How to batch modify large amount of entries without increasing replication latency

We are currently using the following:

OpenLDAP 2.3.43 standard syncprov overlay configuration below 1 master and 3 replicas each on 15k RAID 10.

Problem is batch modifications (perl ldapmod) consisting of modifying a large number of entries ( > 50k ) will cause major replication delays after a few hours. The entry updates tend to be small eg adding a new non-indexed attribute and we throttle down the update to only 3 entries per second.

What is the recommended way to batch modify large amounts of entries without taking the directory (master or replica) offline. Furthermore what is the expected replication throughput eg mods/sec that will sustain no replication delay.

Any and all advice greatly appreciated.

Thank you,

Mark

MASTER replication config ==

overlay syncprov syncprov-checkpoint 50 10 syncprov-sessionlog 100

REPLICA replication config ==

syncrepl rid=RID provider=MASTER_URI type=refreshAndPersist searchbase=SEARCH_BASE binddn=BINDDN credentials=BINDPASSWORD retry="30 +"