Spark缓存策略

当对同一个rdd多次执行action时,如果在磁盘上则每次执行action都会从磁盘将数据加载,如果将其缓存到内存中会提高再次action的读取速度,Spark缓存主要有cache()和persist()两种,当缓存一个rdd时,每一个节点上都会存放这个rdd的partition,当要使用rdd的时候可以直接从内存读出。
cache源码:
def cache(self):
        """
        Persist this RDD with the default storage level (C{MEMORY_ONLY}).
        """
        self.is_cached = True
        self.persist(StorageLevel.MEMORY_ONLY)
        return self

从源码可以看出,cache底层调用的是persist方法,传入的参数是:StorageLevel.MEMORY_ONLY,再看persist()方法:

def persist(self, storageLevel=StorageLevel.MEMORY_ONLY):
        self.is_cached = True
        javaStorageLevel = self.ctx._getJavaStorageLevel(storageLevel)
        self._jrdd.persist(javaStorageLevel)
        return self

persist方法,传入的参数是StorageLevel,从StorageLevel的源码可以看出它的值总共有6种,因此persist()相比cache()在缓存形式上更为丰富,不仅支持内存的方式,还支持内存和磁盘、内存副本等方式。

StorageLevel.DISK_ONLY = StorageLevel(True, False, False, False)
StorageLevel.DISK_ONLY_2 = StorageLevel(True, False, False, False, 2)
StorageLevel.MEMORY_ONLY = StorageLevel(False, True, False, False)
StorageLevel.MEMORY_ONLY_2 = StorageLevel(False, True, False, False, 2)
StorageLevel.MEMORY_AND_DISK = StorageLevel(True, True, False, False)
StorageLevel.MEMORY_AND_DISK_2 = StorageLevel(True, True, False, False, 2)
StorageLevel.OFF_HEAP = StorageLevel(True, True, True, False, 1)

持久化到内存和直接从磁盘读取时间对比:

import os
import time
from pyspark import SparkContext, SparkConf

conf = SparkConf()
sc = SparkContext(conf=conf)

current_dir = os.path.dirname(os.path.realpath(__file__))
file_path = "{}/name_age.txt".format(current_dir)


def cached():
    start_time = time.time()
    text_rdd = sc.textFile("file://{}".format(file_path)).cache()
    text_rdd.count()
    text_rdd.count()
    end_time = time.time()
    print("{}:{}".format("first cache", end_time - start_time))

    start1_time = time.time()
    text1_rdd = sc.textFile("file://{}".format(file_path)).cache()
    text1_rdd.count()
    text1_rdd.count()
    end1_time = time.time()
    print("{}:{}".format("second cache", end1_time - start1_time))


def uncached():
    start_time = time.time()
    text_rdd = sc.textFile("file://{}".format(file_path))
    text_rdd.count()
    text_rdd.count()
    end_time = time.time()
    print("{}:{}".format("first uncache", end_time - start_time))

    start1_time = time.time()
    text1_rdd = sc.textFile("file://{}".format(file_path))
    text1_rdd.count()
    text1_rdd.count()
    end1_time = time.time()
    print("{}:{}".format("second uncache", end1_time - start1_time))


sc.stop()



执行cached()结果:
first cache:1.7104301452636719                                                  
second cache:0.2717571258544922


执行uncached()结果:
first uncache:1.4453039169311523                                                
second uncache:0.49161386489868164

从执行结果可以看出,当第二次执行rdd.count()时,有cache情况下是0.2717571258544922;无cache情况下是0.49161386489868164,由于我的内存空间不足,所以不太明显,当数据量大且内存充足的时候,持久化到内存的效率会远远高于磁盘。

对pyspark有兴趣的小伙伴可以关注我的github,spark for python 持续更新

原文地址:https://www.cnblogs.com/FG123/p/9748772.html