1 Reply Latest reply on Apr 29, 2014 12:54 PM by rvansa

    Infinispan as a primary data store for 1m items

    tromper11

      Hi,

       

      I need to store approx 1million key-value pairs, where the key/value size is about 200bytes/1kb;

       

      I plan to use Infinispan in distributed mode with a (sync.) Oracle/MySQL cache loader; but hopefully all the data shall be in memory and database is for backup only.

       

      The system puts/gets data from Infinispan and Infinispan is trusted to persist/load entries from database.

       

      I did some spiking, but before signing on a full solution, I'd like to know if there's any issue with this approach or anything else that needs to be considered?

       

      Will it work with standard gbit Ethernet, if data needs to be rebalanced ?

       

      Thanks,

        • 1. Re: Infinispan as a primary data store for 1m items
          rvansa

          Read will have good performance (as long as these are loaded into memory, you can specify preload="true" on the cache store), but with sync (write-through) store you'll always wait for the entry to be persisted when you write. Async store would automatically buffer the writes. For durability there are the backup owners in the cluster (in case you don't expect whole cluster to crash due to power drop).

          But of course, your mileage can vary and if you can live with slow writes, your suggested setup is safe.

          1 of 1 people found this helpful