ignite-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrey Mashenkov <andrey.mashen...@gmail.com>
Subject Re: About SQL queries and serialization
Date Fri, 09 Jun 2017 10:38:11 GMT

Please properly subscribe to the mailing list so that the community can
receive email notifications for your messages. To subscribe, send empty
email to [hidden email]
<http://user/SendEmail.jtp?type=node&node=12603&i=0> and
follow simple instructions in
the reply.

1. SQL works only with data that was loaded to cache. So, read-through is
not supported with SQL queries as data is not indexed until it passed to
cache via put or loadcache.

2. Ignite fully relies on H2 planner, but from 2.0 you can use Hints to
force certain index usage.

3. What IgniteSpiAdapter do you mean? It is just abstract class to make
building of pluggable components easier.
Do you mean IndexingSpi which is separate feature for custom data indexing
and required for SpiQuery supported? This have nothing common with SQL

4. All data (keys and values) is serialized with BinaryMarshaller [1] into
BinaryObjects. We avoid Java Serialization by default because it doesn't
meet our needs and has performance issues.
AFAIK from ignite-2.0 there is only BinaryMarshaller can be used.

5. Seems, the only object size limitation is size of byte[] array. However,
transferring big object over network can cause performance issues. E.g.
when node left topology a new backup should be created, that cause a
partitions transfer.

[1] https://apacheignite.readme.io/docs/binary-marshaller

Best regards,
Andrey V. Mashenkov

View raw message