SparkR的第一个测试例子Spark Pi计算

安装SparkR颇费周折，网上看到的各种安装方法，其实最终测试都很不好用。可能是国内有些网站被屏蔽的关系吧。

如install_github("amplab-extras/SparkR-pkg", subdir="pkg")，这条命令，就卡在SBT的环节，即使下载了SBT进行安装也是不行。其错误报码是：“Invalid or corrupt jarfile sbt/sbt-launch-0.13.6.jar”

单独在Spark源码下编译也是无法成功，虽然利用R -e "devtools::install('.')"这样的命令能够生成SparkR的包，但是进行测试的时候，发现网络连接没通过，其核心原因还是因为没有生成sparkr-assembly-0.1.jar，缺少这个当然所有都无法进行联调。

编译完成之后，利用RStudio进行了第一个测试，sparkR进行Spark Pi测试，测试通过，非常开心。

欢迎交流学习. Email : sparkexpert@sina.com

这是在SparkR中输出的结果。

library(SparkR)

[SparkR] Initializing with classpath /home/ndscbigdata/R/x86_64-pc-linux-gnu-library/3.2/SparkR/sparkr-assembly-0.1.jar

>

> #args <- commandArgs(trailing = TRUE)

>

> #if (length(args) < 1) {

> # print("Usage: pi <master> [<slices>]")

> # q("no")

> #}

>

> #sc <- sparkR.init(master="spark://ubuntu-bigdata-5:7077","PiR");

> sc <- sparkR.init(master="local", "PiR")

Launching java with command /usr/lib/jvm/java-8-oracle/bin/java -Xmx512m -cp '/home/ndscbigdata/R/x86_64-pc-linux-gnu-library/3.2/SparkR/sparkr-assembly-0.1.jar:' edu.berkeley.cs.amplab.sparkr.SparkRBackend /tmp/RtmpGq7K9F/backend_port4ae6710a585b

15/10/09 09:31:27 INFO Slf4jLogger: Slf4jLogger started

>

> slices <- ifelse(length(args) > 1, as.integer(args[[2]]), 2)

>

> n <- 100000 * slices

>

> piFunc <- function(elem) {

+ rands <- runif(n = 2, min = -1, max = 1)

+ val <- ifelse((rands[1]^2 + rands[2]^2) < 1, 1.0, 0.0)

+ val

+ }

>

>

> piFuncVec <- function(elems) {

+ message(length(elems))

+ rands1 <- runif(n = length(elems), min = -1, max = 1)

+ rands2 <- runif(n = length(elems), min = -1, max = 1)

+ val <- ifelse((rands1^2 + rands2^2) < 1, 1.0, 0.0)

+ sum(val)

+ }

>

> rdd <- parallelize(sc, 1:n, slices)

> count <- reduce(lapplyPartition(rdd, piFuncVec), sum)

15/10/09 09:31:28 WARN TaskSetManager: Stage 0 contains a task of very large size (391 KB). The maximum recommended task size is 100 KB.

100000

15/10/09 09:31:29 INFO RRDD: Times: boot = 0.768 s, init = 0.003 s, broadcast = 0.000 s, read-input = 0.001 s, compute = 0.066 s, write-output = 0.000 s, total = 0.838 s

100000

15/10/09 09:31:29 INFO RRDD: Times: boot = 0.004 s, init = 0.002 s, broadcast = 0.000 s, read-input = 0.001 s, compute = 0.062 s, write-output = 0.000 s, total = 0.069 s

> cat("Pi is roughly", 4.0 * count / n, "\n")

Pi is roughly 3.14792

> cat("Num elements in RDD ", count(rdd), "\n")

15/10/09 09:31:29 WARN TaskSetManager: Stage 1 contains a task of very large size (391 KB). The maximum recommended task size is 100 KB.

15/10/09 09:31:29 INFO RRDD: Times: boot = 0.005 s, init = 0.002 s, broadcast = 0.000 s, read-input = 0.001 s, compute = 0.000 s, write-output = 0.000 s, total = 0.008 s

15/10/09 09:31:29 INFO RRDD: Times: boot = 0.004 s, init = 0.002 s, broadcast = 0.000 s, read-input = 0.001 s, compute = 0.000 s, write-output = 0.000 s, total = 0.007 s

Num elements in RDD 200000

15/10/09 09:31:29 WARN TaskSetManager: Stage 2 contains a task of very large size (391 KB). The maximum recommended task size is 100 KB.

15/10/09 09:31:29 INFO RRDD: Times: boot = 0.004 s, init = 0.002 s, broadcast = 0.000 s, read-input = 0.001 s, compute = 0.000 s, write-output = 0.000 s, total = 0.007 s

15/10/09 09:31:29 INFO RRDD: Times: boot = 0.006 s, init = 0.002 s, broadcast = 0.000 s, read-input = 0.001 s, compute = 0.000 s, write-output = 0.000 s, total = 0.009 s

15/10/09 09:31:29 WARN TaskSetManager: Stage 3 contains a task of very large size (391 KB). The maximum recommended task size is 100 KB.

15/10/09 09:31:29 INFO RRDD: Times: boot = 0.004 s, init = 0.001 s, broadcast = 0.000 s, read-input = 0.001 s, compute = 0.000 s, write-output = 0.000 s, total = 0.006 s

15/10/09 09:31:29 INFO RRDD: Times: boot = 0.004 s, init = 0.002 s, broadcast = 0.000 s, read-input = 0.001 s, compute = 0.000 s, write-output = 0.001 s, total = 0.008 s