#Instruction

##Download code:

$ git clone https://github.com/NerdonblooR/ambry.git

Most of the code for compaction and compaction scheduling is in:
ambry/ambry-store/src/main/java/com.github.ambry.store/BlobStore.java

##Build Ambry:

$ cd ambry
$ ./gradlew allJar
$ mkdir logs

##Set up metrics dump directory: $ mkdir /tmp/metrics

##Run tests with our scripts:

$ cd ambry/compaction-test
$ python highAccessRate.py -s 0 -t 2 -w 5 -j 1000 --hardwareLayoutFilePath "./demo/HardwareLayout.json" --partitionLayoutFilePath "./demo/PartitionLayout.json" --serverPropertiesPath "./testConfig/config" --bigFileNum 50 --midFileNum 50 --smallFileNum 0 --tinyFileNum 0 --metricPath "/tmp/metrics" --resultPath "./testResult/1" --partitionSize 1024

The script will start Ambry frontend component and backend component, conduct tests and collect performance metrics. When script stops, you can check ambry/compaction-test/testResult/1 for collected performance data

ambry/compaction-test/testResult/1 will contain:
-csv files contain performance metrics for each partition
-server.log contains debug info about compaction code:
check compaction elapsed time via:
$ cat server.log | grep Compaction
-settings for the test: HardwareLayout.json, PartitionLayout.json, config

If the script is running or does not stop normally, you can kill all the background processes created by the script via:
$ cd ambry/compaction-test
$ kill -9 `cat save_pid.txt`

##Explaination for script parameters:
-w <worker threds number per process>
-j <number of requests the main process should fetch before it stops>
-s <1: enable shuffle partition pools, 0: disable shuffle partition pools>
--hardwareLayoutFilePath <hardwareConfigFile>
--partitionLayoutFilePath <partitionConfigFile>
--serverPropertiesPath <file contains the scheduling thresholds>
--bigFileNum <the number of big file a process need to load into ambry before test phase>
--midFileNum <the number of medium file a process need to load into ambry before test phase>
--smallFileNum <the number of small file a process need to load into ambry before test phase>
--tinyFileNum <the number of tiny file a process need to load into ambry before test phase>
--metricPath <must set to /tmp/metrics>
--resultPath <dir to store the test results when test stop>
--partitionSize <size in MB>

Note that: -Based on provided hardware layout file and partition layout file, Ambry will create 3 partitions under /tmp/0, /tmp/1 and /tmp/2.

Please also refer to: https://github.com/linkedin/ambry/wiki/Quick%20Start if you want to interact with Ambry without our script