Hi, Keys are distributed among partitions according to affinity function. Partitions are distributed among nodes according to affinity function. This chain "key->partition->node" is designed to simplify data exchange between nodes on topology changing and due to performance reasons.
RendezvousAffinityFunction is used by default. It tries to distribute keys among partitions in equal manner using key hashcode. Partition distribution among nodes is quite complex as here we have an optimizations to reduce number of partition transfers on unstable topology and save distribution as closer to equal as possible. AffinityKey is intended for data collocation purposes. Actually, AffinityKey is passed in affinity functions for affinity calculations if it is configured. If no affinity key is configured for key type, then entry key itself will be passed to affinity function where it will be used for affinity calculations. On Wed, Jul 19, 2017 at 5:28 PM, Ajay <[email protected]> wrote: > Hi, > > I want to know scenarios, > 1)I have 10 affinity keys like 1-10 suppose when i load this keys between > two servers(2 nodes) it will equally distribute or randomly distribute? > > 2)I have 10 normal keys like 1-10 suppose when i load this keys between two > servers(2 nodes) it will equally distribute or randomly distribute? > > Thanks, > Ajay. > > > > -- > View this message in context: http://apache-ignite-users. > 70518.x6.nabble.com/Ignite2-0-Affinity-Key-Distribution-tp15123.html > Sent from the Apache Ignite Users mailing list archive at Nabble.com. >
