The benefit falls off fairly rapidly as the batch size increases. I did some crude benchmarking here: https://lucidworks.com/2015/10/05/really-batch-updates-solr-2/
If I were going to pick a number, I’d say 100 docs _per shard_. So if you have 10 shards, batch 1,000 docs if possible. Note, the efficiencies here are all about establishing a connection to Solr etc. rather than the actual work Solr does. Best, Erick > On Apr 25, 2019, at 3:13 PM, Markus Jelsma <markus.jel...@openindex.io> wrote: > > Hello, > > There is no definitive rule for this, it depends on your situation such as > size of documents, resource constraints and possible heavy analysis chain. > And in case of (re)indexing a large amount, your autocommit time/limit is > probably more important. > > In our case, some collections are fine with 5000+ batch sizes, but others are > happy with just a hundred. One has small documents and no text analysis, the > other quite the opposite. > > Finding a sweet spot is trial and error. > > Cheers, > Markus > > > > -----Original message----- >> From:Lucky Sharma <goku0...@gmail.com> >> Sent: Thursday 25th April 2019 21:48 >> To: solr-user@lucene.apache.org >> Subject: Solr-Batch Update >> >> Hi all, >> While creating an update request to solr, Its recommended creating >> batch request instead of small updates. What is the optimum batch >> size? Is there any number or any computation which can help us to >> assist on the same. >> >> >> -- >> Warm Regards, >> >> Lucky Sharma >> Contact No :+91 9821559918 >>
