For initial loads, you should pre-split. Otherwise initially all records will go into one shard then have to be balanced into others. Pre-splitting will allow the system to distrubute the data among the shards from the beginning.
This shouldn't be an issue in ongoing operations. It is slightly better in 2.2 than earlier but pre-splitting is still recommended for the initial load - the balancing algorithms aren't designed to work with near-empty systems.
-- Max
Hi,I work on simple test that insert one million documents in a MongoDB collection.When the collection is not sharded, the test last 60 seconds.When the collection is sharded, the test last 140 seconds.For information purpose :The shard key is a random value : doc.__sd.shard = Math.round(Math.random() * 1000)I have 4 servers splited in 2 replicaset (one primary, one secondary and one arbiter)The test load balance the data to be inserted on the 4 mongos.Thank you for your helpRomain--
You received this message because you are subscribed to the Google
Groups "mongodb-user" group.
To post to this group, send email to mongod...@googlegroups.com
To unsubscribe from this group, send email to
mongodb-user...@googlegroups.com
See also the IRC channel -- freenode.net#mongodb
Ok I will try it, I'll keep you informed about the performance issue.