RRP-Xen: Experiment Results

This repository comprises of scheduling latency, file transfer latency, Throughput computation experiments

Redis-Cli Experiments

This directory comprises of results of Redis-Cli Benchmarking techniques Domains tested: 1/7, 2/7, 3/7, 4/7 (The experimental setup assumes Redis-Cli software to be installed in the VM that is being evaluated)

Steps:

0. Run a compute intensive Linux command yes in other VMs that are not being evaluated. (For example: If 2 VMs with availability factors 3/7 and 4/7 are operating in RRP-controlled cpupool. Let us suppose VM with availability factor 3/7 is being evaluated, run yes in 4/7 VM.)

1. Run the command in vantage VM: nohup ./redis-cli --intrinsic-latency 100 2>&1 > output.txt &

2. Immediately after the command is run, try to obtain the PID of this process using (ps -ef)

3. Next, use this PID in chrt -v -f -a -p 1 <PID> to change the SCHEDULING POLICY of this process to SCHED_FIFO

4. Try to maintain a remote counter to keep track of 100s ~ 1min 40s time period.

5. Results are stored in output.txt file. To collect the max latency -> avg run result * worst_run gives the max latency of this run

6. Repeat this experiment 3 times for each domain and gather the averages of these 3 runs.

Note: Repeat the process for following domain pairs <3/7, 4/7>, <1/14, 3/7, IDLE>, ...

If unschedulable using RRP-AAF, omit that pair and proceed with the next pair.

HTTPS-Nginx Latency-Throughput Experiments

Software Requirements for Vantage VM: Nginx application web-server, PHP 7.0 Fast-CGI, ngrok.

Software Requirements for Remote Client Machine: WRK2 tool, Python-3 (to run latency_throughput_extractor.py)

Steps:

0. Run a compute intensive Linux command yes in other VMs that are not being evaluated. (For example: If 2 VMs with availability factors 3/7 and 4/7 are operating in RRP-controlled cpupool. Let us suppose VM with availability factor 3/7 is being evaluated, run yes in 4/7 VM.)

1. Run random file generator program to generate a total of 1GB files (899 1KB files, 1000 100KB files, 1000 1MB files) and place all these files in a tmpfs directory where PHP application resides.

2. Replace /etc/nginx/nginx.conf with the nginx.conf provided in this repo.

3. Run sudo nginx -s reload and sudo nginx -t to confirm Nginx application web server is up and running.

4. Run sudo systemctl restart php7.0-fpm to activate PHP 7.0 fastCGI gateway.

5. Set the priority of nginx worker and master processes from SCHED_OTHER to SCHED_FIFO.

6. Run ngrok and copy the https address.

7. In the client machine, Now open http_client.lua and change ip address in line 82 from the above copied addr.

8. In the client machine, run script_http_automate.sh, the ngrok on Vantage VM should now start serving the file-size requests. This can be confirmed by 200 OK status.

9. If the status is 502 Bad Gateway, it implies something is wrong in the experimental setup.

10. On the client machine, the results are gathered in results directory.