I think you're worrying about the wrong problem ;) Often, the difference between the JSON and SolrInputDocument decoding on the server is dwarfed by the time it takes the client to assemble the docs to send. Quick test: When you start indexing, how hard is the Solr server working (measure crudely by looking at CPU utilization). Very frequently you'll find the server sitting around waiting for the client to send documents.
Very often you'll get _much_ greater throughput gains by racking N clients together all sending to the Solr server than you will get by worrying about whether JSON or SolrInputDocument (or even XML docs) is more efficient on the server. That said, SolrInputDocuments are somewhat faster I think. FWIW Erick On Mon, Aug 11, 2014 at 7:34 AM, <georgelav...@comcast.net> wrote: > I have a large number of documents that I am trying to load into SOLR. > > I am about to begin bench marking this effort, but I thought I would ask > here. I have the documents in JSONArrays already. > > I am most concerned with ingest rate on the server. So I don't mind > performing extra work on the client to speed up the server... > > Assuming I am using ConcurrentUpdateSolrServer, will I get better ingest > performance if I convert all my documents to SolrInputDocuments before > sending, > or if I use the JsonRequestHandler on the server and send the JSONArrays > via a ContentStreamUpdateRequest? > > Thanks, > George >
