This Repository Contains the implementation of a research prototype for Caerus. It builds on the Calvin opensource code at https://github.com/kunrenyale/CalvinDB Currently, we do not support the Calvin and SLOG sequencers; however, we hope to change this. REQUIREMENTS: Caerus has been tested on Ubuntu 20.04. By default, Caerus binds threads to cpus 0-23. Furthermore, sufficient system memory is required as Caerus is an in-memory database; we recommend at least 64GB. Caerus requires packages: unzip g++ cmake libboost-all-dev libevent-dev libdouble-conversion-dev libgoogle-glog-dev libgflags-dev libiberty-dev libreadline-dev liblz4-dev liblzma-dev libsnappy-dev make zlib1g-dev binutils-dev libjemalloc-dev libssl-dev pkg-config libunwind-dev libtool protobuf gflags zeromq glog install-ext will install all of these for you INSTALL INSTRUCTIONS: #install required libraries ./install_ext #setup environment so code will compile source setup #complile code cd src make -j cd ../ # to run a simple a single node locally # NOTE: It is not suggested unless you have a powerful machine WORKLOADS: Caerus supports two workloads, TPC-C and MovR. We would like to support more in the future. --experiment controls which workload will be run Caerus has a built-in load generator; the option --batch_size will dictate the load on the system. Each machine will submit 200 * batch_size transactions per second to the system --locality_size will group regions into locality groups; this defaults to no locality groups TPC-C: --percent_mp controls the percent of multi-warehouse transactions --percent_mr controls the percent of multi-warehouse transactions that are multi-region --hot_records controls the number of warehouses Example: ./bin/scripts/caerus_server --machine_id=0 --experiment="TPC-C" --config=local.conf --max_batch_size=10 --percent_mr=10 --percent_mp=0 --hot_records=240 ./bin/scripts/caerus_server --machine_id=1 --experiment="TPC-C" --config=local.conf --max_batch_size=10 --percent_mr=10 --percent_mp=0 --hot_records=240 Will run TPC-C on Caerus with two replicas and a single partition on the local machine MovR: --percent_mp controls the percent of multi-partition transactions --percent_mr controls the percent of multi-region transactions. When locality_size is set 1/3 of multi-region transactions will be outside the locality group, while 2/3 will be within the locality group --hot_records controls the number of vehicles Example: ./bin/scripts/caerus_server --machine_id=0 --experiment="MovR" --config=local.conf --max_batch_size=10 --percent_mr=10 --percent_mp=0 --hot_records=240 ./bin/scripts/caerus_server --machine_id=1 --experiment="MovR" --config=local.conf --max_batch_size=10 --percent_mr=10 --percent_mp=0 --hot_records=240 Will run MovR on Caerus with two replicas and a single partition on the local machine DEPLOY TO A CLUSTER cluster is a script that will deploy Caerus across a cluster of machines specified in the config file cluster.conf cluster.conf contains the machine information for the cluster. Each line contains the information for a single machine with the format <machine_id>:<replica>:<address>:<port>. #creates the directory specified in --path and pushes the contents of src to all machines ./bin/scripts/cluster --command="push-code" --path="~/caerus" --username=<username> --config="cluster.conf" #copies the config file to all machines ./bin/scripts/cluster --command="put-config" --path="~/caerus" --username=<username> --config="cluster.conf" #installs all required libraries at each machine ./bin/scripts/cluster --command="install-ext" --path="~/caerus" --username=<username> --config="cluster.conf" #compile the code on each machine ./bin/scripts/cluster --command="make" --path="~/caerus" --username=<username> --config="cluster.conf" #runs Caerus across the cluster as specified by the workload parameters --experiment, --hot_records, --max_batch_size, --percent_mr, --percent_mp, --locality_size ./bin/scripts/cluster --command="start" --path="~/caerus" --type=1 --username=<username> --config="cluster.conf" --experiment="MovR" --hot_records=10000 --max_batch_size=10 --percent_mr=10 --percent_mp=0 #kills all running Caerus instances ./bin/scripts/cluster --command="kill" --username=<username> --config="cluster.conf"