The Write Test

##Introduction This benchmark is to test the write performance of Tachyon. It will setup a set of TachyonFS and write data to Tachyon concurrently.

When running, each node starts several threads and each thread contains a Tachyon client. The client will write a list of files to the node's workspace. Note that each node has its own workspace so that the threads of different nodes will not write to the same directory.

User can define the file length to write to Tachyon, and we use the block size in the UserConf of Tachyon.

Note that each thread writes its own files so that there is no cross write operation.

##Configuration The configurations of this benchmark is in conf/testSuite/Write.xml.

Property Name	Meaning	Default Value
threads.num	the threads num of write test	4
files.per.thread	the number of files to write for each write thread	6
file.length.bytes	the file size of write test, in bytes	134217728
grain.bytes	the size of the buffer which is written to Tachyon once	4194304
write.type	the WriteType of the write operation	THROUGH

##Result Demo This benchmark is supported to generate a total report by the tool TachyonPerfCollector. After finished a Write Test, you can generate a total report with the command ./bin/tachyon-perf-collect Write. Here is a demo of the total report:

Write Test - ID : 1409646262537
Finished Successfully
********** System Configuratiom **********
NodeName	Cores	WorkerMemory
localhost	16	6.0GB
Total	16	6.0GB
********** Write Test Settings **********
tachyon.perf.tfs.address	tachyon://localhost:19998
file.length.bytes	134217728
files.per.thread	6
grain.bytes	4194304
threads.num	4
WRITE_TYPE	THROUGH
********** Write Throughput **********
NodeName	WriteThroughput(MB/s)
localhost	90.86506
Total	90.86506
********** Node Details **********
localhost's throughput(MB/s) for each threads:
	[ 21.067068 ][ 21.435152 ][ 24.21949 ][ 24.14335 ]

In the total report, it will list the related system environment, the configurations and the performance result. In the Write Test, we focus on the write throughput and we show the result in MB/s.