Uploaded image for project: 'BrownDog'
  1. BrownDog
  2. BD-683

Scale up algorithm in Elasticity module instantiates a lot of extractors in the subsequent runs

XMLWordPrintableJSON

      When I ran the upload script to submit 1200 images, to total upload time was around 3-5 mins,, it created around 10 VMs for opencv with number of extractor instances 14 for eyes and faces, 20 for profiles.

      After the VMs are suspended except for one, I ran the experiment again, in this run, upload time was 20 mins, the number of opencv VM was 10 but the number of faces and profile extractor instances were 40.
      One possible explanation for long upload time was clowder was busy in the second run. In the second run, the suspended VMs resumed faster than the creation of VMs and more number of extractors instances are deployed faster leading to many requests to clowder for file being uploaded keeping clowder busy.
      So it needs further investigation as if it is the desired behaviour.

              Unassigned Unassigned
              spadhy Smruti Padhy
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

                Created:
                Updated:
                Resolved: