If you intend to scale dynamically, I would recommend to start standalone
server nodes and run nodes embedded in the app as clients . This will
allow you to separate the load and scale two parts of this architecture
separately. For faster access on the client you can utilize near cache .
Also keep in mind that if you lose all nodes except one at a time, you will
most likely lose some data, unless you're using replicated cache. If there
is a big risk that such thing will happen and you want to avoid data loss in
this scenario, replicated cache in a must for you.
View this message in context:
Sent from the Apache Ignite Users mailing list archive at Nabble.com.