1. It is not supported for now, but we plan do fix it [1]
2. Not yet. We split index different way. Every tree can manage certain
partition numbers. Feel the difference.
Mapping "partition -> tree segment" is static and we use as reminder of the
division, like partition_id%N.
Query works same way
trying to understand this:
1) In case where no indexes are involved and you are doing a table scan, it
should automatically try to exploit available CPU cores and process each
partition on a separate thread/core. At least table scan queries should
entertain the idea dynamic parallelism through DML
It can't work on the query level, because internally, it divides all
indexes into N trees instead of one(where N equals queryParallelism). You
can't redefine it after it was created since it will lead to the complete
rebuild of all indexes.
Evgenii
2018-06-06 20:55 GMT+03:00 Sanjeev :
> So it lo
So it looks like that Query parallelism works at Cache level, but it would
make more sense to do it at the Query level, more like a hint in a SQL query
to control how much parallelism is needed. This way it will be very dynamic
and users would have full control. Default could be 1, but OLAP queries
One more thing - it's not necessary for Ignite to send the query for all
server nodes, if it's possible to understand from the query, on which node
all data is placed(i.e. it filters by affinity key), it will send this
query only to the needed node.
Evgenii
2018-05-21 11:25 GMT+03:00 Evgenii Zhur
Hey,
It's configurable by using property queryParallelism and by default, an SQL
query is executed in a single thread on each participating Ignite node.
Please refer this documentation for additional information:
https://apacheignite-sql.readme.io/docs/performance-and-debugging#section-query-para
Hi,
I would like to understand how SQL queries are executed on Ignite Server
Nodes. Each Ignite Server Node has some number of primary partitions it is
responsible for. When a query is sent, let' say through JDBC interface, this
query is routed to all the server nodes where data resides. So the qu
