Distributed data types like IgniteAtomicLong or IgniteAtomicReference are used in the same cases as non distributed JDK counterparts: when you need atomically increment a long value or change a reference. The only conceptual difference is that the distributed data types, that are implemented in Ignite, can be used across different process that run on different physical machines.
IgniteCache in its turn is used as a primarily key-value storage in Ignite where all the data is located. Makes sense? — Denis > On May 6, 2016, at 11:08 AM, zshamrock <[email protected]> wrote: > > What is the difference between Atomic{Reference,Long} and IgniteCache - > memory and performance wise? Which one is more efficient, again comparing > the memory usage and the overall performance. > > As I can see caching could be done by utilizing the atomic instance per key, > or to use a normal cache. > What if the cache is just IgniteCache<String, String> or > IgniteCache<Integer, Integer>, would it be more beneficial to use > AtomicReference and AtomicLong per key correspondingly? > > What is the main design goal and use case for the Atomic{Reference,Long} > usage? Is it designed for something relatively short lived? How do you > control the lifecycle of your Atomic instance? > > Obviously, immediate advantage of using the cache, I see is the more > granular control over the cache behavior - expiry and eviction policy, near > cache, backups, to mention a few. So, when should I consider to use > AtomicReference or AtomicLong instead of IgniteCache? And when I benefit of > doing so. > > > > -- > View this message in context: > http://apache-ignite-users.70518.x6.nabble.com/AtomicReference-Long-vs-Cache-tp4812.html > Sent from the Apache Ignite Users mailing list archive at Nabble.com.
