ClickHouse meetup
in Saint Petersburg
ClickHouse: Present and Future
Previously
— HTTP and executable sources;
— merge optimization, vertical merge;
— distributed query tracing;
— clickhouse-local;
— BETWEEN, || operators;
— UUID - text functions;
New in Query Language
— KILL QUERY;
— LIMIT BY;
— SELECT INTO OUTFILE;
Interfaces
— ability to get progress in HTTP headers;
— ability to skip errors in text formats;
— proper HTTP response codes;
Build
— «proper» build and packages;
— system.build_options table;
Dictionaries
— cached external dictionary performance;
— cached external dictionary instrumentation;
— HTTPS dictionaries;
Instrumentation
— information about index memory usage;
— information about uncompressed column sizes;
— metrics for cache RAM consumption;
— metrics about merges;
Optimizations
— DISTINCT optimization;
— gzip performance in HTTP interface;
— mark cache optimization;
Functions
— proper comparison logic, least, greatest;
— groupUniqArray for all data types;
— decodeURLComponent;
Something Else
— protection against accidental DROP TABLE;
— use_client_time_zone; timezone in config;
— fsync_metadata;
Community
— integration with Grafana, Redash, Apache Zeppelin, Superset;
— proper packages for CentOS, RHEL, GosLinux;
— native protocol driver for Go and C++;
— ability to pass X-ClickHouse-* headers;
— benchmarks NYC Taxi, Percona (Spark);
— Greenplum benchmark;
— English Telegram chat;
— meetings and talks (Brussels, Paris);
ClickHouse vs. Spark
https://www.percona.com/blog/2017/02/13/clickhouse-new-opensource-columnar-database/
ClickHouse vs. Greenplum
TODO (March-April 2017)
— distributed DDL queries;
— configs in ZooKeeper;
— full NULL support;
TODO (Spring-Summer 2017)
— ODBC driver functionality on Windows;
— rewrite query analysis: proper JOIN support;
