转自:http://blog.csdn.NET/wo334499/article/details/51689549
RDD
优点:
- 编译时类型安全
编译时就能检查出类型错误 - 面向对象的编程风格
直接通过类名点的方式来操作数据
缺点:
- 序列化和反序列化的性能开销
无论是集群间的通信, 还是IO操作都需要对对象的结构和数据进行序列化和反序列化. - GC的性能开销
频繁的创建和销毁对象, 势必会增加GC
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18
- 19
- 20
- 21
- 22
- 23
- 24
- 25
- 26
DataFrame
DataFrame引入了schema和off-heap
-
schema : RDD每一行的数据, 结构都是一样的. 这个结构就存储在schema中.Spark通过schame就能够读懂数据, 因此在通信和IO时就只需要序列化和反序列化数据, 而结构的部分就可以省略了.
-
off-heap : 意味着JVM堆以外的内存, 这些内存直接受操作系统管理(而不是JVM)。Spark能够以二进制的形式序列化数据(不包括结构)到off-heap中, 当要操作数据时, 就直接操作off-heap内存. 由于Spark理解schema, 所以知道该如何操作.
off-heap就像地盘, schema就像地图, Spark有地图又有自己地盘了, 就可以自己说了算了, 不再受JVM的限制, 也就不再收GC的困扰了.
通过schema和off-heap, DataFrame解决了RDD的缺点, 但是却丢了RDD的优点. DataFrame不是类型安全的, API也不是面向对象风格的.
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18
- 19
- 20
- 21
- 22
- 23
- 24
- 25
- 26
- 27
DataSet
DataSet结合了RDD和DataFrame的优点, 并带来的一个新的概念Encoder
当序列化数据时, Encoder产生字节码与off-heap进行交互, 能够达到按需访问数据的效果, 而不用反序列化整个对象. Spark还没有提供自定义Encoder的API, 但是未来会加入.
下面看DataFrame和DataSet在2.0.0-preview中的实现
下面这段代码, 在1.6.x中创建的是DataFrame
<code
class="hljs mathematica has-numbering" style="display: block; padding:
0px; color: inherit; box-sizing: border-box; font-family: 'Source Code
Pro', monospace;font-size:undefined; white-space: pre; border-radius:
0px; word-wrap: normal; background: transparent;">//
上文DataFrame示例中提取出来的
val idAgeRDDRow = sc.parallelize(<span class="hljs-keyword"
style="color: rgb(0, 0, 136); box-sizing:
border-box;">Array</span>(<span class="hljs-keyword"
style="color: rgb(0, 0, 136); box-sizing:
border-box;">Row</span>(<span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">1</span>, <span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">30</span>), <span class="hljs-keyword"
style="color: rgb(0, 0, 136); box-sizing:
border-box;">Row</span>(<span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">2</span>, <span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">29</span>), <span class="hljs-keyword"
style="color: rgb(0, 0, 136); box-sizing:
border-box;">Row</span>(<span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">4</span>, <span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">21</span>)))
val schema = StructType(<span class="hljs-keyword" style="color:
rgb(0, 0, 136); box-sizing:
border-box;">Array</span>(StructField(<span
class="hljs-string" style="color: rgb(0, 136, 0); box-sizing:
border-box;">"id"</span>, DataTypes.IntegerType),
StructField(<span class="hljs-string" style="color: rgb(0, 136, 0);
box-sizing: border-box;">"age"</span>, DataTypes.IntegerType)))
val idAgeDF = sqlContext.createDataFrame(idAgeRDDRow,
schema)</code><ul class="pre-numbering" style="box-sizing:
border-box; position: absolute; width: 50px; top: 0px; left: 0px;
margin: 0px; padding: 6px 0px 40px; border-right-width: 1px;
border-right-style: solid; border-right-color: rgb(221, 221, 221);
list-style: none; text-align: right; background-color: rgb(238, 238,
238);"><li style="box-sizing: border-box; padding: 0px
5px;">1</li><li style="box-sizing: border-box; padding: 0px
5px;">2</li><li style="box-sizing: border-box; padding: 0px
5px;">3</li><li style="box-sizing: border-box; padding: 0px
5px;">4</li><li style="box-sizing: border-box; padding: 0px
5px;">5</li><li style="box-sizing: border-box; padding: 0px
5px;">6</li></ul>
但是同样的代码在2.0.0-preview中, 创建的虽然还叫DataFrame
<code
class="hljs python has-numbering" style="display: block; padding: 0px;
color: inherit; box-sizing: border-box; font-family: 'Source Code Pro',
monospace;font-size:undefined; white-space: pre; border-radius: 0px;
word-wrap: normal; background: transparent;">//
sqlContext.createDataFrame(idAgeRDDRow, schema) 方法的实现, 返回值依然是DataFrame
<span class="hljs-function" style="box-sizing:
border-box;"><span class="hljs-keyword" style="color: rgb(0, 0,
136); box-sizing: border-box;">def</span> <span
class="hljs-title" style="box-sizing:
border-box;">createDataFrame</span><span class="hljs-params"
style="color: rgb(102, 0, 102); box-sizing: border-box;">(rowRDD:
RDD[Row], schema: StructType)</span>:</span> DataFrame = {
sparkSession.createDataFrame(rowRDD, schema)
}</code><ul class="pre-numbering" style="box-sizing:
border-box; position: absolute; width: 50px; top: 0px; left: 0px;
margin: 0px; padding: 6px 0px 40px; border-right-width: 1px;
border-right-style: solid; border-right-color: rgb(221, 221, 221);
list-style: none; text-align: right; background-color: rgb(238, 238,
238);"><li style="box-sizing: border-box; padding: 0px
5px;">1</li><li style="box-sizing: border-box; padding: 0px
5px;">2</li><li style="box-sizing: border-box; padding: 0px
5px;">3</li><li style="box-sizing: border-box; padding: 0px
5px;">4</li></ul>
但是其实却是DataSet, 因为DataFrame被声明为Dataset[Row]
<code
class="hljs r has-numbering" style="display: block; padding: 0px;
color: inherit; box-sizing: border-box; font-family: 'Source Code Pro',
monospace;font-size:undefined; white-space: pre; border-radius: 0px;
word-wrap: normal; background: transparent;">package object sql {
// <span class="hljs-keyword" style="color: rgb(0, 0, 136);
box-sizing: border-box;">...</span>省略了不相关的代码
type DataFrame = Dataset[Row]
}</code><ul class="pre-numbering" style="box-sizing:
border-box; position: absolute; width: 50px; top: 0px; left: 0px;
margin: 0px; padding: 6px 0px 40px; border-right-width: 1px;
border-right-style: solid; border-right-color: rgb(221, 221, 221);
list-style: none; text-align: right; background-color: rgb(238, 238,
238);"><li style="box-sizing: border-box; padding: 0px
5px;">1</li><li style="box-sizing: border-box; padding: 0px
5px;">2</li><li style="box-sizing: border-box; padding: 0px
5px;">3</li><li style="box-sizing: border-box; padding: 0px
5px;">4</li><li style="box-sizing: border-box; padding: 0px
5px;">5</li></ul>
因此当我们从1.6.x迁移到2.0.0的时候, 无需任何修改就直接用上了DataSet.
下面是一段DataSet的示例代码
<code
class="hljs scala has-numbering" style="display: block; padding: 0px;
color: inherit; box-sizing: border-box; font-family: 'Source Code Pro',
monospace;font-size:undefined; white-space: pre; border-radius: 0px;
word-wrap: normal; background: transparent;"><span
class="hljs-keyword" style="color: rgb(0, 0, 136); box-sizing:
border-box;">import</span>
org.apache.spark.sql.types.{DataTypes, StructField, StructType}
<span class="hljs-keyword" style="color: rgb(0, 0, 136); box-sizing:
border-box;">import</span> org.apache.spark.sql.{Row,
SQLContext}
<span class="hljs-keyword" style="color: rgb(0, 0, 136); box-sizing:
border-box;">import</span> org.apache.spark.{SparkConf,
SparkContext}
<span class="hljs-class" style="box-sizing: border-box;"><span
class="hljs-keyword" style="color: rgb(0, 0, 136); box-sizing:
border-box;">object</span> <span class="hljs-title"
style="box-sizing: border-box; color: rgb(102, 0,
102);">Test</span> {</span>
<span class="hljs-keyword" style="color: rgb(0, 0, 136);
box-sizing: border-box;">def</span> main(args: Array[String]) {
<span class="hljs-keyword" style="color: rgb(0, 0, 136);
box-sizing: border-box;">val</span> conf = <span
class="hljs-keyword" style="color: rgb(0, 0, 136); box-sizing:
border-box;">new</span> SparkConf().setAppName(<span
class="hljs-string" style="color: rgb(0, 136, 0); box-sizing:
border-box;">"test"</span>).setMaster(<span
class="hljs-string" style="color: rgb(0, 136, 0); box-sizing:
border-box;">"local"</span>) <span class="hljs-comment"
style="color: rgb(136, 0, 0); box-sizing: border-box;">//
调试的时候一定不要用local[*]</span>
<span class="hljs-keyword" style="color: rgb(0, 0, 136);
box-sizing: border-box;">val</span> sc = <span
class="hljs-keyword" style="color: rgb(0, 0, 136); box-sizing:
border-box;">new</span> SparkContext(conf)
<span class="hljs-keyword" style="color: rgb(0, 0, 136);
box-sizing: border-box;">val</span> sqlContext = <span
class="hljs-keyword" style="color: rgb(0, 0, 136); box-sizing:
border-box;">new</span> SQLContext(sc)
<span class="hljs-keyword" style="color: rgb(0, 0, 136);
box-sizing: border-box;">import</span> sqlContext.implicits._
<span class="hljs-keyword" style="color: rgb(0, 0, 136);
box-sizing: border-box;">val</span> idAgeRDDRow =
sc.parallelize(Array(Row(<span class="hljs-number" style="color:
rgb(0, 102, 102); box-sizing: border-box;">1</span>, <span
class="hljs-number" style="color: rgb(0, 102, 102); box-sizing:
border-box;">30</span>), Row(<span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">2</span>, <span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">29</span>), Row(<span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">4</span>, <span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">21</span>)))
<span class="hljs-keyword" style="color: rgb(0, 0, 136);
box-sizing: border-box;">val</span> schema =
StructType(Array(StructField(<span class="hljs-string" style="color:
rgb(0, 136, 0); box-sizing: border-box;">"id"</span>,
DataTypes.IntegerType), StructField(<span class="hljs-string"
style="color: rgb(0, 136, 0); box-sizing:
border-box;">"age"</span>, DataTypes.IntegerType)))
<span class="hljs-comment" style="color: rgb(136, 0, 0);
box-sizing: border-box;">// 在2.0.0-preview中这行代码创建出的DataFrame,
其实是DataSet[Row]</span>
<span class="hljs-keyword" style="color: rgb(0, 0, 136);
box-sizing: border-box;">val</span> idAgeDS =
sqlContext.createDataFrame(idAgeRDDRow, schema)
<span class="hljs-comment" style="color: rgb(136, 0, 0);
box-sizing: border-box;">// 在2.0.0-preview中, 还不支持自定的Encoder, Row类型不行,
自定义的bean也不行</span>
<span class="hljs-comment" style="color: rgb(136, 0, 0);
box-sizing: border-box;">//
官方文档也有写通过bean创建Dataset的例子,但是我运行时并不能成功</span>
<span class="hljs-comment" style="color: rgb(136, 0, 0);
box-sizing: border-box;">// 所以目前需要用创建DataFrame的方法,
来创建DataSet[Row]</span>
<span class="hljs-comment" style="color: rgb(136, 0, 0);
box-sizing: border-box;">//
sqlContext.createDataset(idAgeRDDRow)</span>
<span class="hljs-comment" style="color: rgb(136, 0, 0);
box-sizing: border-box;">// 目前支持String, Integer,
Long等类型直接创建Dataset</span>
Seq(<span class="hljs-number" style="color: rgb(0, 102, 102);
box-sizing: border-box;">1</span>, <span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">2</span>, <span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">3</span>).toDS().show()
sqlContext.createDataset(sc.parallelize(Array(<span
class="hljs-number" style="color: rgb(0, 102, 102); box-sizing:
border-box;">1</span>, <span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">2</span>, <span class="hljs-number"
style="color: rgb(0, 102, 102); box-sizing:
border-box;">3</span>))).show()
}
}</code>
参考
Introducing Apache Spark Datasets
APACHE SPARK: RDD, DATAFRAME OR DATASET?
RDD、DataFrame和DataSet的区别
Spark 2.0.0-preview 官方文档