SLOG

SLOG stands for serializable, low-latency, geo-replicated transactions.

SLOG is a geographically distributed data store that achieves high-throughput and low-latency transactions while guaranteeing strict serializability. SLOG achieves high-throughput, strictly serializable ACID transactions at geo-replicated distance and scale for all transactions submitted across the world while achieving low latency for transactions that initiate from a location close to the home region for data they access.

Research Paper

Install Dependencies

Note: All steps mentioned are suitable for Ubuntu 20.04 LTS

Install gcc,cmake and other essential libraries

./prev-install.sh

Install dependencies for the code and building code

Clone this repo and Run in root directory of repository

./install-deps.sh

Build the code

mkdir build
cd build
cmake ..
make -j$(nproc)

Install dependencies for admin tool.

sudo apt-get install python3-pip -y
pip3 install -r tools/requirements.txt

Install docker

Follow Docker engine install on Ubuntu. Then, configure docker to run without Sudo using Link

Build docker image and push the image to a remote repository:

https://hub.docker.com/repository/docker/ctring/slog

docker build . -t manas11/manas
docker login
# check image id using 
docker images
docker tag <image-id> manas11/manas:latest

docker push manas11/manas

# we pushed our docker image to
https://hub.docker.com/r/manas11/manas

Create local cluster:

Use the file present in examples/cluster.conf as template for creating a local cluster

python3 tools/admin.py local --start --image manas11/manas /home/ubuntu/SLOG/examples/cluster.conf

This creates 4 docker containers with SLOG running, now these containers are ready to exchange data.

Create dataset

python3 tools/gen_data.py -c /var/tmp/slog.conf -s 100 -su mb /var/tmp

Run benchmark tests

The benchmarking tool needs to know which replica key is mastered and which partition it belongs to. To that end, we use the gen_data tool to deterministically generate data across the cluster as well as make a copy on the local machine to be used by the benchmarking tool. Running a benchmark basic workload of 100,500,2000 transactions with a 50% probability of multi-home transactions to study the performance of SLOG. Different kinds of workload can be used with the benchmarking tool. For currently available workloads, source code is mentioned in the benchmark.cpp file present in repository. The parameters of a workload can be found near the beginning of its source code in the workload directory.

build/benchmark -data_dir /var/tmp -config /var/tmp/slog.conf -num_txns=1000 -wl=basic -params="mh=50"

manas11/SLOG