HengWang / Alenka

GPU database engine

Welcome to Alenka - GPU database engine

What is this?

This is a GPU based database engine written to use vector based processing and high bandwidth of modern GPUs

Requirements

• CUDA (nvcc) + Nvidia GPU
• bison
• flex
• Modern GPU Library (included as submodule)

How to build?

git clone --recursive https://github.com/antonmks/Alenka.git
cd Alenka
make

Features :

• Vector-based processing
  CUDA programming model allows a single operation to be applied to an entire set of data at once.

• Smart compression
  Ultra fast compression and decompression on GPU. Database operations on compressed data.

• Column-based storage
  Minimizes disk I/O by only accessing the relevant data.

• Data skipping
  Better performance without indexes.

• Fast Loading
  Gpu based CSV parser loads the data into database at very high speed.

How to use it ?

Create your data files :

Run scripts load_orders.sql, load_lineitem.sql and load_customer.sql to create your database files.

Run your queries from a command prompt or use Alenka JDBC driver from Technica Corporation

Step 1 - Filter data

OFI := FILTER orders BY o_orderdate < 19950315;

CF := FILTER customers BY c_mktsegment == "BUILDING";

LF := FILTER lineitem BY shipdate > 19950315;

Step 2 - Join data

OLC := SELECT o_orderkey AS o_orderkey, o_orderdate AS o_orderdate, o_shippriority AS o_shippriority, price AS price, discount AS discount FROM LF JOIN OFI ON orderkey = o_orderkey JOIN CF ON o_custkey = c_custkey;

Step 3 - Group data

F := SELECT o_orderkey AS o_orderkey1, o_orderdate AS orderdate1, o_shippriority AS priority, SUM(price*(1-discount)) AS sum_revenue, COUNT(o_orderkey) AS cnt
FROM OLC GROUP BY o_orderkey, o_orderdate, o_shippriority;

Step 4 - Order data

RES := ORDER F BY sum_revenue DESC, orderdate1 ASC;

Step 5 - Save the results

STORE RES INTO 'results.txt' USING ('|') LIMIT 10;