You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@carbondata.apache.org by ch...@apache.org on 2017/02/04 02:38:12 UTC
[13/35] incubator-carbondata-site git commit: Updated website for
CarbonData release 1.0.0
http://git-wip-us.apache.org/repos/asf/incubator-carbondata-site/blob/0d4cdb1c/content/docs/latest/useful-tips-on-carbondata.html
----------------------------------------------------------------------
diff --git a/content/docs/latest/useful-tips-on-carbondata.html b/content/docs/latest/useful-tips-on-carbondata.html
new file mode 100644
index 0000000..f4f9cc0
--- /dev/null
+++ b/content/docs/latest/useful-tips-on-carbondata.html
@@ -0,0 +1,209 @@
+<!--
+ Licensed to the Apache Software Foundation (ASF) under one
+ or more contributor license agreements. See the NOTICE file
+ distributed with this work for additional information
+ regarding copyright ownership. The ASF licenses this file
+ to you under the Apache License, Version 2.0 (the
+ "License"); you may not use this file except in compliance
+ with the License. You may obtain a copy of the License at
+
+ http://www.apache.org/licenses/LICENSE-2.0
+
+ Unless required by applicable law or agreed to in writing,
+ software distributed under the License is distributed on an
+ "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ KIND, either express or implied. See the License for the
+ specific language governing permissions and limitations
+ under the License.
+--><h1>Useful Tips</h1><p>This tutorial guides you to create CarbonData Tables and optimize performance. The following sections will elaborate on the above topics :</p>
+<ul>
+ <li><a href="#carbondata-table">Suggestions to create CarbonData Table</a></li>
+ <li><a href="#carbondata-performance">Configurations For Optimizing CarbonData Performance</a></li>
+</ul><h2 id="carbondata-table">Suggestions to Create CarbonData Table</h2><p>Recently CarbonData was used to analyze performance of Telecommunication field. The results of the analysis for table creation with dimensions ranging from 10 thousand to 10 billion rows and 100 to 300 columns have been summarized below. </p><p>The following table describes some of the columns from the table used.</p><p><strong>Table Column Description</strong></p>
+<table class="table table-striped table-bordered">
+ <thead>
+ <tr>
+ <th>Column Name </th>
+ <th>Data Type </th>
+ <th>Cardinality </th>
+ <th>Attribution </th>
+ </tr>
+ </thead>
+ <tbody>
+ <tr>
+ <td>msisdn </td>
+ <td>String </td>
+ <td>30 million </td>
+ <td>Dimension </td>
+ </tr>
+ <tr>
+ <td>BEGIN_TIME </td>
+ <td>BigInt </td>
+ <td>10 Thousand </td>
+ <td>Dimension </td>
+ </tr>
+ <tr>
+ <td>HOST </td>
+ <td>String </td>
+ <td>1 million </td>
+ <td>Dimension </td>
+ </tr>
+ <tr>
+ <td>Dime_1 </td>
+ <td>String </td>
+ <td>1 Thousand </td>
+ <td>Dimension </td>
+ </tr>
+ <tr>
+ <td>counter_1 </td>
+ <td>Numeric(20,0) </td>
+ <td>NA </td>
+ <td>Measure </td>
+ </tr>
+ <tr>
+ <td>... </td>
+ <td>... </td>
+ <td>NA </td>
+ <td>Measure </td>
+ </tr>
+ <tr>
+ <td>counter_100 </td>
+ <td>Numeric(20,0) </td>
+ <td>NA </td>
+ <td>Measure </td>
+ </tr>
+ </tbody>
+</table><p>CarbonData has more than 50 test cases, on the basis of these we have following suggestions to enhance the query performance :</p>
+<ul>
+ <li><strong>Put the frequently-used column filter in the beginning</strong></li>
+</ul><p>For example, MSISDN filter is used in most of the query then we must put the MSISDN in the first column. The create table command can be modified as suggested below :</p><p><code>
+ create table carbondata_table(
+ msisdn String,
+ ...
+ )STORED BY 'org.apache.carbondata.format'
+ TBLPROPERTIES ( 'DICTIONARY_EXCLUDE'='MSISDN,..',
+ 'DICTIONARY_INCLUDE'='...');
+</code></p><p>Now the query with MSISDN in the filter will be more efficient.</p>
+<ul>
+ <li><strong>Put the frequently-used columns in the order of low to high cardinality</strong></li>
+</ul><p>If the table in the specified query has multiple columns which are frequently used to filter the results, it is suggested to put the columns in the order of cardinality low to high. This ordering of frequently used columns improves the compression ratio and enhances the performance of queries with filter on these columns.</p><p>For example if MSISDN, HOST and Dime_1 are frequently-used columns, then the column order of table is suggested as Dime_1>HOST>MSISDN as Dime_1 has the lowest cardinality. The create table command can be modified as suggested below :</p><p><code>
+ create table carbondata_table(
+ Dime_1 String,
+ HOST String,
+ MSISDN String,
+ ...
+ )STORED BY 'org.apache.carbondata.format'
+ TBLPROPERTIES ( 'DICTIONARY_EXCLUDE'='MSISDN,HOST..',
+ 'DICTIONARY_INCLUDE'='Dime_1..');
+</code></p>
+<ul>
+ <li><strong>Put the Dimension type columns in order of low to high cardinality</strong></li>
+</ul><p>If the columns used to filter are not frequently used, then it is suggested to order all the columns of dimension type in order of low to high cardinality. The create table command can be modified as below :</p><p><code>
+ create table carbondata_table(
+ Dime_1 String,
+ BEGIN_TIME bigint
+ HOST String,
+ MSISDN String,
+ ...
+ )STORED BY 'org.apache.carbondata.format'
+ TBLPROPERTIES ( 'DICTIONARY_EXCLUDE'='MSISDN,HOST,IMSI..',
+ 'DICTIONARY_INCLUDE'='Dime_1,END_TIME,BEGIN_TIME..');
+</code></p>
+<ul>
+ <li><strong>For measure type columns with non high accuracy, replace Numeric(20,0) data type with Double data type</strong></li>
+</ul><p>For columns of measure type, not requiring high accuracy, it is suggested to replace Numeric data type with Double to enhance query performance. The create table command can be modified as below :</p><p><code>
+ create table carbondata_table(
+ Dime_1 String,
+ BEGIN_TIME bigint
+ HOST String,
+ MSISDN String,
+ counter_1 double,
+ counter_2 double,
+ ...
+ counter_100 double
+ )STORED BY 'org.apache.carbondata.format'
+ TBLPROPERTIES ( 'DICTIONARY_EXCLUDE'='MSISDN,HOST,IMSI',
+ 'DICTIONARY_INCLUDE'='Dime_1,END_TIME,BEGIN_TIME');
+</code> The result of performance analysis of test-case shows reduction in query execution time from 15 to 3 seconds, thereby improving performance by nearly 5 times.</p>
+<ul>
+ <li><strong>Columns of incremental character should be re-arranged at the end of dimensions</strong></li>
+</ul><p>Consider the following scenario where data is loaded each day and the start_time is incremental for each load, it is suggested to put start_time at the end of dimensions. </p><p>Incremental values are efficient in using min/max index. The create table command can be modified as below :</p><p><code>
+ create table carbondata_table(
+ Dime_1 String,
+ HOST String,
+ MSISDN String,
+ counter_1 double,
+ counter_2 double,
+ BEGIN_TIME bigint,
+ ...
+ counter_100 double
+ )STORED BY 'org.apache.carbondata.format'
+ TBLPROPERTIES ( 'DICTIONARY_EXCLUDE'='MSISDN,HOST,IMSI',
+ 'DICTIONARY_INCLUDE'='Dime_1,END_TIME,BEGIN_TIME');
+</code></p>
+<ul>
+ <li><strong>Avoid adding high cardinality columns to dictionary</strong></li>
+</ul><p>If the system has low memory configuration, then it is suggested to exclude high cardinality columns from the dictionary to enhance load performance. Creation of dictionary for high cardinality columns at time of load will degrade load performance due to excessive memory usage. </p><p>By default CarbonData determines the cardinality at the first data load and allows for dictionary creation only if the cardinality is less than 1 million.</p>
+<h2 id="carbondata-performance">Configurations for Optimizing CarbonData Performance</h2><p>Recently we did some performance POC on CarbonData for Finance and telecommunication Field. It involved detailed queries and aggregation scenarios. After the completion of POC, some of the configurations impacting the performance have been identified and tabulated below :</p>
+<table class="table table-striped table-bordered">
+ <thead>
+ <tr>
+ <th>Parameter </th>
+ <th>Location </th>
+ <th>Used For </th>
+ <th>Description </th>
+ <th>Tuning </th>
+ </tr>
+ </thead>
+ <tbody>
+ <tr>
+ <td>carbon.sort.intermediate.files.limit </td>
+ <td>spark/carbonlib/carbon.properties </td>
+ <td>Data loading </td>
+ <td>During the loading of data, local temp is used to sort the data. This number specifies the minimum number of intermediate files after which the merge sort has to be initiated. </td>
+ <td>Increasing the parameter to a higher value will improve the load performance. For example, when we increase the value from 20 to 100, it increases the data load performance from 35MB/S to more than 50MB/S. Higher values of this parameter consumes more memory during the load. </td>
+ </tr>
+ <tr>
+ <td>carbon.number.of.cores.while.loading </td>
+ <td>spark/carbonlib/carbon.properties </td>
+ <td>Data loading </td>
+ <td>Specifies the number of cores used for data processing during data loading in CarbonData. </td>
+ <td>If you have more number of CPUs, then you can increase the number of CPUs, which will increase the performance. For example if we increase the value from 2 to 4 then the CSV reading performance can increase about 1 times </td>
+ </tr>
+ <tr>
+ <td>carbon.compaction.level.threshold </td>
+ <td>spark/carbonlib/carbon.properties </td>
+ <td>Data loading and Querying </td>
+ <td>For minor compaction, specifies the number of segments to be merged in stage 1 and number of compacted segments to be merged in stage 2. </td>
+ <td>Each CarbonData load will create one segment, if every load is small in size it will generate many small file over a period of time impacting the query performance. Configuring this parameter will merge the small segment to one big segment which will sort the data and improve the performance. For Example in one telecommunication scenario, the performance improves about 2 times after minor compaction. </td>
+ </tr>
+ <tr>
+ <td>spark.sql.shuffle.partitions </td>
+ <td>spark/con/spark-defaults.conf </td>
+ <td>Querying </td>
+ <td>The number of task started when spark shuffle. </td>
+ <td>The value can be 1 to 2 times as much as the executor cores. In an aggregation scenario, reducing the number from 200 to 32 reduced the query time from 17 to 9 seconds. </td>
+ </tr>
+ <tr>
+ <td>num-executors/executor-cores/executor-memory </td>
+ <td>spark/con/spark-defaults.conf </td>
+ <td>Querying </td>
+ <td>The number of executors, CPU cores, and memory used for CarbonData query. </td>
+ <td>In the bank scenario, we provide the 4 CPUs cores and 15 GB for each executor which can get good performance. This 2 value does not mean more the better. It needs to be configured properly in case of limited resources. For example, In the bank scenario, it has enough CPU 32 cores each node but less memory 64 GB each node. So we cannot give more CPU but less memory. For example, when 4 cores and 12GB for each executor. It sometimes happens GC during the query which impact the query performance very much from the 3 second to more than 15 seconds. In this scenario need to increase the memory or decrease the CPU cores. </td>
+ </tr>
+ <tr>
+ <td>carbon.detail.batch.size </td>
+ <td>spark/carbonlib/carbon.properties </td>
+ <td>Data loading </td>
+ <td>The buffer size to store records, returned from the block scan. </td>
+ <td>In limit scenario this parameter is very important. For example your query limit is 1000. But if we set this value to 3000 that means we get 3000 records from scan but spark will only take 1000 rows. So the 2000 remaining are useless. In one Finance test case after we set it to 100, in the limit 1000 scenario the performance increase about 2 times in comparison to if we set this value to 12000. </td>
+ </tr>
+ <tr>
+ <td>carbon.use.local.dir </td>
+ <td>spark/carbonlib/carbon.properties </td>
+ <td>Data loading </td>
+ <td>Whether use YARN local directories for multi-table load disk load balance </td>
+ <td>If this is set it to true CarbonData will use YARN local directories for multi-table load disk load balance, that will improve the data load performance. </td>
+ </tr>
+ </tbody>
+</table>
\ No newline at end of file