文章目录
前言
一、表格存储入门
1.创建实例
2.创建表
二、Java SDK使用介绍
1.官方示例
2.引入库
3.构建AsyncClient 工具类
3.插入数据
4.获取数据
5.官方最佳实际
三、Wide Cloum 模型介绍
四、表设计
总结
前言
因为学过的知识经常忘记,所以就以博客的方式记录下来,偶尔翻看一下加深一下印象。
一、表格存储入门
快速入门,参考官方文档:快速入门 - 表格存储 Tablestore - 阿里云https://help.aliyun.com/document_detail/27286.html
1.创建实例
进入控制台,选择自己合适的地区,创建实例。这里弹出两种购买方式,一种预留模式和按量模式。这里我选择按量模式,因为购买了资源包,一个月1亿的读写CU。预留模式自行研究。两者的区别的话参考官方文档的产品定价。
2.创建表
创建实例完成后,点击实例进入实例管理页面。
点击创建数据表
输入表名,以及主键名,第一个主键为分区键,默认情况下,阿里云会自动管理分区,如果需要预分区的话要联系阿里云技术支持。目前一个表最多支持4个主键,创建完成后点击确认即可创建表。
二、Java SDK使用介绍
1.官方示例
GitHub - aliyun/tablestore-examples: Example code for aliyun tablestore.Example code for aliyun tablestore. Contribute to aliyun/tablestore-examples development by creating an account on GitHub.https://github.com/aliyun/tablestore-examples
2.引入库
4.0.0
com.jackxue
tablestore-study
pom
1.0-SNAPSHOT
wide-column
v2
UTF-8
com.aliyun.openservices
tablestore
5.4.0
commons-io
commons-io
2.4
org.slf4j
slf4j-log4j12
1.7.25
log4j
log4j
1.2.17
3.构建AsyncClient 工具类
package com.jackxue.v2.util;
import com.alicloud.openservices.tablestore.AsyncClient;
import com.alicloud.openservices.tablestore.TunnelClient;
import com.google.gson.Gson;
import org.apache.commons.io.IOUtils;
import java.io.FileInputStream;
import java.io.FileNotFoundException;
import java.io.IOException;
public class TableStoreUtils {
private static AsyncClient asyncClient;
private static TunnelClient tunnelClient;
private static Conf conf;
class Conf {
private String endpoint;
private String accessId;
private String accessKey;
private String instanceName;
public String getEndpoint() {
return endpoint;
}
public void setEndpoint(String endpoint) {
this.endpoint = endpoint;
}
public String getAccessId() {
return accessId;
}
public void setAccessId(String accessId) {
this.accessId = accessId;
}
public String getAccessKey() {
return accessKey;
}
public void setAccessKey(String accessKey) {
this.accessKey = accessKey;
}
public String getInstanceName() {
return instanceName;
}
public void setInstanceName(String instanceName) {
this.instanceName = instanceName;
}
}
static {
FileInputStream fin = null;
try {
fin = new FileInputStream(System.getProperty("user.home") + "/" + "tablestoreConf.json");
String jsonString = IOUtils.toString(fin);
Gson gson = new Gson();
conf = gson.fromJson(jsonString, Conf.class);
asyncClient = new AsyncClient(conf.getEndpoint(),
conf.getAccessId(),
conf.getAccessKey(),
conf.getInstanceName());
tunnelClient = new TunnelClient(conf.getEndpoint(),
conf.getAccessId(),
conf.getAccessKey(),
conf.getInstanceName());
} catch (FileNotFoundException e) {
e.printStackTrace();
} catch (IOException e) {
e.printStackTrace();
} finally {
try {
fin.close();
} catch (IOException e) {
e.printStackTrace();
}
}
}
public static AsyncClient getAsyncClient(){
return asyncClient;
}
public static void close(){
if(asyncClient != null) asyncClient.shutdown();
if(tunnelClient != null){
tunnelClient.shutdown();
}
}
public static TunnelClient getTunnelClient() {
return tunnelClient;
}
}
3.插入数据
@Test
public void test04() throws ParseException, IOException {
String sensor = "270a";
BufferedReader bufferedReader = new BufferedReader(new FileReader("E:\JavaWorkSpaces\tablestore-study\sn3.txt"));
String line;
bufferedReader.readLine(); //去掉首行
long currentLine = 0;
while ((line = bufferedReader.readLine()) != null) {
currentLine++;
String[] split = line.split("t");
String sn = split[0];
String parentSn = split[1];
List columnList = new ArrayList<>();
List columns = Arrays.asList(
new Column("parent_sn", ColumnValue.fromString(parentSn)),
new Column("parent_type", ColumnValue.fromString("采集器")),
new Column("status", ColumnValue.fromString("alert")),
new Column("update_time", ColumnValue.fromLong(System.currentTimeMillis())),
new Column("INV_ST1", ColumnValue.fromString("value1")),
new Column("Et_ge0", ColumnValue.fromString("value2")));
columnList.addAll(columns);
for (int j = 0; j < 200; j++) {
columnList.add(new Column("col" + j, ColumnValue.fromString("col" + j)));
}
deviceLatestInfo.updateDeviceLatestInfo(sensor, sn, columnList);
System.out.println("sn:" + sn + " parent sn:" + parentSn + " line:" + currentLine);
}
bufferedReader.close();
}
public void updateDeviceLatestInfo(String sensor, String sn, List columnList){
RowUpdateChange rowUpdateChange = new RowUpdateChange(tableName, buildPrimaryKey(sensor, sn));
rowUpdateChange.put(columnList);
UpdateRowRequest updateRowRequest = new UpdateRowRequest(rowUpdateChange);
asyncClient.asSyncClient().updateRow(updateRowRequest);
}
public PrimaryKey buildPrimaryKey(String sensor, String sn){
PrimaryKeyBuilder primaryKeyBuilder = PrimaryKeyBuilder.createPrimaryKeyBuilder();
primaryKeyBuilder.addPrimaryKeyColumn("sensor", PrimaryKeyValue.fromString(sensor));
primaryKeyBuilder.addPrimaryKeyColumn("sn", PrimaryKeyValue.fromString(sn));
return primaryKeyBuilder.build();
}
4.获取数据
public Row getDeviceLatestInfo(String sensor, String sn){
SingleRowQueryCriteria rowQueryCriteria = new SingleRowQueryCriteria(tableName, buildPrimaryKey(sensor, sn));
rowQueryCriteria.setMaxVersions(1);
GetRowRequest getRowRequest = new GetRowRequest(rowQueryCriteria);
GetRowResponse getRowResponse = asyncClient.asSyncClient().getRow(getRowRequest);
return getRowResponse.getRow();
}
3.构建AsyncClient 工具类
package com.jackxue.v2.util;
import com.alicloud.openservices.tablestore.AsyncClient;
import com.alicloud.openservices.tablestore.TunnelClient;
import com.google.gson.Gson;
import org.apache.commons.io.IOUtils;
import java.io.FileInputStream;
import java.io.FileNotFoundException;
import java.io.IOException;
public class TableStoreUtils {
private static AsyncClient asyncClient;
private static TunnelClient tunnelClient;
private static Conf conf;
class Conf {
private String endpoint;
private String accessId;
private String accessKey;
private String instanceName;
public String getEndpoint() {
return endpoint;
}
public void setEndpoint(String endpoint) {
this.endpoint = endpoint;
}
public String getAccessId() {
return accessId;
}
public void setAccessId(String accessId) {
this.accessId = accessId;
}
public String getAccessKey() {
return accessKey;
}
public void setAccessKey(String accessKey) {
this.accessKey = accessKey;
}
public String getInstanceName() {
return instanceName;
}
public void setInstanceName(String instanceName) {
this.instanceName = instanceName;
}
}
static {
FileInputStream fin = null;
try {
fin = new FileInputStream(System.getProperty("user.home") + "/" + "tablestoreConf.json");
String jsonString = IOUtils.toString(fin);
Gson gson = new Gson();
conf = gson.fromJson(jsonString, Conf.class);
asyncClient = new AsyncClient(conf.getEndpoint(),
conf.getAccessId(),
conf.getAccessKey(),
conf.getInstanceName());
tunnelClient = new TunnelClient(conf.getEndpoint(),
conf.getAccessId(),
conf.getAccessKey(),
conf.getInstanceName());
} catch (FileNotFoundException e) {
e.printStackTrace();
} catch (IOException e) {
e.printStackTrace();
} finally {
try {
fin.close();
} catch (IOException e) {
e.printStackTrace();
}
}
}
public static AsyncClient getAsyncClient(){
return asyncClient;
}
public static void close(){
if(asyncClient != null) asyncClient.shutdown();
if(tunnelClient != null){
tunnelClient.shutdown();
}
}
public static TunnelClient getTunnelClient() {
return tunnelClient;
}
}
3.插入数据
@Test
public void test04() throws ParseException, IOException {
String sensor = "270a";
BufferedReader bufferedReader = new BufferedReader(new FileReader("E:\JavaWorkSpaces\tablestore-study\sn3.txt"));
String line;
bufferedReader.readLine(); //去掉首行
long currentLine = 0;
while ((line = bufferedReader.readLine()) != null) {
currentLine++;
String[] split = line.split("t");
String sn = split[0];
String parentSn = split[1];
List columnList = new ArrayList<>();
List columns = Arrays.asList(
new Column("parent_sn", ColumnValue.fromString(parentSn)),
new Column("parent_type", ColumnValue.fromString("采集器")),
new Column("status", ColumnValue.fromString("alert")),
new Column("update_time", ColumnValue.fromLong(System.currentTimeMillis())),
new Column("INV_ST1", ColumnValue.fromString("value1")),
new Column("Et_ge0", ColumnValue.fromString("value2")));
columnList.addAll(columns);
for (int j = 0; j < 200; j++) {
columnList.add(new Column("col" + j, ColumnValue.fromString("col" + j)));
}
deviceLatestInfo.updateDeviceLatestInfo(sensor, sn, columnList);
System.out.println("sn:" + sn + " parent sn:" + parentSn + " line:" + currentLine);
}
bufferedReader.close();
}
public void updateDeviceLatestInfo(String sensor, String sn, List columnList){
RowUpdateChange rowUpdateChange = new RowUpdateChange(tableName, buildPrimaryKey(sensor, sn));
rowUpdateChange.put(columnList);
UpdateRowRequest updateRowRequest = new UpdateRowRequest(rowUpdateChange);
asyncClient.asSyncClient().updateRow(updateRowRequest);
}
public PrimaryKey buildPrimaryKey(String sensor, String sn){
PrimaryKeyBuilder primaryKeyBuilder = PrimaryKeyBuilder.createPrimaryKeyBuilder();
primaryKeyBuilder.addPrimaryKeyColumn("sensor", PrimaryKeyValue.fromString(sensor));
primaryKeyBuilder.addPrimaryKeyColumn("sn", PrimaryKeyValue.fromString(sn));
return primaryKeyBuilder.build();
}
4.获取数据
public Row getDeviceLatestInfo(String sensor, String sn){
SingleRowQueryCriteria rowQueryCriteria = new SingleRowQueryCriteria(tableName, buildPrimaryKey(sensor, sn));
rowQueryCriteria.setMaxVersions(1);
GetRowRequest getRowRequest = new GetRowRequest(rowQueryCriteria);
GetRowResponse getRowResponse = asyncClient.asSyncClient().getRow(getRowRequest);
return getRowResponse.getRow();
}
public Row getDeviceLatestInfo(String sensor, String sn){
SingleRowQueryCriteria rowQueryCriteria = new SingleRowQueryCriteria(tableName, buildPrimaryKey(sensor, sn));
rowQueryCriteria.setMaxVersions(1);
GetRowRequest getRowRequest = new GetRowRequest(rowQueryCriteria);
GetRowResponse getRowResponse = asyncClient.asSyncClient().getRow(getRowRequest);
return getRowResponse.getRow();
}
5.官方最佳实际
三、Wide Cloum 模型介绍
上图为官方的图,这个图非常形象的介绍了这个模式的组成。我感觉和hbase的存储模型类似。Hbase 的RowKey 和Primary Keys 相对应,这个模型感觉把Hbase的rowkey拆分成多个的感觉,其他没有什么太大不同。
四、表设计
在大数据的情况下,数据倾斜应该是影响查询等效率的最大原因之一。所以在选择使用分区键时需要考虑让数据尽量的分布均匀。
热点问题:
这个表设计有很明显的热点问题,每次数据写入都是在表的末尾追加数据,而数据是按照分区键的范围进行分区的,也就是每次数据写入都会写入最后一个分区,而无法把写入负载平衡到多台机器上。
解决方案一:
使用MachineIp作为分区键
解决方案二:
使用MachineIp作为分区键,把IP做一次MD5拼接,避免某个IP段热点数据问题,如下图:
-
控制一个分区的数据大小在10GB以内
- 不要使用递增的分区键
- 主键长度不宜过长,控制在1K以内,尽可能小提升查询效率
好记性不如烂笔头,下次把索引和通道总结一下。



