Hi Tuco, Seems, there is a misunderstanding.
Ignite does not have master and slave nodes. All nodes are equals. There are primary and backup partitions distributed among grid nodes. SQL queries are running on index of primary partitions on all data nodes. Actually, index contains backup data, but it is filtering out. If we allow read from backup, then i see 2 issues: 1. how we should decide what part of data of same partition should query will read from primary and what from backup, 2. how we should merge results if we will get a duplicates? Expecially for queries with aggregates, "Select count(*) from table". On Tue, Apr 25, 2017 at 12:45 PM, tuco.ramirez <[email protected]> wrote: > Thanks Andrew, > > A follow up question, if i have 4 machines comprising a cluster of a > distributed cache with backups set as 1, will 2 machines be master and 2 be > slaves, or each machine will hold the 1/4th data of master data and 1/4th > of > slave data of another master. > > If it is former, then we reading from slaves will increase the throughput > as > the queries on master nodes will reduce. > > Thanks > Tuco > > > > -- > View this message in context: http://apache-ignite-users. > 70518.x6.nabble.com/Basic-question-on-scaling-of-reads- > with-ignite-sql-queries-tp11975p12229.html > Sent from the Apache Ignite Users mailing list archive at Nabble.com. > -- Best regards, Andrey V. Mashenkov
