Faster indexing performance

pixolution flow downloads external image files in order to index the analyzed data. This takes some time and is slower than analyzing plain text. Downloading the image and calculating the image descriptor is time consuming.

To gain indexing performance you may send several add requests in parallel to Solr. This way more images are processed at a time and the indexing should be faster.

We experienced that there is a limitation of concurrent connections when using the OpenJDK Runtime Environment. If you are using the OpenJDK, adding several documents in parallel requests and experiencing poor performance we recommend to install the Oracle JDK.

In short:

  1. Add documents in parallel requests
  2. Install Oracle JDK
  3. Calculate image descriptors before indexing outside of Solr (see Standalone Image Descriptor Generation)
  4. Make sure pixolution flow is using the performant math backend (see Native math backends for faster indexing)