Reputation: 328
I am doing performance test on Ignite Cache , I have created distributed cache with 10M entries in a 3 node cluster, for this cache I have created a near cache with max capacity 10M(entire cache with no eviction) in my client , when I compare the performance, I see significant difference in direct key based lookup ,below are the numbers with direct key lookup,
Number of lookups--100000 Distributed Cache--79447ms Near Cache--29521ms
However when I fire a query on the cache (with only indexed fields in the query) , I don't get any performance advantage with near cache, does this mean that near cache does not replicate the indexes locally?, Below are the numbers for query based lookup
Number of lookups--100000 Distributed Cache--97516ms Near Cache--99583ms
Upvotes: 2
Views: 487
Reputation: 8390
Near cache makes sense only for key-value access where you can fetch individual entry to the client node and evict rarely accessed entries when needed.
When using SQL queries, Ignite doesn't know the set of required keys in advance, therefore it has to execute the distributed query anyway. Near cache is not involved in this case.
Upvotes: 4