Hi,
I checked provided test data. I was able to speedup a query execution
with Ignite about 2 times on my machine by using extra configuration
property System.setProperty("IGNITE_MAX_INDEX_PAYLOAD_SIZE", "256");
See a following documentation section about configuring index inline
size [1]. You can try the same in your environment. Shortly inline
size is needed to tune an indexed search speed. By default ignite
index pages can contain very limited pieces of indexed values (default
inline size is 10 bytes). If indexed values do not fit inline size
then actual values will be searched in another page (data page). It
can lead to a performance degradation.
> Not sure how to interpret the above statement. The support for SQL is an attractive feature of Ignite/Gridgain, but if it doesn't perform on a single node with little data I don't see how it will perform on a multi-node cluster.
Actually data distribution is a tradeoff. And usually it sounds as
"doing more work with more resources". And a gain here is not linear.
But as final result you can reach higher throughput by adding more
computational resources. Of course it depends on a particular
workload. Complex joins might be not good candidate here.
[1]
https://apacheignite-sql.readme.io/docs/performance-and-debugging#section-increasing-index-inline-sizeср, 18 сент. 2019 г. в 11:03, spoutnik_be <
[hidden email]>:
>
> Unfortunately, I am nowhere near Silicon Valley these days ;-)
>
> Any update on possible optimizations that could bring us somewhat closer
> than H2 timings?
>
> Thanks, L.
>
>
>
> --
> Sent from:
http://apache-ignite-users.70518.x6.nabble.com/--
Best regards,
Ivan Pavlukhin