead, we will just report on some specified recall rate, though we may need to just state the HNSW parameters.