Hi, 1. Yes, use simple types if it is possible. These types has much smaller footprint than POJOs. Ignite has performance optimization for simple types (and some other JDK classes like String). Ignite use BinaryObjects [1] concept that allows to query certain user-object's field without deserialization of whole object, but this costs some memory overhead.
2. See [2] for details. Inline size is a portion of key\column which will be inlined into index tree. So, when field value fit to inline size then Ignite will be able to comparison instantly when scaning\updateing index. Otherwise Ignite will go for data page (where entry actually resides) to read field whole value for further comparison. Proper inline size may significantly speed-up SQL index scans and updates operations. [1] https://apacheignite.readme.io/docs/binary-marshaller#basic-concepts [2] https://apacheignite-sql.readme.io/docs/create-index#section-index-inlining On Wed, Jun 6, 2018 at 2:03 PM, the_palakkaran <[email protected]> wrote: > Hi Andrew, > > Thanks for the reply. > > I have two more doubts: > > 1. So rather than a Custom key, you are suggesting to use simple types like > Integer, Long, etc, right? Why is this so? > > 2. What does indices with proper inline size mean ? Also, more indices does > not necessarily give better performance right? > > > > > > -- > Sent from: http://apache-ignite-users.70518.x6.nabble.com/ > -- Best regards, Andrey V. Mashenkov
