2 通过案例对SparkStreaming透彻理解之二

Spark Core是基于RDD形成的，RDD之间都会有依赖关系。而Spark Streaming是在RDD之上增加了时间维度，DStream就是RDD的模板，随着时间的流逝不断地实例化DStream，以数据进行填充DStream。DStream的依赖关系构成Dstream Graph，根据DStream Graph的依赖转换成RDD的依赖。DStream Graph就是静态的RDD DAG模板。
DStream之是逻辑级别，而RDD才是物理级别。Dstrem只是在RDD的基础上加上了时间的维度，所以整个Spark Streaming就是时空维度。

private[streaming] var generatedRDDs = new HashMap[Time, RDD[T]] ()

DStream在计算的时候compute需要传入一个时间参数，通过时间获取相应的RDD，然后再对RDD进行计算

def compute(validTime: Time): Option[RDD[T]]

我们运行一个简单的Spark Streaming例子

objectNetworkWordCount {

  defmain(args:Array[String]) {
    if (args.length< 2) {
      System.err.println("Usage: NetworkWordCount<hostname> <port>")
      System.exit(1)
    }

    val sparkConf= newSparkConf().setAppName("NetworkWordCount").setMaster("local[2]")
    val ssc = newStreamingContext(sparkConf,Seconds(1))
    val lines= ssc.socketTextStream(args(0), args(1).toInt,StorageLevel.MEMORY_AND_DISK_SER)
    val words= lines.flatMap(_.split(""))
    val wordCounts= words.map(x => (x,1)).reduceByKey(_ + _)
    wordCounts.print()
    ssc.start()
    ssc.awaitTermination()
  }
}

我们查看SparkStreaming的运行日志，就可以看出和RDD的运行几乎是一致的:

2016-05-20 06:54:10,056 INFO  [JobScheduler] scheduler.JobScheduler (Logging.scala:logInfo(58)) - Finished job streaming job 1463698450000 ms.0 from job set of time 1463698450000 ms
2016-05-20 06:54:10,056 INFO  [JobScheduler] scheduler.JobScheduler (Logging.scala:logInfo(58)) - Total delay: 0.056 s for time 1463698450000 ms (execution: 0.044 s)
2016-05-20 06:54:10,057 INFO  [JobGenerator] rdd.ShuffledRDD (Logging.scala:logInfo(58)) - Removing RDD 4 from persistence list
2016-05-20 06:54:10,062 INFO  [JobGenerator] rdd.MapPartitionsRDD (Logging.scala:logInfo(58)) - Removing RDD 3 from persistence list
2016-05-20 06:54:10,062 INFO  [block-manager-slave-async-thread-pool-0] storage.BlockManager (Logging.scala:logInfo(58)) - Removing RDD 4
2016-05-20 06:54:10,062 INFO  [block-manager-slave-async-thread-pool-1] storage.BlockManager (Logging.scala:logInfo(58)) - Removing RDD 3
2016-05-20 06:54:10,063 INFO  [JobGenerator] rdd.MapPartitionsRDD (Logging.scala:logInfo(58)) - Removing RDD 2 from persistence list
2016-05-20 06:54:10,063 INFO  [block-manager-slave-async-thread-pool-2] storage.BlockManager (Logging.scala:logInfo(58)) - Removing RDD 2
2016-05-20 06:54:10,064 INFO  [JobGenerator] rdd.BlockRDD (Logging.scala:logInfo(58)) - Removing RDD 1 from persistence list
2016-05-20 06:54:10,067 INFO  [block-manager-slave-async-thread-pool-7] storage.BlockManager (Logging.scala:logInfo(58)) - Removing RDD 1
2016-05-20 06:54:10,067 INFO  [JobGenerator] dstream.SocketInputDStream (Logging.scala:logInfo(58)) - Removing blocks of RDD BlockRDD[1] at socketTextStream at NetworkWordCount.scala:21 of time 1463698450000 ms
2016-05-20 06:54:10,068 INFO  [JobGenerator] scheduler.ReceivedBlockTracker (Logging.scala:logInfo(58)) - Deleting batches ArrayBuffer()
2016-05-20 06:54:10,068 INFO  [JobGenerator] scheduler.InputInfoTracker (Logging.scala:logInfo(58)) - remove old batch metadata: 
2016-05-20 06:54:15,015 INFO  [Spark Context Cleaner] spark.ContextCleaner (Logging.scala:logInfo(58)) - Cleaned accumulator 5
2016-05-20 06:54:15,015 INFO  [JobGenerator] scheduler.JobScheduler (Logging.scala:logInfo(58)) - Added jobs for time 1463698455000 ms
2016-05-20 06:54:15,016 INFO  [dispatcher-event-loop-0] storage.BlockManagerInfo (Logging.scala:logInfo(58)) - Removed broadcast_4_piece0 on localhost:62612 in memory (size: 1626.0 B, free: 1311.0 MB)
2016-05-20 06:54:15,016 INFO  [JobScheduler] scheduler.JobScheduler (Logging.scala:logInfo(58)) - Starting job streaming job 1463698455000 ms.0 from job set of time 1463698455000 ms
2016-05-20 06:54:15,020 INFO  [streaming-job-executor-0] spark.SparkContext (Logging.scala:logInfo(58)) - Starting job: print at NetworkWordCount.scala:26
2016-05-20 06:54:15,021 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Registering RDD 11 (map at NetworkWordCount.scala:25)
2016-05-20 06:54:15,021 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Got job 5 (print at NetworkWordCount.scala:26) with 1 output partitions
2016-05-20 06:54:15,021 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Final stage: ResultStage 10 (print at NetworkWordCount.scala:26)
2016-05-20 06:54:15,021 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Parents of final stage: List(ShuffleMapStage 9)
2016-05-20 06:54:15,021 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Missing parents: List()
2016-05-20 06:54:15,022 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Submitting ResultStage 10 (ShuffledRDD[12] at reduceByKey at NetworkWordCount.scala:25), which has no missing parents
2016-05-20 06:54:15,024 INFO  [dag-scheduler-event-loop] storage.MemoryStore (Logging.scala:logInfo(58)) - Block broadcast_5 stored as values in memory (estimated size 2.6 KB, free 60.9 KB)
2016-05-20 06:54:15,026 INFO  [dag-scheduler-event-loop] storage.MemoryStore (Logging.scala:logInfo(58)) - Block broadcast_5_piece0 stored as bytes in memory (estimated size 1627.0 B, free 62.5 KB)
2016-05-20 06:54:15,027 INFO  [dispatcher-event-loop-2] storage.BlockManagerInfo (Logging.scala:logInfo(58)) - Added broadcast_5_piece0 in memory on localhost:62612 (size: 1627.0 B, free: 1311.0 MB)
2016-05-20 06:54:15,027 INFO  [dag-scheduler-event-loop] spark.SparkContext (Logging.scala:logInfo(58)) - Created broadcast 5 from broadcast at DAGScheduler.scala:1006
2016-05-20 06:54:15,027 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Submitting 1 missing tasks from ResultStage 10 (ShuffledRDD[12] at reduceByKey at NetworkWordCount.scala:25)
2016-05-20 06:54:15,028 INFO  [dag-scheduler-event-loop] scheduler.TaskSchedulerImpl (Logging.scala:logInfo(58)) - Adding task set 10.0 with 1 tasks
2016-05-20 06:54:15,028 INFO  [dispatcher-event-loop-3] scheduler.TaskSetManager (Logging.scala:logInfo(58)) - Starting task 0.0 in stage 10.0 (TID 5, localhost, partition 0,PROCESS_LOCAL, 1894 bytes)
2016-05-20 06:54:15,029 INFO  [Executor task launch worker-1] executor.Executor (Logging.scala:logInfo(58)) - Running task 0.0 in stage 10.0 (TID 5)
2016-05-20 06:54:15,031 INFO  [Executor task launch worker-1] storage.ShuffleBlockFetcherIterator (Logging.scala:logInfo(58)) - Getting 0 non-empty blocks out of 0 blocks
2016-05-20 06:54:15,031 INFO  [Executor task launch worker-1] storage.ShuffleBlockFetcherIterator (Logging.scala:logInfo(58)) - Started 0 remote fetches in 0 ms
2016-05-20 06:54:15,032 INFO  [Executor task launch worker-1] executor.Executor (Logging.scala:logInfo(58)) - Finished task 0.0 in stage 10.0 (TID 5). 1161 bytes result sent to driver
2016-05-20 06:54:15,033 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - ResultStage 10 (print at NetworkWordCount.scala:26) finished in 0.005 s
2016-05-20 06:54:15,033 INFO  [task-result-getter-0] scheduler.TaskSetManager (Logging.scala:logInfo(58)) - Finished task 0.0 in stage 10.0 (TID 5) in 5 ms on localhost (1/1)
2016-05-20 06:54:15,034 INFO  [task-result-getter-0] scheduler.TaskSchedulerImpl (Logging.scala:logInfo(58)) - Removed TaskSet 10.0, whose tasks have all completed, from pool 
2016-05-20 06:54:15,034 INFO  [streaming-job-executor-0] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Job 5 finished: print at NetworkWordCount.scala:26, took 0.013596 s
2016-05-20 06:54:15,040 INFO  [Spark Context Cleaner] spark.ContextCleaner (Logging.scala:logInfo(58)) - Cleaned accumulator 6
2016-05-20 06:54:15,041 INFO  [dispatcher-event-loop-1] storage.BlockManagerInfo (Logging.scala:logInfo(58)) - Removed broadcast_5_piece0 on localhost:62612 in memory (size: 1627.0 B, free: 1311.0 MB)
2016-05-20 06:54:15,043 INFO  [streaming-job-executor-0] spark.SparkContext (Logging.scala:logInfo(58)) - Starting job: print at NetworkWordCount.scala:26
2016-05-20 06:54:15,044 INFO  [dag-scheduler-event-loop] spark.MapOutputTrackerMaster (Logging.scala:logInfo(58)) - Size of output statuses for shuffle 2 is 82 bytes
2016-05-20 06:54:15,044 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Got job 6 (print at NetworkWordCount.scala:26) with 1 output partitions
2016-05-20 06:54:15,044 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Final stage: ResultStage 12 (print at NetworkWordCount.scala:26)
2016-05-20 06:54:15,044 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Parents of final stage: List(ShuffleMapStage 11)
2016-05-20 06:54:15,044 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Missing parents: List()
2016-05-20 06:54:15,045 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Submitting ResultStage 12 (ShuffledRDD[12] at reduceByKey at NetworkWordCount.scala:25), which has no missing parents
2016-05-20 06:54:15,046 INFO  [dag-scheduler-event-loop] storage.MemoryStore (Logging.scala:logInfo(58)) - Block broadcast_6 stored as values in memory (estimated size 2.6 KB, free 60.9 KB)
2016-05-20 06:54:15,048 INFO  [dag-scheduler-event-loop] storage.MemoryStore (Logging.scala:logInfo(58)) - Block broadcast_6_piece0 stored as bytes in memory (estimated size 1627.0 B, free 62.5 KB)
2016-05-20 06:54:15,050 INFO  [dispatcher-event-loop-0] storage.BlockManagerInfo (Logging.scala:logInfo(58)) - Added broadcast_6_piece0 in memory on localhost:62612 (size: 1627.0 B, free: 1311.0 MB)
2016-05-20 06:54:15,050 INFO  [dag-scheduler-event-loop] spark.SparkContext (Logging.scala:logInfo(58)) - Created broadcast 6 from broadcast at DAGScheduler.scala:1006
2016-05-20 06:54:15,050 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Submitting 1 missing tasks from ResultStage 12 (ShuffledRDD[12] at reduceByKey at NetworkWordCount.scala:25)
2016-05-20 06:54:15,050 INFO  [dag-scheduler-event-loop] scheduler.TaskSchedulerImpl (Logging.scala:logInfo(58)) - Adding task set 12.0 with 1 tasks
2016-05-20 06:54:15,051 INFO  [dispatcher-event-loop-3] scheduler.TaskSetManager (Logging.scala:logInfo(58)) - Starting task 0.0 in stage 12.0 (TID 6, localhost, partition 1,PROCESS_LOCAL, 1894 bytes)
2016-05-20 06:54:15,052 INFO  [Executor task launch worker-1] executor.Executor (Logging.scala:logInfo(58)) - Running task 0.0 in stage 12.0 (TID 6)
2016-05-20 06:54:15,054 INFO  [Executor task launch worker-1] storage.ShuffleBlockFetcherIterator (Logging.scala:logInfo(58)) - Getting 0 non-empty blocks out of 0 blocks
2016-05-20 06:54:15,054 INFO  [Executor task launch worker-1] storage.ShuffleBlockFetcherIterator (Logging.scala:logInfo(58)) - Started 0 remote fetches in 1 ms
2016-05-20 06:54:15,055 INFO  [Executor task launch worker-1] executor.Executor (Logging.scala:logInfo(58)) - Finished task 0.0 in stage 12.0 (TID 6). 1161 bytes result sent to driver
2016-05-20 06:54:15,055 INFO  [task-result-getter-1] scheduler.TaskSetManager (Logging.scala:logInfo(58)) - Finished task 0.0 in stage 12.0 (TID 6) in 4 ms on localhost (1/1)
2016-05-20 06:54:15,056 INFO  [dag-scheduler-event-loop] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - ResultStage 12 (print at NetworkWordCount.scala:26) finished in 0.004 s
2016-05-20 06:54:15,056 INFO  [task-result-getter-1] scheduler.TaskSchedulerImpl (Logging.scala:logInfo(58)) - Removed TaskSet 12.0, whose tasks have all completed, from pool 
2016-05-20 06:54:15,056 INFO  [streaming-job-executor-0] scheduler.DAGScheduler (Logging.scala:logInfo(58)) - Job 6 finished: print at NetworkWordCount.scala:26, took 0.012564 s

动态的job控制器会根据我们设定的时间间隔收集到数据，让静态的Dstream Graph活起来,而来不断产生job执行。job的具体生成在以后介绍
如果数据处理不过来，就可以限流，Spark Streaming在运行的过程中可以动态地调整自己的资源,具体内容以后介绍

最后编辑于：2017.12.03 05:39:14

人面猴
序言：七十年代末，一起剥皮案震惊了整个滨河市，随后出现的几起案子，更是在滨河造成了极大的恐慌，老刑警刘岩，带你破解...
沈念sama阅读 159,290评论 4赞 363
死咒
序言：滨河连续发生了三起死亡事件，死亡现场离奇诡异，居然都是意外死亡，警方通过查阅死者的电脑和手机，发现死者居然都...
沈念sama阅读 67,399评论 1赞 294
救了他两次的神仙让他今天三更去死
文/潘晓璐我一进店门，熙熙楼的掌柜王于贵愁眉苦脸地迎上来，“玉大人，你说我怎么就摊上这事。” “怎么了？”我有些...
开封第一讲书人阅读 109,021评论 0赞 243
道士缉凶录：失踪的卖姜人
文/不坏的土叔我叫张陵，是天一观的道长。经常有香客问我，道长，这世上最难降的妖魔是什么？我笑而不...
开封第一讲书人阅读 44,034评论 0赞 207
港岛之恋（遗憾婚礼）
正文为了忘掉前任，我火速办了婚礼，结果婚礼上，老公的妹妹穿的比我还像新娘。我一直安慰自己，他们只是感情好，可当我...
茶点故事阅读 52,412评论 3赞 287
恶毒庶女顶嫁案：这布局不是一般人想出来的
文/花漫我一把揭开白布。她就那样静静地躺着，像睡着了一般。火红的嫁衣衬着肌肤如雪。梳的纹丝不乱的头发上，一...
开封第一讲书人阅读 40,651评论 1赞 219
城市分裂传说
那天，我揣着相机与录音，去河边找鬼。笑死，一个胖子当着我的面吹牛，可吹牛的内容都是我干的。我是一名探鬼主播，决...
沈念sama阅读 31,902评论 2赞 313
双鸳鸯连环套：你想象不到人心有多黑
文/苍兰香墨我猛地睁开眼，长吁一口气：“原来是场噩梦啊……” “哼！你这毒妇竟也来了？” 一声冷哼从身侧响起，我...
开封第一讲书人阅读 30,605评论 0赞 199
万荣杀人案实录
序言：老挝万荣一对情侣失踪，失踪者是张志新（化名）和其女友刘颖，没想到半个月后，有当地人在树林里发现了一具尸体，经...
沈念sama阅读 34,339评论 1赞 246
护林员之死
正文独居荒郊野岭守林人离奇死亡，尸身上长有42处带血的脓包…… 初始之章·张勋以下内容为张勋视角年9月15日...
茶点故事阅读 30,586评论 2赞 246
白月光启示录
正文我和宋清朗相恋三年，在试婚纱的时候发现自己被绿了。大学时的朋友给我发了我未婚夫和他白月光在一起吃饭的照片。...
茶点故事阅读 32,076评论 1赞 261
活死人
序言：一个原本活蹦乱跳的男人离奇死亡，死状恐怖，灵堂内的尸体忽然破棺而出，到底是诈尸还是另有隐情，我是刑警宁泽，带...
沈念sama阅读 28,400评论 2赞 253
日本核电站爆炸内幕
正文年R本政府宣布，位于F岛的核电站，受9级特大地震影响，放射性物质发生泄漏。R本人自食恶果不足惜，却给世界环境...
茶点故事阅读 33,060评论 3赞 236
男人毒药：我在死后第九天来索命
文/蒙蒙一、第九天我趴在偏房一处隐蔽的房顶上张望。院中可真热闹，春花似锦、人声如沸。这庄子的主人今日做“春日...
开封第一讲书人阅读 26,083评论 0赞 8
一桩弑父案，背后竟有这般阴谋
文/苍兰香墨我抬头看了看天上的太阳。三九已至，却和暖如春，着一层夹袄步出监牢的瞬间，已是汗流浃背。一阵脚步声响...
开封第一讲书人阅读 26,851评论 0赞 195
情欲美人皮
我被黑心中介骗来泰国打工，没想到刚下飞机就差点儿被人妖公主榨干…… 1. 我叫王不留，地道东北人。一个月前我还...
沈念sama阅读 35,685评论 2赞 274
代替公主和亲
正文我出身青楼，却偏偏与公主长得像，于是被迫代替她去往敌国和亲。传闻我的和亲对象是个残疾皇子，可洞房花烛夜当晚...
茶点故事阅读 35,595评论 2赞 270

2 通过案例对SparkStreaming透彻理解之二

推荐阅读更多精彩内容