Download presentation
Presentation is loading. Please wait.
Published byBlaise Hart Modified over 9 years ago
1
Firewater The “split” personality of LucidDB A quick whirl through combining a purpose built BI database with scale “out” capabilities aka – LucidDB roadmap for “Bigger and Faster”
2
First and only 100% Open Source DB purpose built for doing Analytics Long Code History (10+ years) Foundation backing Commercial Support through DynamoBI
3
Column Store Vertical Partitioning (solves I/O problems) Great for BI, awful for OLTP Bitmap IDXes, Hash Joins/Aggs Labels (read consistent reporting) UDXs, plugin framework, extendable catalog
4
We're winning the “I/O war” Row Store I/O Bound CPU Cycles to Spare Column Store I/O efficient CPU bound THOU SHALT PARALLELIZE THY PROCESSING THY DATA SHALL BE PARTITIONED GETTING AHEAD OF THE I/O CURVE GIVES US A NEW WAR: CPU and Massively Parallel Processing
5
SERVER1 LUCIDDB SELECT SUM(SALES) FROM 100M_ROW_TBL GROUP BY YEAR, MONTH SERVER1 LUCIDDB SERVER2 LUCIDDB SERVER3 LUCIDDB SERVER0 FIREWATER SELECT SUM(SALES) FROM 100M_ROW_TBL GROUP BY YEAR, MONTH
6
1 SERVER, SMP, 8 CPUs Available First “run” thrown out to tickle the cache SELECT T."Carrier" as "CARRIER", C, C2, C*1000/C2 as "C3" FROM (SELECT "Carrier", count("Carrier") AS C FROM OTP."ontime" WHERE "DepDelay">10 GROUP BY "Carrier") T JOIN (SELECT "Carrier", count("Carrier") AS C2 FROM OTP."ontime" GROUP BY "Carrier") T2 ON (T."Carrier" = T2."Carrier") ORDER BY 3 DESC;
7
DEMO
8
SMP MPP
9
What's Working Projection Pushdown Filter Pushdown ( where j > 3) GROUP BY pushdown (all aggs) GROUP BY + distinct count Combinations of these things
10
What's Not (aka Roadmap) Hash and Replicated tables and * schema Joins Parallel Loading GOAL: Full support for Mondrian * queries MONDRIAN Page Views Cube LucidDB Firewater * queries
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.