Faione / tailbench-in-container

build and run tailbench in container

tailbench-in-container

build and run tailbench in container, inspired by briankoco/tailbench-docker

Component

• TailBench v0.9: apps and tools's source code and dependencies
• TailBench Datasets: realword workload dataset
• Dockerfile: build a compiling env container on centos
• docker-compose.yaml: run a building env container, build if image not found

after download all dependencies, the dir must be like follow:

├── docker-compose.yaml
├── Dockerfile
├── tailbench.inputs
└── tailbench-v0.9

TailBench

you can download tailbench and datasets from here, but there may exist some bugs in tailbench, so you can clone from a fixed repo by:

$ git submodule update --init 

I have made some configs for building and running in container on branch nochrt, if you want to use the raw config just run:

$ cd tailbench-v0.9
$ git checkout main

Container Env

if you have docker-compose installed, you can to build and run env container and mount tailbench onto it just by

# add `--build` to build image locally,
# `-d` means running backend
$ docker-compose up -d

# enter into container
$ docker exec -it tailbenchd-tailbenchenv-1 /bin/bash

but if you want to start step by step, you do as follow

$ make image

this will build a image base on centos7, and install nessessary packages as follow:

make automake file \
gcc gcc-c++ \
java-1.8.0-openjdk java-1.8.0-openjdk-devel \
gperftools google-perftools \
libtool bison autoconf numpy scipy swig ant \
zlib-devel libuuid-devel opencv-devel jemalloc-devel numactl-devel \
libdb-cxx-devel libaio-devel openssl-devel readline-devel \
libgtop2-devel glib-devel boost-devel\
python python-devel python-pip

after all thinges done, you can run a container and start compling

• tailbench v0.9
• tailbench.inputs
• tailbenchenv:centos7

# flag `rm` will destroy container while existing, remove if you want to keep container remaining
$ docker run -it --rm -v ${PWD}/tailbench-v0.9:/src/tailbench -v ${PWD}/tailbench.inputs:/src/dataset faione/tailbenchenv:centos7 /bin/bash

Build and Run

tailbench source code will mount at /src/tailbench, you can build by BUILD-INSTRUCTIONS

$ cd /src/tailbench

$ ./build.sh

each app has two scripts to run bench mark, to run all:

$ cd /src/tailbench

# sleep for bench to exit
$ for app in img-dnn masstree shore specjbb moses silo sphinx xapian;do \
  echo ${app}; \
  cd ${app};\
  ./run_networked.sh; \
  sleep 10; \
  cd -; \
  done;

check the result by:

$ for app in img-dnn masstree specjbb moses silo sphinx xapian;do \
  echo ${app}; \
  python utilities/parselats.py ${app}/lats.bin 95; \
  done;