Copy the herd-build/payload.zip to the present directory.
For example, if you run the experiments from your laptop, you will need to scp the payload from the deployment machine where you built the binaries.
First, ensure that hugepages are enabled, as the are required by HERD. In the deployment machines run the following command:
# `node0` refers to the first NUMA node. Make sure to adapt this to your setup.
# Select the NUMA node closer to the RDMA NIC.
echo 8192 | sudo tee /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepagesThem, simply run
./herd_stock.shFind the generated information under logs/stock_*/m1/logs/workers.txt.
Note: Edit
herd_stock.shto run various configurations.
Simply run
./herd_isactive.shFind the generated information under logs/isactive_*/m4/logs/workers.txt.
The reduction in throughput is uKharon's overhead.
Note: Edit
herd_isactive.shto run various configurations.