试用Cassandra,其写效率太差
Cassandra是由Facebook贡献的开源分布式数据库。其遵从NoSql理念，是结合了Dynamo与BigTable的产物。最近Twitter和Digg都将其数据库由MySql迁往Cassandra。看到其发展势头不错，我就下载下来，做了个测试。
测试环境：
分别在两台机器上部署cassandra.这里说明下关键配置：
配置文件路径是%Cassandra_Home%\conf\storage-conf.xml
<Storage><!--两台机器的ClusterName必须相同，作为集群标识 --> <ClusterName>BurceServers</ClusterName> <AutoBootstrap>false</AutoBootstrap> <Keyspaces> <Keyspace Name="Keyspace1"> <KeysCachedFraction>0.01</KeysCachedFraction> <ColumnFamily CompareWith="BytesType" Name="Standard1"/> <ColumnFamily CompareWith="UTF8Type" Name="Standard2"/> <ColumnFamily CompareWith="TimeUUIDType" Name="StandardByUUID1"/> <ColumnFamily ColumnType="Super" CompareWith="UTF8Type" CompareSubcolumnsWith="UTF8Type" Name="Super1" Comment="A column family with supercolumns, whose column and subcolumn names are UTF8 strings"/> </Keyspace> </Keyspaces> <Partitioner>org.apache.cassandra.dht.RandomPartitioner</Partitioner> <InitialToken></InitialToken> <EndPointSnitch>org.apache.cassandra.locator.EndPointSnitch</EndPointSnitch> <ReplicaPlacementStrategy>org.apache.cassandra.locator.RackUnawareStrategy</ReplicaPlacementStrategy> <ReplicationFactor>1</ReplicationFactor> <CommitLogDirectory>c:/cassandra/lib/cassandra/commitlog</CommitLogDirectory> <DataFileDirectories> <DataFileDirectory>c:/cassandra/lib/cassandra/data</DataFileDirectory> </DataFileDirectories> <CalloutLocation>c:/cassandra/lib/cassandra/callouts</CalloutLocation> <StagingFileDirectory>c:/cassandra/lib/cassandra/staging</StagingFileDirectory><!--在这里可以添加多个cassandra服务器--> <Seeds> <Seed>10.219.101.101</Seed><Seed>10.219.101.121</Seed> </Seeds> <RpcTimeoutInMillis>5000</RpcTimeoutInMillis> <CommitLogRotationThresholdInMB>128</CommitLogRotationThresholdInMB><!--监听地址必须是本机IP--> <ListenAddress>10.219.101.101</ListenAddress> <StoragePort>7000</StoragePort> <ControlPort>7001</ControlPort><!--基于Thrift的cassandra客户端监听地址--> <ThriftAddress>10.219.101.101</ThriftAddress> <ThriftPort>9160</ThriftPort> <ThriftFramedTransport>false</ThriftFramedTransport> <SlicedBufferSizeInKB>64</SlicedBufferSizeInKB> <ColumnIndexSizeInKB>64</ColumnIndexSizeInKB> <MemtableSizeInMB>64</MemtableSizeInMB> <MemtableObjectCountInMillions>0.1</MemtableObjectCountInMillions> <MemtableFlushAfterMinutes>60</MemtableFlushAfterMinutes> <ConcurrentReads>8</ConcurrentReads> <ConcurrentWrites>32</ConcurrentWrites> <CommitLogSync>periodic</CommitLogSync> <CommitLogSyncPeriodInMS>10000</CommitLogSyncPeriodInMS> <GCGraceSeconds>864000</GCGraceSeconds> <BinaryMemtableSizeInMB>256</BinaryMemtableSizeInMB></Storage>
除增加了一个cassandra的服务器外，基本采用默认配置。
测试代码：
/** * */package com.tpri.sis.test;import java.io.UnsupportedEncodingException;import java.nio.charset.Charset;import me.prettyprint.cassandra.service.CassandraClient;import org.apache.cassandra.service.Cassandra;import org.apache.cassandra.service.ColumnPath;import org.apache.cassandra.service.ConsistencyLevel;import org.apache.cassandra.service.InvalidRequestException;import org.apache.cassandra.service.TimedOutException;import org.apache.cassandra.service.UnavailableException;import org.apache.commons.lang.RandomStringUtils;import org.apache.commons.lang.StringUtils;import org.apache.thrift.TException;import org.apache.thrift.protocol.TBinaryProtocol;import org.apache.thrift.protocol.TProtocol;import org.apache.thrift.transport.TSocket;import org.apache.thrift.transport.TTransport;import org.apache.thrift.transport.TTransportException;/** * @author brucepang * */public class CassandraClientDemo {/** * */public CassandraClientDemo() {}/** * @param args */public static void main(String[] args) {try {TTransport tr = new TSocket("10.219.101.101", 9160);TProtocol pro = new TBinaryProtocol(tr);Cassandra.Client cli = new Cassandra.Client(pro);tr.open();String key = null;String name, age;ColumnPath namePath = new ColumnPath("Standard1", null, "name".getBytes("UTF-8"));ColumnPath agePath = new ColumnPath("Standard1", null, "age".getBytes("UTF-8"));String keySpace = "Keyspace1";long time = 0;long l1 = System.currentTimeMillis();for (int i = 0; i < 100; i++) {key = String.valueOf(i);name = RandomStringUtils.random(5,"abcdefghefsdf");time = System.currentTimeMillis();cli.insert(keySpace, key, namePath, name.getBytes("UTF-8"),time, ConsistencyLevel.ONE);cli.insert(keySpace, key, agePath, key.getBytes("UTF-8"), time,ConsistencyLevel.ONE);}long l2 = System.currentTimeMillis();long ch = l2 - l1;System.out.println(ch);} catch (TTransportException e) {e.printStackTrace();} catch (UnsupportedEncodingException e) {// TODO Auto-generated catch blocke.printStackTrace();} catch (InvalidRequestException e) {// TODO Auto-generated catch blocke.printStackTrace();} catch (UnavailableException e) {// TODO Auto-generated catch blocke.printStackTrace();} catch (TimedOutException e) {// TODO Auto-generated catch blocke.printStackTrace();} catch (TException e) {// TODO Auto-generated catch blocke.printStackTrace();}}}测试结果：1 写100条数据，耗时59922毫秒，将近1分钟；
1 楼 wing5jface 2010-03-19 如果是单机的话，我跑你的程序，只用了94ms呀 2 楼 pangyi 2010-03-22 单机效率还可以。cassandra的主要特性是其分布特性。
3 楼 waterdh 2010-04-02 可能机器问题吧。
我的2台集群测试下来效果还不错，单线程插入2w条数据花了13s。
4 楼 zdyhlp 2010-04-19 我在windows机器上单节点试了下。4秒写了10000条记录。
TTransport tr = new TSocket("localhost", 9160);
TProtocol pro = new TBinaryProtocol(tr);
Cassandra.Client cli = new Cassandra.Client(pro);
tr.open();
String key = null;
String name, age;
ColumnPath namePath = new ColumnPath("Standard1");
namePath.setColumn("name".getBytes("UTF-8"));
ColumnPath agePath = new ColumnPath("Standard1");
agePath.setColumn("age".getBytes("UTF-8"));
5 楼 pangyi 2010-05-17 最近试用了下MongoDB,其读写效率都不错。 6 楼 machoche 2010-06-07 有些库找不到？ 7 楼 linliangyi2007 2010-07-10 楼主杯具了，cassandra被你用成残废了。你的配置中seed为啥又两个啊，做群集seed 1个就够了。