Oracle Compression Test

Overview

The following tests calculate the amount of disk space required to store 10+ million time:value samples in a Oracle EE 12c v12.2.0.1 database.

Results

Schema Compressed Data Size Index Size Total Size Row Count Bytes per Row Bytes per Sample
Trade Table No 92,274,688 45,088,768 137,363,456 2,045,514 67.2 13.4
Trade Table Yes 53,477,376 42,991,616 96,468,992 2,045,514 47.2 9.4
Universal Table No 276,824,064 260,046,848 536,870,912 10,227,570 52.5 52.5
Universal Table Yes 184,549,376 218,103,808 402,653,184 10,227,570 39.3 39.3

Dataset

The dataset represents 20+ years of historical minute stock trade data available from the Kibot company.

The one minute trade statistics are available for IBM stock traded on the New York Stock Exchange. The recording starts on February 1st, 1998 and lasts until the last trading day.

The data is provided in the commonly used OHLCV format.

Date,Time,Open,High,Low,Close,Volume
01/02/1998,09:30,104.5,104.5,104.5,104.5,67000
...
09/08/2017,17:38,142.45,142.45,142.45,142.45,3556

The records can be downloaded from http://api.kibot.com/?action=history&symbol=IBM&interval=1&unadjusted=0&bp=1&user=guest.

The file contains over 2 million lines. The OHLC metrics contain values with up to four decimal places. The volume metric is an integer. The dates are recorded in US/Eastern time.

Each row consists of 5 metrics for a given 1-minute interval:

09/08/2017,15:42,142.53,142.5399,142.49,142.49,10031
...
time   = 09/08/2017 15:42
open   = 142.53
high   = 142.5399
low    = 142.49
close  = 142.49
volume = 10031

Schema Alternatives

The tests are performed using two schema options:

  • Trade Table schema uses a named column for each input metric.
  • Universal Table schema uses a single metric ID column for all input metrics.

The Trade Table schema requires less disk space however the underlying table can not be extended to store different sets of columns for different instrument types. As such, multiple tables must to be created to store data for various instrument types.

The Universal Table schema allows adding new metrics without altering the tables. This can be done by inserting a new record into the Metrics table (a dictionary) and using foreign keys when inserting data into the data table.

Trade Table Schema

  • TradeHistory Table
DESCRIBE TradeHistory;
Name                     Null?    Type
----------------------------------- -------- ------------------------
INSTRUMENT                 NOT NULL NUMBER(7)
OPEN                          NUMBER(7,4)
HIGH                          NUMBER(7,4)
LOW                          NUMBER(7,4)
CLOSE                          NUMBER(7,4)
VOLUME                       NUMBER(8)
TIME                     NOT NULL TIMESTAMP(0)

SELECT * FROM TradeHistory FETCH FIRST 5 ROWS ONLY;

INSTRUMENT|     OPEN|        HIGH|    LOW|     CLOSE|    VOLUME|TIME
----------|----------|----------|----------|----------|----------|----------------------
     1|    104.5|       104.5|     104.5|     104.5|     67000|02-JAN-98 09.30.00 AM
     1|    104.38|       104.5|    104.38|    104.38|     10800|02-JAN-98 09.31.00 AM
     1|    104.44|       104.5|    104.38|     104.5|     13300|02-JAN-98 09.32.00 AM
     1|    104.44|       104.5|    104.38|    104.38|     16800|02-JAN-98 09.33.00 AM
     1|    104.38|       104.5|    104.38|    104.38|      4801|02-JAN-98 09.34.00 AM
  • Instruments Table
DESCRIBE Instruments;

 Name             Null?      Type
 ----------------------- -------- ----------------
 ID             NOT NULL NUMBER(7)
 NAME                  VARCHAR2(20)

SELECT * FROM Instruments;

    ID|NAME
----------|--------------------
     1|IBM

Universal Table Schema

  • UniversalHistory Table
DESCRIBE UniversalHistory;

Name                     Null?    Type
 ----------------------------------- -------- ------------------------
 INSTRUMENT                 NOT NULL NUMBER(7)
 METRIC                  NOT NULL NUMBER(7)
 TIME                     NOT NULL TIMESTAMP(0)
 VALUE                          NUMBER(12,4)

SELECT * FROM UniversalHistory FETCH FIRST 5 ROWS ONLY;

INSTRUMENT|    METRIC|TIME                                     |     VALUE
----------|----------|---------------------------------------------------------------------------|----------
     1|        1|02-JAN-98 09.30.00 AM                             |     104.5
     1|        1|02-JAN-98 09.31.00 AM                             |    104.38
     1|        1|02-JAN-98 09.32.00 AM                             |    104.44
     1|        1|02-JAN-98 09.33.00 AM                             |    104.44
     1|        1|02-JAN-98 09.34.00 AM                             |    104.38
  • Instruments Table
DESCRIBE Instruments;

Name                     Null?    Type
 ----------------------------------- -------- ------------------------
 ID                     NOT NULL NUMBER(7)
 NAME                          VARCHAR2(20)

SELECT * FROM Instruments;

    ID|NAME
----------|--------------------
     1|IBM
  • Metrics Table
DESCRIBE Metrics;

Name                     Null?    Type
 ----------------------------------- -------- ------------------------
 ID                     NOT NULL NUMBER(7)
 NAME                          VARCHAR2(20)

SELECT * FROM Metrics;

    ID|NAME
----------|--------------------
     1|Open
     2|High
     3|Low
     4|Close
     5|Volume

Executing Tests

Download Input Data

Create directory /tmp/test.

mkdir /tmp/test

Download the dataset.

curl --compressed -o /tmp/test/IBM_adjusted.txt \
  "http://api.kibot.com/?action=history&symbol=IBM&interval=1&unadjusted=0&bp=1&user=guest"

Verify the row count:

wc -l IBM_adjusted.txt
2045514 IBM_adjusted.txt

Launch Oracle Database Container

Build a Oracle EE 12c v12.2.0.1 container as described here. Start the container with /tmp/test directory.

docker run --name oracle \
  -e ORACLE_SID=axibase \
  -e ORACLE_PWD=axibase \
  -v /tmp/test:/data \
  -d oracle/database:12.2.0.1-ee

Execute SQL scripts for the Trade Table Schema

curl -o /tmp/test/oracle-trade-table.sql \
 "https://raw.githubusercontent.com/axibase/atsd/master/administration/compaction/oracle-trade-table.sql"
docker exec -u root oracle bash -c "chmod 777 /data" && \
 cat /tmp/test/oracle-trade-table.sql | \
 docker exec -i oracle sqlplus -S system/axibase | grep '|' --color=never
SEGMENT_NAME          |    BYTES
--------------------------|----------
TRADEHISTORY          |  92274688

SEGMENT_NAME          |    BYTES
--------------------------|----------
TRADEHISTORY_PK       |  45088768

TABLE_NAME  |ROWS_COUNT
------------|----------
TRADEHISTORY|    2045514

SEGMENT_NAME          |    BYTES
--------------------------|----------
TRADEHISTORY_COMPRESSED   |  53477376

SEGMENT_NAME          |    BYTES
--------------------------|----------
TRADEHISTORY_COMPRESSED_PK|  42991616

TABLE_NAME           |ROWS_COUNT
-----------------------|----------
TRADEHISTORY_COMPRESSED|   2045514

Execute SQL scripts for the Universal Table Schema

curl -o /tmp/test/oracle-universal-table.sql \
 "https://raw.githubusercontent.com/axibase/atsd/master/administration/compaction/oracle-universal-table.sql"
docker exec -u root oracle bash -c "chmod 777 /data" && \
 cat /tmp/test/oracle-universal-table.sql | \
 docker exec -i oracle sqlplus -S system/axibase | grep '|' --color=never
SEGMENT_NAME              |     BYTES
------------------------------|----------
UNIVERSALHISTORY          | 276824064

SEGMENT_NAME              |     BYTES
------------------------------|----------
UNIVERSALHISTORY_PK          | 260046848

TABLE_NAME    |ROWS_COUNT
----------------|----------
UNIVERSALHISTORY|  10227570

SEGMENT_NAME              |     BYTES
------------------------------|----------
UNIVERSALHISTORY_COMPRESSED   | 184549376

SEGMENT_NAME              |     BYTES
------------------------------|----------
UNIVERSALHISTORY_COMPRESSED_PK| 218103808

TABLE_NAME           |ROWS_COUNT
---------------------------|----------
UNIVERSALHISTORY_COMPRESSED|  10227570