hbase性能调优之压缩测试

文章概述：

1、顺序写

2、顺序读

3、随机写

4、随机读

5、SCAN数据

---

0 性能测试工具

hbase org.apache.hadoop.hbase.PerformanceEvaluation

Usage: java org.apache.hadoop.hbase.PerformanceEvaluation

  [--nomapred] [--rows=ROWS] [--table=NAME]

  [--compress=TYPE] [--blockEncoding=TYPE] [-D<property=value>]* <command> <nclients>

 

Options:

 nomapred        Run multiple clients using threads (rather than use mapreduce)

 rows            Rows each client runs. Default: One million

 sampleRate      Execute test on a sample of total rows. Only supported by randomRead. Default: 1.0

 table           Alternate table name. Default: 'TestTable'

 compress        Compression type to use (GZ, LZO, ...). Default: 'NONE'

 flushCommits    Used to determine if the test should flush the table. Default: false

 writeToWAL      Set writeToWAL on puts. Default: True

 presplit        Create presplit table. Recommended for accurate perf analysis (see guide).  Default: disabled

 inmemory        Tries to keep the HFiles of the CF inmemory as far as possible. Not guaranteed that reads are always served from memory.  Default: false

 latency         Set to report operation latencies. Currently only supported by randomRead test. Default: False

 

 Note: -D properties will be applied to the conf used.

  For example:

   -Dmapred.output.compress=true

   -Dmapreduce.task.timeout=60000

 

Command:

 filterScan      Run scan test using a filter to find a specific row based on it's value (make sure to use --rows=20)

 randomRead      Run random read test

 randomSeekScan  Run random seek and scan 100 test

 randomWrite     Run random write test

 scan            Run scan test (read every row)

 scanRange10     Run random seek scan with both start and stop row (max 10 rows)

 scanRange100    Run random seek scan with both start and stop row (max 100 rows)

 scanRange1000   Run random seek scan with both start and stop row (max 1000 rows)

 scanRange10000  Run random seek scan with both start and stop row (max 10000 rows)

 sequentialRead  Run sequential read test

 sequentialWrite Run sequential write test

 

Args:

 nclients        Integer. Required. Total number of clients (and HRegionServers)

                 running: 1 <= value <= 500

Examples:

 To run a single evaluation client:

 $ bin/hbase org.apache.hadoop.hbase.PerformanceEvaluation sequentialWrite 1

---

1 顺序写测试

测试基准：10个并发客户端，写入200万行数据

1.1 无压缩顺序写 

hbase org.apache.hadoop.hbase.PerformanceEvaluation --rows=2000000 --nomapred --table=none_test randomRead 10

1.2 LZO顺序写

hbase org.apache.hadoop.hbase.PerformanceEvaluation --rows=2000000 --nomapred --compress=LZO --table=none_test randomRead 10

1.3 有无压缩对比

对比指标	不压缩	LZO压缩
插入100万行数据平均时间
文件大小(1000万行数据)	19.2G	4.7G

---

2 顺序读测试

2.1 无压缩顺序读

 

2.2 LZO顺序读

 

2.3 有无压缩对比

 

 

 

 

 

 

 

参考文献：

[1] 性能调优 | HBase表操作使用LZO